> Do we have evidence that Musk has been fatigued and felt worthless and just wanted to lie around in bed and not cared about Mars or anything for two straight weeks?
"Devastated by the breakup with Amber Heard and the news that his father had a child with the woman he had raised as his stepdaughter, Musk went through periods when he oscillated between depression, stupor, giddiness, and manic energy. He would fall into foul moods that led to almost catatonic trances and depressive paralysis. Then, as if a switch flipped, he would become giddy and replay old Monty Python skits of silly walks and wacky debates, breaking into his stuttering laugh. Professionally and emotionally, the summer of 2017 through the fall of 2018 would be the most hellacious period of his life, even worse than the crises of 2008. “That was the time of most concentrated pain I’ve ever had,” he says. “Eighteen months of unrelenting insanity. It was mind-bogglingly painful.”
At one point in late 2017, he was scheduled to be on a Tesla earnings call with Wall Street analysts. Jon McNeill, who was then Tesla’s president, found him lying on the floor of the conference room with the lights off. McNeill went over and lay down next to him in the corner. “Hey, pal,” McNeill said. “We’ve got an earnings call to do.”
“I can’t do it,” Musk said.
“You have to,” McNeill replied.
It took McNeill a half-hour to get him moving. “He came from a comatose state to a place where we could actually get him in the chair, get other people in the room, get him through his opening statement, and then cover for him,” McNeill recalls. Once it was over, Musk said, “I’ve got to lay down, I’ve got to shut off the lights. I just need some time alone.” McNeill said the same scene played out five or six times, including once when he had to lie on the conference room floor next to Musk to get his approval for a new website design.
Around that time, Musk was asked by a user on Twitter if he was bipolar. “Yeah,” he answered. But he added that he had not been medically diagnosed. “Bad feelings correlate to bad events, so maybe the real problem is getting carried away for what I sign up for.” One day, when they were sitting in the Tesla conference room after one of Musk’s spells, McNeill asked him directly whether he was bipolar. When Musk said probably yes, McNeill pushed his chair back from the table and turned to talk to Musk eye to eye. “Look, I have a relative who is bipolar,” McNeill said. “I’ve had close experience with this. If you get good treatment and your meds dialed right, you can get back to who you are. The world needs you.” It was a healthy conversation, McNeill says, and Musk seemed to have a clear desire to get out of his messed-up headspace.
But it didn’t happen. His way of dealing with his mental problems, he says when I ask, “is just take the pain and make sure you really care about what you’re doing.”"
I'm surprised that nobody has raised the possibility of a personality disorder, perhaps some version of borderline personality.
Some reasons to think in this direction:
1. Musk may have cycles, but they are very rapid. He can flip on a switch. Borderline is/was often misdiagnosed as Bipolar, but these are weeks+ cycles. I've not seen evidence, for example, that Musk has ever had a six month+ major depressive episode. At the same time, diagnosing Borderlines as Bipolar is common, in part because it's a more glamorous disorder.
2. The list of symptoms of borderline matches pretty well, including "Unstable and chaotic interpersonal relationships"; "Impulsive or reckless behaviors"; "Chronic feelings of emptiness"; "Inappropriate, intense anger that can be difficult to control"; "Transient, stress-related paranoid or severe dissociative symptoms" (see, e.g., Twitter behavior) https://en.wikipedia.org/wiki/Borderline_personality_disorder#Signs_and_symptoms
3. Pretty extreme forms of child abuse and abandonment — both high risk factors for borderline. Although the most significant risk factor, child sexual abuse, is absent, it is notable that his truly awful father had leanings in that direction.
The main difference, of course, is that Musk is extremely high-functioning. Another difference is that Borderline symptoms tend to subside in later life, but Musk shows no sign of regularizing his behavior.
My impression is that people with BPD are very concerned with abandonment and cycle in their relationships between trying to be close and provoking conflicts due to hypersensitive perception that the other person is critical or unloving. They are manipulative in a desperate way (not devious like a psychopath) for example threatening to commit suicide to get other people to take care of them. They often feel terrible and have difficulty regulating emotions so they are prone to take drugs or engage in activities like cutting. This doesn’t match Elon Musk very well.
1. @gwern: Appeciate your comment but this is completely the opposite of how he behaves in public and has run Twitter. How is constantly tweeting all day instead of focusing on his long term vision a evidence of a great undisturbed undivided focus and vision? He keeps putting out these I work 18-20 hours a day, sleep on the Twitter floors, sleep on the factor floors PR stuff and people keep lapping it up (I am sure he has done that occassionally but not nearly to the level of his PR would suggest.)
Why is the person who is visibly constantly running into expensive rakes is the "most focused person in the world"?. He foolishingly bid for Twitter at $44B, then tried to get out the deal repeatedly, tried to lie to the Delaware chancellery courts which would have been a losing proposition, was then put into a position to buy Twitter at a wildly overvalued $44B price and after having bought Twitter, then lied to the world that he did it for noble free speech and public square platform reasons.
This is how he normally operates. Him stepping on rakes and him tagging along without fatal blows is considered evidence of his unrelenting focus rather than the obvious occam's razor observation that Most Business Mistakes are not Fatal as Arnold Kling put it (https://arnoldkling.substack.com/p/most-business-mistakes-are-not-fatal)
2. @SSC original post:
No, the user experience is not the same on X after firing 80-90% of the staff. It is noticably worse.
What is the evidence that he is running Twitter well? Also a social media company is not measured purely on "user experience". There is no moat for social media companies other than network effects. When you bring your company into serious disrupte you risk your company's moat so company is not on firm reputational and network effects ground as it was pre-Elon. He has made Twitter not "cool" anymore.
Twitter is by Elon own admission down by 90% in value, so how is it "doing well" and how could SSC possibly not mention this or even factor it into SSC's prediction that X will be a "success"? Also what is success? Is it just not folding or becoming the $3T company his cult has been saying it has the potential to be which would be a richer company than Apple (and yes, they bandy about that $3T figure. Just search on YouTube and you'll find a lot of videos with that exact figure).
Also there is nothing in the post about the dozens of lawsuits and the downright shameful non-payment of bills to vendors, non-payment of rents at various of its locations. The richest man in the world is stiffing vendors and landlords and this doesn't even get a mention in the post about how Elon is running Twitter? He is on track to pay a lot of fines for stiffing vendors and ex-X employees. That is not is X "doing well".
Would have liked more discussion on if and how $TSLA is overvalued which is his primary claim to the title of the richest man on Earth
Thanks. I made a mistake. You do not in fact find a lot of claims with that $3T figure. I confabulated that figure from the news that Apple crossed the $3T market cap news I had read. The figure the Musk cults/fans bandy about is a Trillion dollars. ("Twitter will become a trillion dollar company" is the popular claim).
> Appeciate your comment but this is completely the opposite of how he behaves in public and has run Twitter. How is constantly tweeting an evidence of a great undisturbed undivided focus and vision?
I don't understand your point. Yes, of course physical catatonia is the opposite of 'how he behaves in public'; when you have catatonia, how are you going out in public? (The Tesla CEO is practically physically dragging him into the earnings call in that anecdote.) That's my point. Objections based on 'well, I don't see him during his public performances being depressed!' are self-refuting.
As for 'constantly tweeting', how do you know? Tweets are extremely easy, can be handled by assistants, and he periodically announces he's going to stop tweeting for a while - not that you're exactly firing up data analytics software to try to infer his latent cycles from timing during day and overall amount such that you would observe any absences... (Probably a good project for someone, let me know if anyone does it.)
> He keeps putting out these I work 18-20 hours a day, sleep on the Twitter floors, sleep on the factor floors PR stuff and people keep lapping it up (I am sure he has done that occassionally but not nearly to the level of his PR would suggest.)
I'm sure he does do all that quite a lot, although I don't know what level his PR suggests in any falsifiable quantitative sense... Those are easy when you're in a hypomanic phase, and so much less impressive than they look. You only slept 4 hours last night, Musk? Yeah, you and every other bipolar entrepreneur, big deal. (I spent an hour or two today reading through accounts on Hacker News today of bipolar techie types, usually with formal diagnoses, to check my impression, and yeah, Musk looks even more typical in like of them - remarkable only in quantity, not quality.)
"He would fall into foul moods that led to almost catatonic trances and depressive paralysis. Then, as if a switch flipped, he would become giddy and replay old Monty Python skits of silly walks and wacky debates, breaking into his stuttering laugh."
This isn't how I expect bipolar to work. There is no "switch flipping" (except very occasionally when a manic episode follows directly after a depressive one). A patient will be depressed for weeks or months, then gradually come out of it, and after weeks or months of coming out of it, get back to normal. Being "moody" in the sense of having mood swings is kind of the opposite of bipolar; I would associate it more with borderline or PTSD.
Throughout late 2017, Musk was continuing to sound as "manic" as ever - for example, committing to build hyperloops from NYC to DC and saying the government had given approval (it hadn't).
I am absolutely willing to believe Musk has extremely bad times, mood swings, and horrible lows, but you are going to have to do something other than keep repeating that if you want me to believe it's bipolar.
But isn't it starting to sound like a mood-based pathology? At the point that someone can be fairly described as having "hyperthymic temperament, hypomanic signs, and moderate to severe MDD" and where they present with pretty severe depressive symptoms (based on the excerpt gwern provided, although I want to note that if you're rousable with a little encouragement you might be in the vicinity of catatonia but you aren't there yet) you (well, I) tend to think the same sort of thing has gone wrong for them that goes wrong in bipolar. Akin to the idea that some instances of MDD have more in common with BD than other forms of depression, etiologically speaking--are "on the bipolar spectrum," but the pt doesn't manifest enough manic symptoms to meet the dx threshold of hypomania (I've seen this--their depressive symptoms are more severe and melancholic, they have bipolar-like temporal patterns of wellness and unwellness, they have assorted soft signs, they don't respond well to antidepressent monotherapy, but they don't have proper hypomanic episodes, or they are too short and infrequent to catch). Musk could have such a high mood baseline (which would be uncommon, as dysthymic baselines predominate) that his discrete up-episodes are harder to detect, and that his "moments of normality" within the depression look like hypomania because that's his normal. Hypomania is easy enough to miss even when it's a big deviation from the norm. (I also think it's plausible that the rapid switching was brought on by drug use, legal or illegal.) I would also like to see something more pathognomonic of a classic manic episode to (mentally) slap a dx on him, but my suspicions are definitely raised.
Of course the point of the dx is prognostication and treatment, so the real question is--what kinds of pathological behavior would you expect from Musk in the future, and what kinds of treatments might he respond to?
I was actually going to comment in the other direction, on the indefensible grounds that I'm a fairly good...mental health phenotype detector/pattern matcher? And that Musk just doesn't strike me as bipolar, the way that, say, RDJ, Eminem, and Kanye do (all of whom I clocked before their diagnoses were public). But my phenotype-detector is obviously more prone to Type II errors.
ETA like a month later: I realized shortly after I wrote this wall of text that every reported symptom that led to bipolar suspicion can also be explained by stimulant abuse. It's impossible to get a good diagnostic picture while abuse is ongoing, and stimulant abuse makes easy hash of phenomena like quasi-catatonia, wild mood swings, and risk-indifference.
Is there any reason, apart from the classification schemes of 19th century psychiatrists, why mania needs to be accompanied by depression? Can't someone just have the mania without the depression? You can certainly have depression without mania.
What if Musk just has the light side of the force without the dark side?
To someone like me, i.e., not a professional, Musk's behavior seems more like a positive trait (capability to work very hard for extended periods of time) that in some people perhaps is strong and unmanageable enough to turn into bipolar. He's hardly unique in this!
The fact that he was very sad about Amber Heard leaving him (apparently without even a goodbye gift in bed) doesn't really strike me as pathological either.
What's with the obsession of pathologizing everyone and everything? Why can't Elon Musk just be a normal person who gets distraught when his girlfriend abandons him and his father has a semi-incestuous relationship with his stepdaughter?
My pre-1995 SAT scores slightly exceeded Musk’s (Wow! That surprises me), but I don’t think my IQ is 136-140. I was IQ tested as a 9-year-old and had an IQ of 128.
Just for reference, I got a 1480 pre-1995 (IIRC 760 math, 720 verbal) and tested at 136 in high school. At least back then, the verbal part of the SAT was mostly about spending long hours studying words you've never heard before (nice gatekeeping!). The IQ test seemed to be more well-rounded. I nailed some areas like spatial relations (I grew up playing with Tangrams) and repeating a long number backwards, and sucked at some others like story problems. I shudder to think how badly I'd so at some of that stuff now.
"At least back then, the verbal part of the SAT was mostly about spending long hours studying words you've never heard before (nice gatekeeping!)."
It was actually about being a smart person who read a reasonable amount across a variety of fields and learned those words naturally. But I guess some people could or had to game it.
The scores kind of surprised me too: SpaceX engineers were said to be pretty elite, such that I'd expect many of them to score higher (adjusted for the drift in meaning), and we read about Musk impressing them as an engineering polymath. The idea that what's exceptional was the combo of talent and focus/drive, that makes sense, but still I think I'd chalk this up more to the SAT scores being noisy in this range.
Remember that IQ is only moderately good at predicting any particular talent; the example I usually use is that Kasparov has tested IQ 135. He is a "merely" 99th-percentile smart person who happens to be amazing at chess.
Chess success and IQ are only weakly correlated with the strongest correlation being for kids. That's because raw calculations are a small part of a game compared to memorization (especially of opening lines) and pattern recognition. The trend is becoming even stronger with the raise in popularity of faster game formats like rapid (usually 10 minutes/player for an entire game) and bullet (usually 1 minute/player for an entire game) that don't allow a player to calculate for a long time a single move.
What does it matter? If Musk had an IQ of 50, would his accomplishments be any the less impressive? Surely IQ itself is only a proxy, which we care about only because it lets us make (probabilistic) predictions about someones likelihood of succeeding at [given task]. But in this case we don't need proxies, we have thirty years of actual record to judge on.
On the Twitter point, speaking anecdotally from personal experience, I’ve seen a significant amount of followers and correspondents drop off Twitter post-Elon, and it has not slowed down. I’m mostly the film Twitter space, and a lot of folks moved to Mastodon, there were a lot of complaints then about Mastodon, and then people started to migrate to Bluesky, which seems to provoke much less technical misery. Twitter Blue remains intensely unpopular.
Bluesky seems like a weird choice, considering how long Jack Dorsey has been at the tops of Twitter, all the while its shittification kept going on (and also the Trup era !).
But I guess these people are more concerned about Twitter being evil because of Elon, rather than because of being a platform ?
(Also, while Bluesky is a protocol, Jack Dorsey still doesn't particularly seem to be willing to make amends, considering how Bluesky *Social* is proprietary.)
I think that while some folks are leaving out of a personal distaste for Musk, a lot of it is because of the increased chaos and decreased functionality since he took over. Personally, I didn’t bother to make a Bluesky account until this week, when Musk started threatening to charge for Twitter access. The constant arbitrary rule changes, interesting people and entities leaving, and increased bugginess have made Twitter a genuinely worse experience than it was before (which is saying something).
As for why Bluesky is drawing folks, I think largely because the functionality is VERY similar and easy to learn, albeit more rudimentary.
> P.S. the interface is so slow and laggy, holy cow
Maybe this has changed relatively recently in Teslas, as all my experience in them comes in the past 2-3 years, but the idea that somehow the Big 3 carmakers are *ahead* of Tesla in infotainment systems is completely crazy to me. Every Big 3 car I've ever been in has a UI straight from Nokia in 2004 with processing power to match. They are only remotely usable if you turn them into dumb glass by letting your phone take over everything, and even then "slow and laggy" is usually a great description of how they react.
Overall, I appreciated reading the comment, but the whole time I felt like I was reading one of those reviews of Apple products from a decade or so ago comparing how many gigahertz are in the CPU and whether or not it has a serial port to plug in a mouse, when that's not what customers care about *at all*. I don't know what "rigidity" means or what "panel alignment" even is.
Hm, that's possible, but a couple paragraphs above:
> It really makes you question the customer sometimes, because if we put out a touchscreen that failed like that, we'd rightly be ridiculed. CEOs have lost their jobs over far less.
So just substitute that quote for the one I picked if so.
Yeah, ambiguous. My prior is based on the knowledge that the touchscreen failures were hardware-related, and substack's interface is slow and glitchy.
The biggest reason no other automaker had such a huge and beautiful touchscreen was/has been that no such screens were available that were automotive-qualified. Musk seems to have a tendency to ignore such nuisance obstacles (sarcasm), with predictable results.
Fwiw, the "predictable result" was that Tesla had to replace a whole bunch of faulty touch screens, which for our Tesla they did in our driveway at no cost to us, and the upside is that you have a massive screen that's actually big enough for the camera and Google Maps to be useful - smaller screens mean squinting and missing key details
We really need to have an Open Thread about the ACT substack experience. I don't know if it substack or the layer on top to make it more like the old SSC interface. But it is slow! And laggy! I read most everything on an ipad but I can't read ACT on an ipad because it implodes. Even on a PC, ACT lags. Its weird that no one talks about this more.
At this point I have abandoned reading Scott’s Posts on Substack and only read them directly in my Mail app. I use the built in iOS Mail app instead of Gmail because it doesn’t truncate long emails.
I was also confused by the rigidity comment. The "panel alignment" stuff I _think_ refers to thinks like the gaps between the doors and the body etc, which Tesla is pretty famously bad about (or at least, was a few years ago, apparently they have gotten better). Basically the "fit and finish" stuff.
But the rigidity comment I didn't understand. I'm curious what it is important for. It can't be for safety/structural integrity because my understanding is that Teslas outperform other vehicles on tihs by so far that they basically broke the scale when they first came out.
But overall, I think my response to all of that is: Consumers were willing to deal with all of that because everything else about the car was so much better.
I have a new electric car (not a Tesla), and while there are definitely things I would like to improve about it, I don't think I'd ever go back to an ICE vehicle again, and I'd be willing to accept a fair number of tradeoffs, including bad cosmetics/fit and finish, for the advantages I get from an EV.
In other words: reading that comment felt like watching Goodhart's law in action. The engineers at the big company had a ton of metrics they needed to meet around quality etc., but meeting those metrics meant that they couldn't move fast on the far more important things that consumers (or at least one very particular segment of consumers) cared about.
In a fully mature market, yeah, I'd like an EV that didn't have to make all those tradeoffs. But if my only options for getting a vehicle now are "take the tradeoffs or wait 5 years", I'll take the tradeoffs every time.
And yeah, even though I hate all touchscreen interfaces (and, other than overall price, was probably my biggest strike against Teslas when I was shopping), you can't argue that their touchscreen interface isn't better than everyone else's. All of them suck, but Tesla's sucks the least.
Goodhart's Law is a great reference point there. I'm sure the tensile strength of some of those Tesla parts is lower than spec 713(a)(b) for a GM car or whatever, but as a customer, what I'm interfacing with is the vastly-superior Tesla charging port. And I can buy a Tesla on their app today, for the price they say it is, and drive it around charging it at the big, reliable Tesla charger network.
You're also interfacing with the rest of the Tesla, including NVH, rate and handling / ride quality (what body stiffness contributes to, FYI), ergonomics, perceived touch / sound quality of the interfaces, wear (think seat and interior carpet), etc. As well as the lack of buttons, offset gauge information, weird stalks, etc.
In $current_year, the Model 3 is a lot better than the Model Ss of the past, but you're fundamentally paying German OEM prices for a much lesser vehicle with the key killer app being an EV powertrain. It's been interesting seeing who draws their line where over what - and will be very interesting now that the major OEMs are getting serious about EVs.
> You're also interfacing with the rest of the Tesla, including ...
Right, but I think this is where Car People go wrong, thinking that normal customers rank these things anywhere close to where they do. I care vastly more about the size and UI design of the infotainment screen than I do the perceived touch/sound quality of the interfaces. A lot of this stuff veers into mechanical keyboard or automatic watch territory, I think.
This is where things get difficult. As a consumer I have no visibility into things like "body rigidity", component reliability, etc. Teslas look great and test-drive amazing. It's only when you're in a crash 6 years later suddenly these boring automotive qualification issues make a difference.
Well, and the average vehicle on US roads is ~12 years old. When Teslas start hitting the used market in appreciable numbers we'll get to learn a lot about legacy auto requirements.
When you ask customers to rate where they consider these features, they always rate quite low compared to the basics (cost, fuel economy, safety, brand perception/signal). But there's a very sharp cliff - if the turn signal stalk feels like cheap crap (or the blinker sound sounds cheap) during a test drive, or if the door handle feels particularly bad, this can impact a cross-shop. Most cross-shops aren't really genuine (most customers have some idea of which car / brand they want to buy already, and are looking for justification), but 1. don't add fuel to the fire with bad design and 2. surprise and delight with better than expected design.
For the longest time Tesla was somewhat immune to this - if you wanted a particular type of EV, it was them or nothing. We'll see how the next few years pan out as actual competitors enter the sector.
I think that the general direction of this thread ("automakers prioritize many things that are customer-invisible and that in an EV rather than an ICE don't matter to the terminal use-case") is one that makes sense overall but I would definitely *not* say that NVH is only for Car People - if anything, it's the opposite. Car People like hearing the "engine note." I just want to hear my audiobook and/or music.
Goodhart in action again I think. To this normal person (nor a Car Person) the Tesla handles so much better than any ICE car that it's no contest. And it also handles vastly better than the electric cars from the legacy manufacturers (I test drove half a dozen). Maybe in five years time this will be different, but right now, it's no contest.
1: where I live, the difference between charging my vehicle and the price I would be paying for fuel saves me between $100-$130/month. This means that when calculating fuel + monthly payment, I could not have gotten a new ICE vehicle that met my requirements for cheaper than I got the EV, although that probably wouldn't have been the case if I'd been willing to buy used, since the used EV market is basically nonexistent at the moment. And once I'm done paying it off, the savings continue.
2: Relatedly, never having to visit a gas station is very nice. It's a minor thing, but I _never_ have to do it anymore.
3: I haven't owned this vehicle very long, but so far maintenance is considerably easier. Tire rotations are about it. Admittedly this one could change as I've got less than a year of data. _Theoretically_ EVs should have less maintenance, but that doesn't stop bad engineering from creating problems.
I will admit that at least part of my rosy outlook is that this vehicle is a significant upgrade from my previous vehicle which was a beater I bought in grad school that didn't even have power windows or AC. Just about anything I bought would be a significant upgrade, completely aside from ICE/EV status. But having fuel just basically never be a consideration for 95+% of the driving we do is great. And I've even taken it on a couple 10+ hour drives that involved multiple fast charging stops, and haven't really had any issues, so even for road trips, it's not anywhere close to enough of a downgrade from ICE to negate the benefits in day-to-day driving.
I have similar affection for EVs, with one minor caveat- routine maintenance is much easier, but maintenance for things that actually go terribly wrong is much more difficult (and mechanics tend to be out of their depth with EVs, especially in rural areas). You can't fix the stator-rotor by eyeballing it, it's difficult or impossible to replace a single part of the battery pack rather than the whole thing (despite the fact that usually only a small area is the problem child when things start going wrong) and many EVs have much more complex electronics than midline ICE vehicles. However, if you have the money to burn on full part replacement and live in an area with high EV use, these problems mostly go away. And they'll probably go away entirely in the next couple of decades as EVs become the majority of the automobile market.
That's fair. For my specific vehicle, according to estimated replacement costs of the battery, it will take ~6 years of my fuel savings to break even. Even if I only start counting from when the vehicle is done being payed off, the battery is likely to last longer than that unless I get pretty unlucky. Plus I think it comes with a 10 year warranty. So at least for my particular case, even if I have to pay for a replacement battery after 10 years, I'm still coming out ahead. But that's all pretty speculative.
Ah yes, I have recently discussed something like this with an artist with a 380Mm car with regular trips to other continents : no car after ~2002 is an option, because they have not so much electronics, but locked down computers that no mechanic outside of Europe knows how to fix and/or has the proprietary computer suitcase required for diagnostics.
Since EVs are much simpler, I wonder if we'll see again repairable cars that are EVs, sooner than repairable ICEs ?
Unlikely; you need computer-controlled battery management to make EVs practical. Those computers are simple to make, simple to install, and simple to program to brick your car if touched by a mechanic who isn't paying Danegeld to the automaker.
I think this highlights the software vs. hardware concept of quality.
The Big 3 interfaces are ugly, lack functionality, and are slow but they are basically bomb-proof. Unlike the Tesla interface it will almost never go into a reboot loop or have some other major issue that requires a repair. The repair/failure stats are probably on par with other parts of the car, like how you expect 1 in x0,000 transmissions to fail during warranty or whatever. I have a Honda and the interface isn't great but I have done some weird stuff (disconnect battery while running diagnostics, mash random buttons) and it always recovers unlike my other computers.
Tesla however has a nice UI but a higher failure rate and more bugs. However, the bugs can be fixed and many failures can be fixed remotely. Legacy carmakers can't do that so they have to have highly reliable computers and interfaces.
Tesla has software-style high quality, where things are easy and pleasant to use. Big 3 have hardware-style high quality, where absolute failure rates requiring repair or replacement are low.
Additionally - NHTSA will have whatever body part of you they'd like if they find out you have a UI failure that gets anywhere near something they care about. Losing speed indication, for example, is frowned upon with similar intensity as a hood fly-up or loss of propulsion.
How Tesla has managed to outrun this is kind of interesting - their high software churn rate basically outruns the US recall system. A lot of what they change and update is safety-adjacent, but they outrun ODI with what you'd call (at any other OEM) "silent recalls". Whether this is better (debatably faster fixes) or worse (the recall system doesn't hold them correct to account on software issues and the high churn rate may introduce new failure modes) is an exercise for the reader.
My Bolt’s tendency to have infotainment boot-up fail and refuse to reboot itself for ~5 minutes, causing me to not have a backup camera when I turn it on, begs to differ with the idea GM’s software, at least, is bomb proof.
In the Tesla, there’s a steering-wheel button combination to reboot now, which I’ve used a couple times in a few years of ownership. Annoying, but much better than 5 minutes of can’t-even-turn-off-the-radio-which-insists-on-turning-on-when-I-don’t-want-it.
Overall the Bolt’s rate of infotainment-failed is about 5x/month of ownership in my experience.
Yeah, that comment read to me like the analog of a comment from a programmer about how Mac OS is so much worse than Linux. Like, yeah, maybe there are specific metrics on which Teslas are worse, and maybe those metrics are what the legacy automakers have optimized heavily for, but those metrics have roughly zero overlap with what J Random Consumer cares about. [Speaking as a random consumer who bought a model Y six months ago after comparing to half a dozen legacy manufacturers, because it simply blew all the offerings from the legacy manufacturers out of the water. I hadn't intended to buy a Tesla going in, but in the `head to head comparison' it was simply a walkover].
I'm late to the party, but re. The Boring Company, it's the most-visionary and most-important of his projects. I've been advocating just such a project for, I dunno, 20 years now.
Boring tunnels underground, evacuating them of air, and running vehicles through them, is the only way to achieve hypersonic travel on Earth. The theoretical limits on speed then are just the ability of the human body to tolerate g-forces. It's the obvious, necessary, and /final/ transportation technology for Earth; it can never be surpassed by anything short of teleportation.
Also, the first person to build a trans-American hypersonic tunnel system gets to choose where to build the transfer points. Buy a bunch of prairie in Kansas and build your west coast / east coast tunnel interchange there. A megalopolis will spring up over the next decade, and you'll have made up to a trillion dollar profit on those land purchases.
Entirely ignoring the physical problems -- the economic aspects of a hypothetical "hypersonic commute" ought to terrify people. In particular, the same people currently willing to pay half or more of their wage for the privilege of living in SV. Suppose the vacuum tunnel is built, now you can pay that same half+ of your wage for tickets on this train. (Not to mention, tickets to the late Concorde are likely to look like a bargain in comparison.) And now you get to "enjoy" all of the down-sides of space travel, right here on Earth: e.g. microscopic rock falls from the tunnel, now you're perforated as if from a meteorite strike, and breathing vacuum.
Why is the idea of moving meat every day hundreds (or, as contemplated, thousands) of miles so that it can drink bad coffee, watch Powerpoints, and push buttons, still seen as anything but insanity?
Because fundamentally, people want to be in places with other people, and transporting meat quickly is the only alternative to packing all those people in one place?
The massive rents people accept for SV should be a sign that we really could benefit from better transportation.
Much of the time the people "want" to be in places with others simply because this is the only way to get paid. There are alternatives to moving them in the flesh, but for so long as nothing (e.g. COVID) is forcing the issue, employers would rather force the meat to move at its own expense, or fight in bidding wars for cockroach flats.
Picture if employers were required to justify in-person presence requirements for office work to the same standard that they are currently required to do for IQ testing. (i.e. to an effectively-impossible standard, in front of the proverbial "rock with the word "No" written on it".)
As for better transportation, it seems that Moloch tends to eat most of the "win" in the long run. Compare the experience of having a car today to having one in e.g. 1930s USA.
If a for-profit teleporter were to be built between SV and Topeka, Kansas, the workers will be paying the teleporter operator something similar to the current difference in cost of living between the two locales, while rent-seekers quickly bid up the "new" land. And if the teleporter becomes gratis, the commuters will simply find their wages cut. Rather like the "cost of living adjustment" megacorps presently do (where you can move from e.g. Washington DC office of $company to the one in Taipei, and your wage will drop precisely in half.)
Plenty of people are in the bay area not because it's the only place they get money, but also because they have friends and family there, or like the culture there, or so on. There are many reasons to want to travel, commute for work is just one of them.
Plenty of people also actually prefer working face to face, and not alone at home in front a computer screen - it would be nice if this could be up to individual preference instead of company-wide directives, but having the option is still good.
I disagree with Moloch eating all or most of the win. I live in switzerland, my experience of transportation has been fairly comfortable, if expensive, to date. Making everything even faster would be strictly better, it would mean the "city" as measured by time to reach the center could expand tenfold. Removing transportation would be horrific and would crater the livability of the place.
It is kind of cheating if you're comparing the ease of transportation in a time when fewer people existed and fewer people traveled. I'm also not sure cars are the best example of transportation that scales well.
If a for-profit teleporter were to be built between SV and Kansas, the teleporter would hopefully be state-subsidized rather quickly, with 100 other teleporters built to connect SV to all over the country, so the rent-seekers would bid up land all over the nation. Isn't this what already happened with railway in all the nations where it's significant?
As for megacorps paying less, currently the money is being wasted towards landlords for uncomfortable housing. I don't know whether or how much of that money workers would be able to benefit from if housing became overall cheaper, but we would probably all benefit of it at least indirectly, there aren't many wastes worse than the rent currently going on.
It isn't clear that concentrated "Meccas" like SV would continue to exist at all if transportation could be anything near "a network of free teleporters". If workers could disperse to arbitrary geographies, why would employers refrain from doing so? (Or, for that matter, friends/family/cultural orgs, etc.)
Maybe! In fact, I almost hope so. That sounds quite utopic to me.
That said, given that in reality we probably won't have teleporters, we'll still have "hubs" or stations between which you can go *very* fast over long distances, and then use more humble means to get to your specific place. So you'll still have these hubs, in one form or another.
Yes it is clear. The lower transport costs become, the more significant tiny differences in transport costs become. If in place A costs are $100, and in place B costs are $100.10, the difference is immaterial. Lower the cost to $0.3 and $0.4, the same difference as before is now significant.
Costs are not only money; they also include annoyance, or lack of fortuitous meetings with others who also happen to be travelling in the same places.
As transport costs decrease, we should expect more agglomeration effects, industries converging on particular cities, unless other costs increase enough to stop that happening as may well be the case with SF and its poo problem.
Switzerland keeps Moloch at bay through hard work by millions of people to maintain a dominant culture that values solving coordination problems over individuals achieving personal utility maxima.
> Compare the experience of having a car today to having one in e.g. 1930s USA.
It's quiet, comfortable, air-conditioned, ~never breaks, streams the entirety of human knowledge to you in audio form, uses satellites in space to pinpoint your location and signal for distress if you crash, and is an order of magnitude cheaper?
Who said anything about commute ? The SUBsonic hyperloop would be somewhere between high speed rail and (supersonic) air flight : as you say only rich people would be able to afford it for commute.
SUBsonic hyperloop would be *much* safer than space travel. Air flight is very unsafe too (losing propulsion in flight !!), and yet somehow got popular.
Air travel is one of the safest modes of travel, and most commercial planes are designed to be able to fly with partial loss of propulsion (e.g. loss of 1 out of 2 engines) and to have good control even with total loss of propulsion. Once they're up to cruising height they're in a very safe situation.
The only mode of transport with better safety than planes is the one I insist upon but everyone disregards, elevators!
By the way, what's the point of subsonic hyperloop? that's just a high speed train on a gigantic vacuum tube, at that point you might as well lose the myriad of engineering issues that the vacuum tube brings, you're getting very little bang for you buck anyway.
I question this. We have had the capability for supersonic travel for decades, yet it has never taken off (aside from Concorde). Maglev trains, though much faster than current trains, have also failed to see any application outside niche demonstrations. I'm not sure why exactly - perhaps there's just not enough demand for high speed travel? But if these two technologies have not been widely adopted, I don't see why hypersonic tunnels would be.
The fastest that a Concorde could take you from New York City to San Francisco, ignoring time spent going through airport security and sitting on the runway, would be about 2 hours. The fastest a vacuum tunnel could take you, with a continuous acceleration of 3g, would be twelve and a half minutes.
A sad consequence of this is that it would separate urban and rural people even further. Every major American city would be closer to every other major American city than to any rural city.
My take is that the Concorde ultimately disappeared primarily because it couldn't reach the volume of passengers and routes required to start driving economies of scale and demand for better designs. With the first-generation design, it was only able to compete for first-class passengers, and even for those it couldn't pull in much more money, if any at all, than subsonic first-class seats.
The operating cost premium of Concorde flights made selling economy-class seats infeasible, especially since there wasn't really enough room in the fuselage to put nicer seats and differentiate the classes. So a transatlantic round-trip Concorde ticket cost (according to Wikipedia) about $15,000 in today's dollars, and they were competing with first-class subsonic flights, which go for a similar price.
So you're rich or you work for a company with deep pockets, and for the same cost, you cut your travel time in half or a little bit more but also get a less comfortable seat. Meh? It's certainly a nice option to have, but most of the reason people hate long plane rides is that they're uncomfortable, and they're not nearly so bad in international first class. And you're only spending a few hours on the routes the Concorde ran on in the first place. (Nowadays it'd be even worse, since you can get wifi in your subsonic first-class seat and keep working on the plane, so you don't even lose much time for the company that's paying for your first-class ticket.) So while most people would take a supersonic flight if the cost is equal, the benefit is marginal, so you can't really charge much more for a supersonic flight than for any other flight...
...especially because you need to attract as many people as possible – you have to fill an entire 100-seat plane with first-class passengers! Which is the second problem, because there just aren't many routes that have that much routine demand from people willing to shell out $15K per flight (oh, and by the way they also have to be overwater so you don't have everyone angry at you for making obnoxious sonic booms everywhere, so NYC–LA, London–Dubai, etc., are out).
British Airways at least was able to hit the sweet spot and make a nice profit on operations, contrary to popular belief – but only on a handful of routes like London–NYC. With the current market consisting of a handful of planes on a handful of airlines, it's nearly impossible to make money on a better plane design unless you do so much better you can massively expand the market (and the improvement required here in one go would be huge, given the gap between first-class and economy fares). So nobody tried, and the technology stagnated.
A combination of rising maintenance costs because the design was aging and a nasty accident in 2000 didn't help the Concorde, but ultimately at that point it seems the main problem was that no airline saw any way they could end up in the black keeping it versus focusing on subsonic first class – unless the tech got cheaper, it was a lot of extra bother to run these flights without a whole lot of benefit.
I have nowhere near enough domain knowledge to guess whether early hypersonic tunnels would have larger or smaller operational and amortized design/construction costs than the Concorde.
A very large distance, if you count the hours that all of the overhead that comes with flying with a rocket brings. This video goes into some of the details: https://www.youtube.com/watch?v=02pFTSMbevY
And let's not even mention the fuel tab. And the fact that historically, something like 1% of rockets go boom. And all the other reasons why this is, and will forever be, sci-fi.
Fuel costs are likely to be negligible by the end of the century (converting CO2 via abundant solar energy) and rockets as reliable as 1970s aircraft. The time scope given was infinite, but even looking to 75 years from now it seems possible.
Sound is a big unknown, there hasn't been much pressure to develop quieter launches yet. Underground launches, active sound suppression technologies, these are potential mitigants.
To be clear I really don't think this will happen in the next 50 years, but in 100 years I think it will be possible. Maybe the economics will not work out, maybe people will not put up with the sound, but it will for sure be easier than a 5000km vac-train.
It wouldn't be faster at the theoretical limits, because the rocket would have to go a longer distance, and pass through the atmosphere on the way up and the way down. And it would require much, much more energy. So it's strictly inferior, in a theoretical sense, although it's technically easier (since we've been able to do that for 60 years now).
The point of the tunnel is to let you create a vacuum, which is the only advantage space has over airplanes. You could of course use a rocket motor to go through the tunnel, but there's no need for that, since less-flammable technologies can give you higher acceleration already.
Theoretically it's possible to maintain a tunnel at a perfect vacuum. Realistically I haven't seen anything to show that that's going to work. The current upper limit for vac-trains is around 1000km/h. Going faster just gets exponentially more difficult.
Yes, yes, of course things we've already done are easier than things we haven't done. I'm talking about Musk's /vision/. Travel by evacuated tunnel will be very hard, and might not work within Musk's lifetime; but it is the theoretical best mode of travel. It is the logical endpoint of Earth transportation technology.
The whole point of the concept is to prevent the formation of a supersonic air cushion in front of the vehicle that would then only keep growing as more and more air piles up in front of it, prevented by the speed of sound from going around fast enough.
"Supersonic" with a pressure of 1 atm (and some reasonable range of temperature) means over 1230 km/h. Hyperloop max speed would be 1220 km/h, I wonder if it's a coincidence ?
(BTW, "hypersonic" means more than 5 times the speed of sound.)
Of course, hyperloop is supposed to work at only 0.001 atm, which is equivalent to that of the stratopause which is at ~50 km (between the stratosphere and the mesosphere, interestingly with a temperature around 0°C), at which point the speed of sound is... still 1190 km/h ! Yeah, not a coincidence, still, it's weird that hyperloop *is* effectively ("barely") supersonic at that speed, maybe they have forgotten to adjust for air pressure in these early calculations ?? (Again, breaking the sound barrier is something to absolutely avoid, consider also the damage from repeated supersonic shocks !)
Your typical commercial flight operates instead around 10km, with a typical temperature of -50°C, and a pressure of 0.240 atm, which corresponds to a speed of sound of already only 1080 km/h.
Huh, I wouldn't have guessed that the speed of sound was so much more sensitive to temperature than pressure !!
Wait, you mean the tunnels aren't supposed to be evacuated of air? I guess at some point I assumed that's where he was going eventually. How embarassing. :P
I'm less-excited about them then, but they'd still be great.
Re the boring company, it's worth noting that (unlike rockets or EVs), building tunnels for an order of magnitude (or more) less than is typical in US infrastructure projects is already regularly done in many other countries (including countries with similar or higher labour costs) - and that's for full-sized train tunnels, not just small-diameter tunnels for cars. That the Boring Company has failed to even match that for their smaller tunnels would imply they're probably not doing a great job on the engineering side.
Edit: I looked up some numbers and this is at least somewhat wrong, looks like they actually did match the low end of european costs (at least, assuming we can take their claimed costs at face value).
It's confusing (much like healthcare costs). Broadly speaking with rail tunnels there's two political issues - unreasonable requirements (e.g. asking for overly large tunnel diameters and big stations, or requiring mined stations instead of allowing cut and cover) and paperwork costs (e.g. getting into legal disputes with contractors or overhiring planning consultants).
But on the other hand, American (and latam) construction costs for roads are also an order of magnitude higher, and they don't seem to have higher technical requirements (the legal requirements might still be an issue, but I'm more confused about that). I don't know of any in-depth study for this, but it seems like there's a general cost premium for US infrastructure projects that's a bit hard to explain.
"I’m not sure the Boring Company is interesting enough for this to matter."
Has anyone pointed out that of *course* the Boring company is not interesting; it is, of course, boring. That's straightforward nominative determinism.
I always thought he named it that on purpose as a pun; it's a company where they bore through the earth, and also it's just a less interesting company.
> Or did they all start learning to paint and spending time with their friends and families?
This coincided with Zvi starting to release novel-length posts on AI every week, so I just switched my twitter time to those (and presumably everyone else has too).
My social media engagement statistics plummeted at the time Zvi started the weekly AI roundups. Much better use of time than snarking for someone else's AI training funnel.
"then of his top-300,000-most-intelligent-Americans cohort, he would be in the top 30 most intense, or alternatively, for his top-30,000-most-intense-Americans cohort, he would be in the top 30 most intelligent."
This rings true, in some approximation of the specific numbers. Over some decades I've met or known or been related to a few people who I'd place that high on one or the other of those lists, but, pondering it a bit now, no one who I could list that high on _both_. That _combination_ seems exceedingly rare.
It also reminds me of a sports quote: "Michael Jordan is what happens when a sport's greatest athlete and its greatest competitor happen to be the same person."
In summary, Musk claims that he graduated UPenn in 1995, was accepted to Stanford's physics program, attended for two days, and then dropped out. But none of this is quite true. Musk did attend UPenn, but wasn't awarded a degree in 1995, but rather in 1997, two years after he supposedly graduated- from the eminently bribeable Wharton school. Coursework that Musk posted reveals that he was taking sophomore classes in his junior year. And while UPenn now claims Musk got a physics degree from them as well, the 1997 graduation announcements don't include his name (https://pbs.twimg.com/media/FhyREt6XoAAbh2p?format=jpg&name=large). The implication is fairly obvious: Musk struggled in school, didn't manage to graduate from UPenn in 1995, went to SF to work at tech startups, and then after he'd raised some money and there was the potential for embarrassment (and legal issues) if it became clear that he didn't have a degree, he arranged with UPenn to be given a diploma two years later. Musk has made vague allusions to unfinished classes as a reason for this mixup, but at the time he was on a student visa- it is not believable that someone would "forget" that they hadn't attained the degree that would keep them from being an illegal immigrant.
Meanwhile, the story about getting into Stanford's physics grad program is total bunk. You do not get to start a graduate phd program without an undergraduate degree- if Musk was ever accepted (doubtful, if he wasn't a good student), his acceptance would have been rescinded the spring prior to his starting semester, when he failed to graduate. Musk has been inconsistent about what department he was accepted to (MSE, and Applied Physics have both been claimed), claims that he was accepted to work on a topic ("advanced capacitors and batteries") that the professor he claims he was hired by did not work on at the time, and when subpoenaed, Stanford's graduate admissions office was unable to locate any records relating to Musk (https://pbs.twimg.com/media/FhySnhNXgAQnFfW?format=png&name=medium). Plus, Musk also has claimed that he moved to SF specifically to go into the tech business, something that couldn't be true if he moved to start a 5 year PhD program.
The synthesis of all this is that Musk had an unexceptional academic trajectory, for someone getting into tech in the 90s. He transferred into an elite university, gamed a bit too much, struggled in his classes, didn't quite manage to graduate, but moved to SF at the right time and made loads of money at a tech startup anyway. Nothing about this would be objectionable, but Musk has told stories about his academic career to cultivate a persona of being a scientific genius, and it's just not true.
"At first, I, too, felt like there were a lot of oddities surrounding Musk’s academic record, particularly the Stanford days. But, as I dug in, there were solid explanations for all of the inconsistencies and plenty of evidence to undermine the cases of Musk’s detractors.
During the course of my reporting, for example, I found evidence that contradicted O’Reilly’s timeline of events. Peter Nicholson, the banker whom Musk had worked for in Canada, took a stroll with Musk along the boardwalk in Toronto before Musk left for Stanford and chatted about the incarnations of something like Zip2. Musk had already started writing some of the early software to support the idea he’d outlined to Kimbal. “He was agonizing whether to do a PhD at Stanford or take this piece of software he’d made in his spare time and make a business out of it,” Nicholson said. “He called the thing the Virtual City Navigator. I told him there was this crazy Internet thing going on, and that people will pay big money for damn near anything. This software was a golden opportunity. He could do a PhD anytime.” Kimbal and other members of Musk’s family have similar memories.
Musk, speaking at length for the first time on the subject, denied everything alleged by O’Reilly and does not even recall meeting the man. “He’s a total scumbag,” Musk said. “O’Reilly is like a failed physicist who became a serial litigate. And I told the guy, 'Look, I’m not going to settle an unjust case. So it’s just like don’t even try.’ But he still kept at it. His case was tossed out twice on demur, which means that basically even if all the facts in his case were true, he would still lose.
“He’d tried his best to like torture me through my friends and personally [by filing the lawsuit]. And then we’ve got summary judgment. He lost the summary judgment. He appealed summary judgment, then several months later lost the appeal and I was like, 'Okay, fuck it. Let’s file for fees.’ And we were awarded fees from when he appealed. And that’s when we sent the sheriff after him and he claimed that he had no money basically. Whether he did or didn’t I don’t know. He certainly claimed he had no money. So we were like either we’ve got to like impound his car or tap his wife’s income. Those didn’t seem like great choices. So, we decided that he doesn’t have to pay back the money he owes me, so long as he doesn’t sue anyone else on frivolous grounds. And, in fact, late last year or early this year [2014], he tried to do just that thing. But, whoever he sued was aware of the nature of my judgment and contacted the lawyer I used, who then told O’Reilly, 'Look, you need to drop the case against these guys or everyone’s going to ask for the money. It’s kind of pointless to sue them on frivolous grounds because you’re going to have fork over the winnings to Elon.’ It’s like go do something productive with your life.”
As for his academic records, Musk produced a document for me dated June 22, 2009, that came from Judith Haccou, the director of graduate admissions in the office of the registrar at Stanford University. It read, “As per special request from my colleagues in the School of Engineering, I have searched Stanford’s admission data base and acknowledge that you applied and were admitted to the graduate program in Material Science Engineering in 1995. Since you did not enroll, Stanford is not able to issue you an official certification document.”
Musk also had an explanation for the weird timing on his degrees from Penn. “I had a History and an English credit that I agreed with Penn that I would do at Stanford,” he said. “Then I put Stanford on deferment. Later, Penn’s requirements changed so that you don’t need the English and History credit. So then they awarded me the degree in ’97 when it was clear I was not going to go to grad school, and their requirement was no longer there.
“I finished everything that was needed for a Wharton degree in ’94. They’d actually mailed me a Wharton degree. I decided to spend another year and finished the physics degree, but then there was that History and English credit thing. I was only reminded about the History and English thing when I tried to get an H-1B visa and called the school to get a copy of my graduation certificate, and they said I hadn’t graduated. Then they looked into the new requirements, and said it was fine.”"
I don't think that biography quite assuages my doubts- yes, O'Reilly lost lawsuits, but as Musk himself points out, he lost on summary judgement, before any factual allegations had to be proven. It wouldn't have mattered if Musk hadn't actually gotten into Stanford, or if he'd been given a degree on the sly a couple years late- for all legal purposes, Musk is a graduate of UPenn. In a later lawsuit where similar facts were brought up, Musk quietly settled.
And the PhD program part of the story still has issues, as well. I haven't heard of any PhD program admitting students on the promise that they'll finish their bachelors *after* they get the PhD. Maybe Musk was given provisional admission and didn't end up attending after he didn't graduate. But it strains credulity that someone on a student immigrant visa, without a degree and therefore unable to transition to a professional visa, would just drop out of a PhD program and not realize that they hadn't attained a bachelors degree. The immigrants I know are all very aware of the requirements around degrees and jobs, because screwing something like that up can get you deported and banned from returning to the US.
">>>yes, O'Reilly lost lawsuits, but as Musk himself points out, he lost on summary judgement, before any factual allegations had to be proven. "
This is misleading and best, straight up incorrect at worst, and possibly conflating a motion to dismiss with a motion for summary judgment. Summary judgment is generally based on the factual record as fully developed (at a minimum, in relevant part, but often for the whole case) -- there are no more facts that are going to come in. If you win on summary judgment it means that you have established that *no reasonable juror* could find in the other side's favor because there is *no issue of material fact* that could support their position, and thus the issue shouldn't proceed to trial and you are entitled to judgment as a matter of law.
Isn't there a lot of separately implausible things that all need to be true for this to make sense, compared to the relatively simple hypothesis that someone with the means, motive and opportunity to embellish his academic record and then cover it up, did so?
Weird, to me it seems extremely implausible that these prestigious schools were all conspiring with Musk to fake his degrees/acceptance compared to the simple hypothesis that he's telling the truth.
I could very well be wrong, I was vaguely aware of there being allegations against the authenticity of Musk's academic merits before, but this is the first time I've taken a deeper look at the arguments for and against this claim.
But at least going by the comment from wax above (which I haven't verified, but the comment provides plenty of sources, and nothing in it triggers my bullshit detector), there are just too many odd things that happen to be just right for Musk's claim to hold up: Penn just happening to drop the requirement for the courses he was missing a year or two later (and then happening to reinstate these requirements at some later point) Stanford somehow accepting him into the PhD program before his bachelor was technically completed... to me it just reads too much like one of those contrived sitcom plots where, through some implausible sequence of perfectly innocent reasons, two characters just happen to look like they're engaged in something risqué the moment a third character walks in. Which, I am sure, must have happened to some people at some point in real life, but far less often than people get caught in what looks like a risqué situation because it is.
Schools can and do change graduation requirements as frequently as each academic year, and graduate schools frequently accept undergraduate applicants before they have completed their bachelor's on the condition that they complete their bachelor's (and usually maintain a certain GPA). These are both very basic things that happen.
2023 Musk, sure, elite colleges give out honorary degrees and other weird stuff to particularly rich, famous, or powerful people. 1995-7 Musk? There's very little reason for them to know him or care at that point. Why would they want to conspire with some 25-year-old about faking a degree?
(i) I know someone who has a PhD from Stanford, and remembers Musk (very briefly) being in his PhD program. (ii) what does it matter anyway? That he got into a PhD program at Stanford is not even in the top-ten most impressive things about Elon Musk.
I'm a Physics professor. I know another Physics professor who has a PhD from Stanford, and who tells me that he remembers Elon Musk as being in his PhD program (same year), up until Musk dropped out. I consider this dispositive. Musk was admitted, and he showed up (admittedly he didn't stick around for very long).
Yes. Apparently Musk tried to convince my friend to drop out with him and join him in founding [some startup] but failed to persuade. But apparently he gave a very intense sales pitch.
Having a Master’s before getting a PhD is the usual process but there are schools and programs that accept applicants into PhD programs that do not have a Masters.
> AI experts, is this a big deal? Can other AI teams not access Twitter data through the public web? Is it a substantial amount of text compared to other corpuses? Is the structure (280-character blurbs written by morons) a limiting factor? Or is this a genuine treasure?
My guess: the main thing Twitter will be good for is having a good source of links to high quality articles. This was what Reddit was most useful for wrt training GPTs back when its API was free [1]. Highly upvoted articles linked on Reddit were assumed to be good, and so included in the training runs. Otherwise you just get a bunch of SEO garbage. Similarly, highly liked tweets linking to stuff by verified users likely won't be SEO garbage, so will be useful for building datasets with.
Twitter overall has value as training data for LLMs but it's not crucial. I wouldn't consider links to be very valuable. (I do have relevant experience).
Why wouldn't you think links are so valuable? I wouldn't be that surprised if Twitter links to more quality content than it generates (likely it generates a lot, but much is also likely bots).
Would you want to limit your web training data to only linked articles? Clearly not, as this would significantly reduce the size of your training dataset and adversely affect the resulting model.
So what's left is using Twitter links as some type of a quality signal. But it's more likely these links point to stuff that's controversial, political, or appeals to a wide audience. This is the kind of content that's easy to find, so it's not what you're missing in your training data. And that's assuming you're even using ratings of content pieces to train your LLM in the first place.
What's been done recently is using a known, good LLM to rate and filter training data for a smaller LLM but that's quite different. You could likely hand-craft a rating system that would be better than any link signal from Twitter, simply based on the domain name where the content appears.
Maybe, but I’m skeptical. We do filter datasets based on social media engagement. I’m guessing you think The Pile shouldn’t have gone with their Reddit link approach?
These crawled links accounted for only 62.77 GiB of The Pile. At 4 chars/token, that's 15.7B tokens. Newer open datasets, which don't use this strategy, reach 6.3T tokens (https://arxiv.org/pdf/2309.09400.pdf), making this just 0.25% - a rounding error.
I've seen no evidence that this strategy of using social media links improves the model compared to adding more data from top domains, or that it's even a decent indicator of high quality diverse training data.
My understanding was that Deleted Tweets, Direct Messages, and GroupChats were high value for Twitter, would that content also be valuable from the perspective of LLM training? It is data that isn’t publicly available.
Personally I stopped using Twitter when they started requiring Twitter Blue to use Tweetdeck. The old Twitter interface was good, the new one sucks, but hey, at least there's Tweetdeck! Except, oops, now you have to pay for that. So I just stopped reading Twitter.
I think the Twitter corpus contains some alpha that isn't captured by other corpora, even if one ignores the temporal aspect. Reddit probably has a more valuable corpus right now but they are smaller and seem to have antagonized many of their their core users, just like X.
I'm very non-confrontational (even cowardly, at least with regard to social interaction). I wouldn't be able to handle working for him even if that was somehow an option for me. However, I get a lot of vicarious satisfaction from watching someone tech-inclined who's both bold and powerful enough to tell the whole world to shove it, even when I think the world is right and he's wrong.
I apply Stafford Beer's "the purpose of a system is what it does".
Musk has done two unique (and to me very valuable) things.
First, he changed the mindset of many western governments, and China's government, from "electric cars are a cute idea but totally impractical" in 2010, to, in 2022, "you may not sell anything BUT electric cars as of [a date in the 2030s]." In twelve years!
Second, he is in the process of creating an in-principle uncensorable world-wide communications network that cannot be stifled by government action, or by the absence of government (say in a rural village in Afghanistan or Eritrea.)
I suppose you're right in places where there is no government but I don't think you can operate Starlink in China or Russia for long without censoring.
That presumes you can get the equipment and have a way to pay for the subscription. I think it's far more likely that they will offer the service with the censorship in order to make money - at least that's what Twitter/X is doing.
<speculation>Perhaps X is just playing nice with the expectation that enough dissidents will "obtain" Starlink terminals that the content will be effectively uncensored.</speculation>
I actually give Musk less credit for the EV adoption piece. 2 points. 1) incumbent electric cars were fairly boring and lame before tesla, but tesla's existence (pre-musk) is proof that the market would figure out the appeal of EVs eventually 2) the price of lithium-ion batteries is a massive driver in EV affordability, and the price dropping is purely the result of mining/refining/assembly infrastructure maturing
> If Parag or whoever employed a thousand censors to keep the ADL happy, and the ADL becoming unhappy cuts Twitter profits by 60%, then there’s a strong business case for those censors!
Wait. Zoom out from Musk for a moment. Surely there is something about this state of affairs about which we should be very unhappy. What you have described is essentially an extortion racket, not unlike the ones that the mob sometimes ran. Saying "there's a strong business case" for paying the extortion money doesn't seem like something we want to normalize. Why is OK for outside groups to impose such costs? What if there were an analogous, competing group making demands in the opposite direction? Would that be just fine too?
Oh, OK. It sounded like you were holding it to be actually neutral. I agree that, descriptively, this is how big tech has come to work.
But I also think that you can steelman Musk's actions as a higher-level judgment that the business landscape itself has become corrupt and needs to change, even if that costs Twitter in the short run. A bit analogous to SpaceX pursuing a different business model from previous spaceflight contractors.
Yeah, if you trace back all the references, my original claim was "Parag must have been really dumb to hire 10x more people than he needs", and this was what I was updating on.
Re: hiring 10x more people than necessary, it may be something discussed in https://danluu.com/in-house/ [0], where, tl;dr Twitter had several world class experts at (simplifying the systems involved) operating systems and programming languages who are nominally not directly working in the product, but regularly can have their expertise called on to solve "unsolvable" bugs. When I think of the number of people lost, I would assume that there are lots of examples of this during the earlier Twitter purges. And their disappearance won't be "missed" until the next large "unsolvable" outage comes and lo and behold, all of the expertise has left the company.
[0] Note that while the blog post makes many confident claims about cost savings and counterfactual a regarding incidents, I have not actually seen their reasoning at work and I have a low grade suspicion that the blog author is far less careful about finding alternate explanations compared to their apparently high confidence level. This is a vibe and should not be interpreted as well reasoned.
As a software engineer, I find it highly implausible that the experts of old Twitter had some magic secret sauce that Elon would find difficult to replace via new hires.
> tl;dr Twitter had several world class experts at (simplifying the systems involved) operating systems and programming languages who are nominally not directly working in the product, but regularly can have their expertise called on to solve "unsolvable" bugs.
In reality this practice is more about prestige and recruiting than unsolvable bugs. You get these CS celebrities at big companies, they do a bunch of research and sometimes you can apply it, and they give talks with your logo on the slide deck which helps you recruit the programmers the next few levels down from them.
For companies like Google and Facebook that are throwing off billions of otherwise-unspendable dollars, this is worth it. At Twitter maybe not.
Why are you assuming that these high level technical experts got fired, rather than various low level, and especially non-technical employees (that also weren't even at Twitter a few years before that) ?
There was a lot of talk previously on Twitter about how technical talent got hollowed out, and how it was sure to fail around the World Cup because it was a period of peak usage. Of course that didn't happen, but if it was a called shot it'd be for the more recent spate of problems and it was just called too early.
Also there is a certain common dynamic to lay offs / hiring where if a particularly competent member, or a member with high status within engineerin, but maybe not management leaves that can also cause a chain reaction of team members who lose faith in management then leave, or whose job can no longer be performed without the star member.
Obviously not great evidence, which is why I presented this as speculation but hardly unsupported. Do you have a model that fits more of the facts (layoffs leading to instability) or do you dispute that some details aren't facts? (Twitter is still stable, crucial engineers were re-hired and so on)
Also I think the pre-Elon regime was worse on censorship and I like that he tells the ADL/The Left to get lost and has largely not engaged in censorship of the Right.
Wait how is it an extortion racket? Is ADL asking that Twitter pay them? There really is a lot more antisemitic stuff on Twitter these days, I don't see why it's unreasonable for that to make advertisers more reluctant.
Its only an extortion racket if you think that the things that ADL points to are things that otherwise noone would care about or have no impact on the business, which seems unlikely. You could equally say that money spent on health and safety is extortion money for insurance companies/health department/etc
The things the ADL has been complaining about Twitter allowing/condoning seem like normal, common-or-garden antisemitism. The sort of things an advertiser might be worried about even if there wasn't an organization devoted to pointing it out to them.
The ADL define anti-semitism as, amongst many other over-broad things, being critical of the ADL itself, so their definition of the term is deceptive and manipulative. Read their report and see for yourself:
If you look at the first tweet they call out as evidence of Musk-enabled Jew hate, it's literally someone tweeting the ADL's own words back at them. They claimed he took their words out of context (he didn't), and that he was to blame for the ADL disliking any replies, including indirect replies.
Did Trump literally say the words "expelling warmongers, driving out globalists, casting out communists, and throwing off those who hate our country", or did he not ? (If not, did he says something very similar ?)
(My suspicion is that this is yet another example of the Twitter character limits causing damage.)
Because the alternative is "the ADL is not allowed to talk about content it finds distasteful on Twitter" and that's not how free speech works. The ADL is not putting a gun to anyone's head, it's saying "we don't like this content " and advertisers are voluntarily saying "we agree, it's bad for business for us to be associated with that content."
>What if there were an analogous, competing group making demands in the opposite direction?
They'd be protected by the same right to free speech that protects the ADL, and advertisers would have to decide which group they find more convincing on if they should pull their ads or not. The marketplace of ideas in its most literal form.
The thing any conversation about Musk inevitably reveals about the VAST majority of people who otherwise consider themselves intelligent is that they are unbelievably susceptible to manipulation and propaganda, and that there are many many midwits who love to look down their noses at someone that is deemed 'smart'
I’m very conflicted about Musk, because it seems to have it’s own particular version of Gell-mann’s amnesia.
When he talks about something you’re really good at, it’s hard to not see he is talking BS, but when he talks about other stuff, he really sounds like an expert. We should be discounting his views in everything else, but instead most people inflate his opinions of what they know is wrong.
A lot of people here know a lot about AI, but if someone expressed Musk musings on AI around here, there would be much less incentive to try to “read” deep insights.
It’s also interesting as someone who has worked in manufacturing that a lot of his “bright” ideas seem to be the kind that get you in trouble *on the long run*.
For instance, while the aluminium frame story makes him sound super smart if you stop where the book does, it takes on another look once you realize Tesla has had to settle lots of lawsuits related to the aluminium frame. The reason for not using aluminium is that you *really* want your car frame to be a single BIG piece, not several pieces welded together, especially in an EV where your frame protects the battery in case of impact. By welding you create stress points that tend to shear on crashes, which is one of the reasons Teslas are notorious for going in flames even after relatively minor crashes.
Musk proposed using Starship for P2P, one of (if not *the*) loudest rockets there is. If planes are inconvenient because of how far they need to be from urban centers and how much space they require, heavy rockets are hardly a good solution.
Then there’s the obvious risk factor. Even the best rockets aren’t reliable enough for commercial passenger travel. We’d need orders of magnitude improvements on reliability for the idea to be feasible.
Looked with any detail, the concept is an obvious non starter.
is the idea here that within-the-hour transport to anywhere on Earth's surface isn't useful, because you might have to take an hour-long train ride to get to the rocket-port? or that it's completely impossible to make a rocket reliable enough to routinely carry paying passengers on?
these are implementation details, and the idea that they make the whole idea an "obvious non-starter" is risible.
When someone is off orders of magnitude, is not “an implementation detail”, to say that is to fail at basic engineering. Moving something from ~1% failure rate to ~0.01% is a gigantic task, because you’re into heavy diminishing returns zone. And that’s assuming it can be done, which it might not.
By that logic making hypersonic planes is just “an implementation detail”...
There’s LOTs more issues by the way:
- Rockets don’t take off on bad weather, but more importantly *they can’t do vertical landing on bad weather either*.
- Fueling a rocket takes hours, and it’s done just before launch for safety reasons. Changing this requires compensating with even better failure rates.
- The economics barely work out even if you *could* reuse the rocket several times a day, AFAIK SpaceX quickest turnaround for reuse is still measured in weeks.
- Depending on location, a suitable launchpad/landing pad is far enough that lots of routes are pointless (optimistically: 10m boarding boat + 2h boat + 10m boarding rocket + 1h + 10m + 2h second boat= 5h30m i.e. NY to LA).
- A lot of people would simply be unable to handle the g load, reducing the g load necessarily increases flight time heavily.
- At least 10x the emissions/passenger trip as a regular flight.
I could keep going. This is just a *bad* idea, there’s simply no way around it, and wishing the end result (travel anywhere in 1h) does nothing to change the physics of it.
Has anyone done the maths on how loud it would actually be on shore at that estimated distance, and how that compares to, say, a helicopter flying overhead?
"It has worked on a much smaller scale, for a much shorter distance, under experimental conditions, with almost non-existant concerns for safety, economic feasibility, or public and environmental impact"
is an extremely weak rebuttal in general, and in this case in particular. If you really are an "engineer" (as opposed to, say, a mathematician or a theoretical physicist), you should already know that. It's like claiming that economically viable fusion power is easy because I can build a fusor in my garage.
The claim being rebutted is "he okayed the idea of point to point rocket travel, which is so full of holes that IMO it puts into question how much he knows". All that's needed to rebut that is a simple existence proof, which Delta Clipper Experimental provided before some of the people in this thread were born.
Would point to point rocket travel be economically viable for passengers? Dunno. It could wind up like the Concorde, technically feasible but economically a failure. But that's a question requiring detailed analysis and probably building prototypes. Almost certainly an intercontinental passenger transport would be a different design than a Mars craft.
Would there be lots of detail to work out? Oh, Hell yeah. Would the ride be rough enough to filter out some passengers and cargo? Depends on the vehicle design.
Would some adventurers want to ride it even if it cost ten times what a 747 does? Yeah, but not enough to pay for it. That's why you do business case analysis--something I have done as an engineer.
> All that's needed to rebut that is a simple existence proof, which Delta Clipper Experimental provided before some of the people in this thread were born.
No, it didn't, because in the context of this thread, "point to point rocket travel" clearly doesn't refer to a short hop of a couple 100m by an experienced test pilot in a vehicle with room for exactly zero passengers, but to Musk's plans (or concept art?) for commercial, airliner-like transport of regular people. Nobody doubts that SpaceX can launch and land rockets vertically – but that's not what Xavi meant by "idea which is full of holes". The DC-X tests only demonstrated that rockets can launch and land vertically, but that's only a miniscule part of what's needed to show that the idea isn't "full of holes".
To add to that, the space shuttle regularly landed at a different place to where it launched, and could in principle have been used for point to point transport. There are major logistical issues, however, as well as national security issues (I doubt Starship will ever be allowed anywhere near China, for example). Personally I doubt Starship will be used for point-to-point transport, but its not hard to imagine that something else could be.
I’m perfectly willing to concede you can design a craft for point to point suborbital transfer, but it’s clear that that vehicle isn’t Starship.
To propose Starship is a viable craft for that purpose is IMO bonkers. Even if SpaceX can iron out the details, Starship will by design be very different from what you want from an “space airliner”, which it’s not surprising if it was designed as a craft to get heavy payloads to Mars.
"but I’ve not seen a single instance where he showed “knowing rockets” past what I’d expect for the average KSP player."
Where have you been looking? Because anyone talking about rockets to the general public, is going to cap their exposition at roughly the KSP/Scott Manly level if they want to actually communicate. KSP may be a silly-looking bit of entertainment, but it actually encodes rocketry knowledge at about the highest level an audience without engineering degrees can handle.
As someone with a couple of aerospace engineering degrees, Elon knows his stuff. He's not in the top tier of rocket scientists, but he's good enough to work with and effectively manage the top tier.
I’m not expecting him to start trying to explain real orbital mechanics, but I’d expect him to have insights that require good knowledge. I’ve found 0 evidence of that, including his SpaceX presentations.
I’m not saying “KSP is bad”, I’ve spent my fair share of time on it. I’m saying “if that’s your bar for genius level rocketry, your bar is far too low”.
Again, all I get is people who say “he knows his stuff”, but I’m still waiting for actual *evidence* of that claim, and so far the evidence I got says he doesn’t fundamentally understand a lot of the physics involved.
You've moved the bar from your claim that for experts "it’s hard to not see he is talking BS". An expert just told you otherwise. Even if Musk really doesn't know his stuff, this still defeats that claim. Clearly it's not such obvious BS to others who know this stuff.
I think your reading of my point is a particularly uncharitable one. You’re translating “I see experts claim Musk points are BS” to “all experts disagree with Musk”, thus one counter example is enough to disprove the point. That’s a ludicrous bar to pass, there’s still experts who e.g. think cold fusion is possible.
In my view, if someone is claimed to be a genius, I’d expect experts to disagree on the margins, but I’d also expect a broad agreement of the basics. That’s not what I see with Musk.
I see some experts agreeing with Musk, but then I see vocal critics who say he doesn’t understand something fundamental about their particular area of expertise. As an example:
- Pesenti (among others) on AI. Some are more diplomatic than others.
- Nicolelis on how the brain works.
- Theo on how internet advertising works.
- Bryan Cantrill on how software development works.
I'm not assuming that you think every expert disagrees with Musk. I'm just objecting to the idea that there's Gell-Mann amnesia for Musk or that it's obvious on average to experts that Musk is talking BS.
I'm not sure what all the people in your examples have said about Musk, but Pesenti was objecting to Musk saying in a Lex Fridman interview, "I think we're missing a few key ideas for artificial general intelligence. But it's gonna be upon us very quickly"
1. Pesenti objects to Musk even using the term AGI, which he says isn't a thing. But it was Lex Fridman, an AI researcher, who asked Musk if he thinks AGI would will come from the current approaches or if some new ideas need to be invented. Musk was just answering the question as it was posed.
2. Pesenti is confident human-level AI won't happen within his lifetime. This is itself a controversial opinion among AI researchers. And Musk only vaguely says he thinks general AI will come soon, but he gives no indication of whether he thinks soon is within 10 years or more like 40 years. There's nothing objectionable about Musk's statement here and plenty of AI experts have said similar things.
3. Pesenti is upset that Musk is concerned about the dangers of an AGI taking over. A lot of AI researchers are concerned about this.
So I don't think Pesenti counts. He's jumping on a vague answer Musk gave in a casual podcast that plenty of AI experts would agree with.
It would be genuinely bizarre if he didn't know about rockets at this point though, right?
I did not understand why this quote was a big deal:
"When I met Elon it was apparent to me that although he had a scientific mind and he understood scientific principles, he did not know anything about rockets. Nothing. That was in 2001. By 2007 he knew everything about rockets"
That is literally **6 years** of extremely devoted and focused time. How could an intelligent person NOT become incredibly knowledgeable in that time period?
He started on rockets in 2001 based on the quote. Again, this is not really noteworthy. An intelligent nerd knew many things about their domain of interest after years of hyper focus? Common.
Are you aware that they have slowly been progressing towards a "single big piece"? They already apparently have some of the largest presses in the industry, and are soon to be rolling out even larger ones that will essentially form the entire body from a single piece instead of the 2-3 it is currently made of.
So we should give Tesla a pass for knowingly making dangerous cars because they’re now doing better?
Even if (and it’s a big if) other carmakers are being excessively cautious, I think it’s bonkers to consider someone a genius because he’s aggressively cost cutting by not caring about safety while designing a car.
Teslas routinely ace road-safety tests. A welded frame may well be less safe, comparatively, than a single-piece forged one. But it's obvious that early-generation Teslas were not actually "unsafe" on any objective scale.
They also have a track record of reporting false results of their safety ratings (and getting called on that) covering flaws by having customers sign NDAs, and there’s leaked code that it’s a strong indicator that they have “special modes” for safety testing.
Just off the top of my head, since the setup of crash tests is known in advance, you could adjust the airbag deployment. You could also tweak the auto braking to be way more sensitive, because test setups don't have floating debris or incoming sun rays that will cause phantom braking. Once you know the details of a situation, there's lots of stuff to optimize around.
There's lots of evidence, but of course looking for "Tesla" and "Safety" pulls out Tesla's press releases and articles that suspiciously tout the info in the press release. For starters, the obligatory wiki link:
Here you can find info on troubling behaviors like the stealth recalls, the misleading safety ratings, etc.. There's also some of the notable Tesla issues. Note that this issues weren't addressed by Tesla through active monitoring or feedback response, the recalls were either safety regulatory body initiated or initiated in response to a lawsuit.
Just couting NTHSA recalls, Tesla apparently takes the cake:
And that's with less models and less cars on the road that most of the other brands.
Finally, while the data is now old, this is informative of a lot of my views regarding safety testing in the automotive industry (I've worked in safety related electronics):
Briefly: instead of focusing on a particular rating, you look at how well existing cars test on *new tests* (tests that did not exist when the car was designed, in this case driver-side small overlap). The idea of this kind of study is to see if the automaker was optimizing for safety or optimizing for good ratings.
My conclusion with all of this data, is that Tesla wants high safety ratings, but as a company it puts safety far back in the priorities list. Their approach to safety sounds similar to the Web approach to development, ship fast, let the users find and report the bugs, fix what you can through updates. The stakes are a bit higher with a car than a 404 page though.
This perception also meshes well with other data points, like the high number of OSHA related complaints Tesla has, and the concerning number of accidents on the factory floor.
Safety of EVs has more to do with having a lower center of gravity, more weight from battery backs, and that battery packs add to the strucural integrity of the car. I haven't seen the forged frame touted as beneficial for safety, rather that it's theoretically cheaper to manufacture. The downside is that a single-piece frame can lead to higher rates of unrepairable structural damages for minor-moderate collisions.
A) Please provide evidence both that Teslas are significantly unsafe and that this is due to the frame compromises made by choosing aluminum and
B) this proves that Tesla (and likely Musk) knew exactly the thing you claimed they didn't know. I was showing you that your example of a supposed lack of basic knowledge was simply wrong.
They may have chosen a tradeoff that you think is a bad one, but they clearly knew about the welding difficulty and that a single piece would be preferable, which directly contradicts your Gell-Mann amnesia argument.
A would be very hard, even if Tesla was open about their safety incidents (which they are not, as I explained in another comment on the thread.
B requires me to show what Musk knew at that time which I'm not sure is even knowable..
What I can provide you is evidence of stuff like welds on structural elements that would be considered very deficient in any other manufacturers, and any mechanical engineer can tell you they'll compromise the results of any safety ratings the car might have. Safety ratings are measured in part by crashing actual units, which means the results are only valid as long as the manufactured units and the tested units are similar enough, i.e. there's a strict quality control that ensures that the cars rolling off the plant are almost identical to the ones tested.
Terrible welds on Model S structural elements (Spanish, the mechanic in question is repeatedly asking Tesla if the part really complies with their QA, and wants written confirmation before installing it):
This video also shows another failure caused by the aluminium alloy, the mechanic explains that driving the car over a regular bump that the customer has in their access road has damaged the frame. One would expect the screw or the front wishbone to get damaged first. This kind of damage can easily impact stability on the highway. It also explains that as a consequence from that extra slack, the screws were continuously hitting the battery.
The aluminium frame has other implications, the main one being that the variability on the frame seems to be the underlying cause of misalignment issues. Some of these misalignments are catastrophic in terms of safety:
In this video, a misaligned rear door is rendered permanently closed, even in the case of emergency.
There's also a LOT of videos, articles, etc... showing quality issues, including this very funny (but still concerning IMO) finding from a Model Y owner:
The question is does it work at what cost. The case against Musk isn't that he has acheived nothing but that he's acheived things by a combination of luck, throwing money at the problem and marketing, not any particular organizational genius
if you could completely revolutionize orbital-lift rockets by "marketing and throwing money at the problem", Boeing et al. would have done that a long time ago.
Musk would belong on any list of genius CEOs just for SpaceX alone, leaving aside everything else.
"Luck, throwing money at the problem and marketing" could plausibly make someone rich, but I don't think it's plausible that someone could become *as rich as Musk is* without some serious talent.
There are Hard problems that no one currently has a solution to. Musk has solved *more than one* of these Hard problems (rockets and electric cars), helped solve electronic transactions (which isn't a Hard problem, but still impressive), is casually working on tunnel efficiency with the Boring company, and is working on Twitter.
Since no one solved electric cars or rockets before, despite many billions of dollars on both, it appears that Musk may be able to identify real solutions to problems that other very smart and very capable people failed to solve. It seems obvious that solving these kinds of problems necessarily involves doing things differently than previously tried. That Musk has tried different things that have been successful is very strong evidence in favor of his ability to do it, rather than evidence that he managed to luck into solving multiple problems that tens of thousands of people spent billions of dollars trying to solve but didn't.
This is irrelevant to the discussion, but I wonder about the derivation of IQ from SAT scores. Is it really as simple as saying, well he's probably 1 in a 1,000 good at the SAT, so he's probably at least that intelligent?
"The correlation between SAT and IQ in a representative sample is about 0.80 if there were no measurement error and no restriction of range. The fact that this correlation is not 1.00 means is that you cannot take 99th centile SAT and convert to 99th centile IQ. You have to apply some regression towards the mean. You have to use the z scores for this as centiles are non-linear transformations. 99th centile SAT in the table is about 1450 (rounded). If we compute it using the mean of 1010 and SD of 200, we get 1475, which is pretty close. In z scores, this is 2.33. If we want to get the IQ for this, it is 1.86 (2.33 * 0.8), which is 128. In the same fashion, a perfect SAT of 1600 is 99.8th centile, and about 2.95 z, which is then 2.36 z on IQ, which is about 135. This is a lot less impressive than sometimes thought and presented on various websites because these websites implicitly assume a correlation of 1.00 between IQ and SAT scales."
When you correct for this it lowers my estimate of his IQ about 5 points, which I've edited into the post.
What's the correlation between one IQ score test and another IQ score test for the same person using different questions on a different day? I'd be surprised if it were much greater than 0.8.
Even scores on an IQ test would be expected to regress towards the mean on a different IQ test though, so that's not unique to the SAT.
It seems like the fair thing to do is either regress all test scores based on their correlations with other IQ tests/g-loading, whether they're officially IQ tests or not, which would confusingly mean no one's "IQ" is actually what they scored on an IQ test.
Or the simpler option of treating scores on all decent tests as more or less equivalent using centiles/z scores as you originally did.
In the case of Elon we have a strong prior that he's not of average intelligence so it doesn't make much sense to regress him towards the mean of the general public. Perhaps we should even expect him to do better on an IQ test battery with spatial tests.
I'm not sure how to think about this; maybe @emilkirkegaard or @cremieuxrecueil can weigh in. My guess would be that all IQ tests are correlated about 0.9 with each other and g, but that IQ = score on an IQ test and so it's by definition correct, and the SAT is correlated more like 0.8 and is not exactly an IQ test so it's not by definition equivalent to your IQ. I could be totally wrong about that though.
I think the right way to perform the Bayesian update is to treat the IQ/SAT miscorrelation as noise, correct it by regressing it, and then once we have an effective IQ score, combine it (at that stage) with our prior about how smart he is. Otherwise we risk double-counting.
"all IQ tests are correlated about 0.9 with each other and g"
The big batteries maybe.
"IQ = score on an IQ test and so it's by definition correct"
By definition yes, but it's still slightly misleading(and what could be worse than that?) for SAT scores to be subjected to regressions that IQ tests are spared from simply because they're called IQ tests. Hath not a SAT load?
But I guess it doesn't matter that much, even a simple "SAT score divided by 10" IQ conversion probably worked ok for the old SAT.
If you would always apply regression to the mean, wouldn't that equate to always lowering the IQ measurement? I'm confused why this number should always go down, rather than potentially up. If it only correlates at 0.80, that doesn't imply that it's always higher, only that they're not the same thing.
So scores below 100 always move up and scores above 100 always go down? That still doesn't make sense to me. A mismeasurement can go up or down, and there's no reason to think that a mismeasurement of a certain kind can only go one way or another.
Twitter: If Musk's eventual goal is to transform the company to an Everything App, it's probably premature to judge the results. This would be someone buying a property to flip, then claiming it was a disaster when the house is stripped down to the studs. It might be too early to tell whether this will turn out.
"Musk was an early investor in Tesla because the founders approached him": This seems to argue the opposite what you did in your review. Sure, Tesla approached Musk, but having recently exited PayPal flush with cash, I'm sure lots of people in SV approached Musk. His reply to Tesla seems to have been, "sure, and I've been funding a guy to do cutting-edge battery research!" which doesn't seem like a 'right-place-right-time' situation, so much as 'fortune favors the prepared'.
I take your point that it's unfair to attribute a single motivation to something as complicated as a person, but it's probably also unfair to say someone like Musk is motivated by the whims of the moment. It's clear with many of his ventures they are things he thought about for a long time. The book makes it seem like he was interested in space and rockets, but didn't really believe he could build them until they talked to the Russians and he did some back-of-the-envelope calculations. In the same way, it seems like he was interested in electric cars, but didn't believe he could make them happen until Tesla approached him.
In both cases, once he perceived his vision was 'possible', he devoted large amounts of time to making them reality. You don't build Tesla or SpaceX on a whim.
I can also say firsthand that I was a Birdwatch beta tester since ~mid 2021 and the feature worked more or less the same way that Community Notes works now. It's great, but it certainly wasn't Musk's project. Credit to him for doubling down on it, but not pioneering it.
Hard to say from the outside, but from what I've observed:
* He's called out the feature a few times and drawn attention to it being important to Twitter's future
* When I started using the beta, the notes were actually only visible to other Birdwatch testers - they went visible to everyone right around the time of the acquisition (Oct '22), so presumably not Elon's direct doing but adjacent to it
* Anecdotally, it feels like it's gained more visibility in the UI and more people have been invited into the program as notes writers, which presumably he would have signed off on
"I’m not sure whether this means that everyone else is an idiot with a pointless bureaucracy fetish, or that only a few very special people like Musk can make the non-bureaucratic version work."
It's the latter, though I'm sure Musk has access to a unique (temporally speaking) pool of disaffected talent. One mismatch that marketeering types sometimes seem to have with the general populace is that, yeah, the market CAN produce things with amazing quality and superior efficiency, but there is often a risk of total failure where you get nothing. When you apply this dynamic to things people rely on and pay taxes for (or sometimes, just the latter), they get understandably leery. And the more a successful firm begins to be relied on, the more they start to stagnate.
> It's the latter, though I'm sure Musk has access to a unique (temporally speaking) pool of disaffected talent
Yeah, Musk seems to have pulled off a more extreme version of what a lot of startups in Silicon Valley try to do of selling (mostly young) people on a dream and a mission, that makes them willing to work long hours at low pay, or put in extraordinary effort, that they wouldn't at a company that said "come work 9-5 and make progressive technical improvements".
Whether you see that as exploitative or just good leadership probably depends on the case in question. Anyone who has worked in industries like this will have seen young enthusiastic people get burned out by believing in something that turned out to be just spin, and not getting compensated for the real value of what they were providing. But there are also all the goto examples of startups that have done well
> It is basically impossible to become a world-class software developer if you start after you’ve achieved career success in another industry.
Does anyone have thoughts on why this would be? I can’t think of a counter example to contradict this statement. But it doesn’t seem like there’s any fundamental reason why you couldn’t become world class starting a bit later. Is it just that it’s tough to put in the hours once you’re in your 30’s? You're more susceptible to carpal tunnel and neck pain and it just physically gets more difficult to put in the same number of hours.
I’m not sure I agree with the premise to begin with, it’s tough to be a world class developer to begin with, so it’s not unexpected that it’s hard to find counter examples, but I’m pretty sure there are some.
The only thing I can think of as an explanation would be that it’s hard to “think like a computer” if you aren’t use to it, and it’s an essential skill, but I doubt it’s unlearnable past your 30s.
"World-class" is such a vague term that it's neither true nor false. Does it mean "capable of becoming a staff-level engineer at a FAANG" or does it mean "capable of being Don Knuth"?
The biggest problem is motivation; people who become software engineers later in life are usually the people who didn't really want to become software engineers, they just fell into it after their initial cooler career choice didn't work out. So they're going through the motions to earn a paycheque. If they had the personality and drive to become really amazing software engineers then they would have done that in the first place.
Knuth is a talented mathematician who likes algorithms. I'm not sure I would think of this as the standard of "world-class developer": it's special but in a slightly different direction.
I was 32 when Hofstadter's "Gödel, Escher, Bach" inspired me to learn coding. I was not a world class developer but had a pretty successful 30 year career.
This depends how you define "world class developer". If you define it as someone who's primarily on the software engineering side of things then the transition is rare because almost no one switches from being successful in one field to being an IC in another field. If you just mean someone who's good at programming required at his job, plenty of people have e.g. learned astrophysics and then become good at programming astrophysics simulators or whatever.
Having seen the unusably poor quality of some academic simulators, gonna have to quibble with that last claim. Academics usually seem to learn just enough programming to get the computer to spit out numbers that support their pre-existing beliefs, and then stop. Does the program actually work? Do what it claims to do? Run on anything other than the researcher's laptop? Not considered relevant.
Computing changes a lot faster than most other industries do, which means that you have to be continuously learning to a greater extent than in other fields. And that knowledge compounds. Yes, some stuff I learned 30 years ago is now obsolete, but a surprising amount isn't. The knowledge I'm learning today doesn't replace yesterday's, it augments or builds on it.
So if you don't start young, it's very hard to catch up. You can learn software development alright but you'll never be world class because all your time will be going into learning the new stuff, not the older stuff that's still important.
I'm not sure about the original author's intention but it seems plausible to me. Software requires an absurd degree of detail orientation and willingness to dive down a dozen abstractions, far away from the problem you're nominally trying to solve. There's so much detail in between processors, memory, operating systems, programming languages, system architectures, if you've achieved success in some other industry it does seem pretty unlikely you'll put in the work to master these concepts, all of which you kind of need to be considered a "world-class developer" imo.
On the Twitter/X user experience: I have a lot of right leaning friends in the indie author space. There's been more of those people coming back to the site. Those who stayed on report a better experience--more interaction, less shadowbanning, quicker fixes when the mods screw up. I suspect the opposite is happening in left-leaning twitter spaces, and the site is staying even on users on net.
That would only work if there is an equal number of right and left leaning users to draw on. Since the most "online" demographics tend to be young educated urban people, who are more left leaning on average, I wouldn't expect that to be the case. Though it may be one of those things where what matters is actually a small number of highly engaged users, and Musk is able to make more money from the smaller pool of right wing whales.
There's probably way more "right leaning" people available given that left and right in pre-Musk twitter terms were heavily tilted towards the extreme left being defined as normal. Musk himself was like that, right? He tweeted once that famous meme where someone starts out being center-left and then stays still as others dash way further to the left, meaning he ends up on the right. These two camps aren't a 50/50 split by any means.
Also given that one of the chief complaints about Facebook is the preponderance of older people, it doesn't seem like lack of right wing people would be a problem, especially as in many parts of the world older and more conservative people are richer.
My impression is that normies tend to not have strong detailed political opinions, instead mostly going with the flow. Sure, if you ask them out of the blue, the lefty ones wouldn't articulate the hottest woke talking points, but neither would they be bothered when encountering those in the wild, being on the right side of history and so on.
After all that I’ve read about Mr Musk there is nothing I would buy, nothing I would say nor, anything that would make me give him any more of my time.
Musk is a Tigger. A personality type which leads Rabbits to become a bit too keen on Unbouncing him.
One day he will inevitably say that he knows how to climb trees and get stuck at the top of one with Roo on his back. It's something of a worry that with AI humanity could be playing the role of the baby kangaroo.
> It will use Tesla cars, driven by Tesla employees. In my view this is basically an underground Uber system, but it will probably have more expensive fares to regain the capital costs of building the tunnel (Boring Company is paying for the tunnels, and casinos are paying for the stations, they do NOT have funding from City of Las Vegas AFAIK). But this expensive Uber system is exciting??
It is known that the goal is not to have Tesla employees drive, but rather to use self-driving cars. This should be much easier than achieving general self-driving capabilities.
The capital costs of building the tunnels appear to be very low compared to those of subways in the U.S. I can see the system becoming profitable.
If anyone thinks that a public transport system with private vehicles is ridiculously inefficient, think of the number of times you've been on a whole damn train thundering down the line at less than ten percent of its passenger capacity. (For me it's a lot, anyway.)
Basic computation: a car with one person going at 30mph requires about 30 meters of buffer, plus 3 meters if the car. In terms of space usage that's worse than a single person per train car (or an entire New York subway train with six people in it). And that's for a road at full capacity *and also* ignoring roads being empty ever *and also* ignoring the massive space costs of last mile roads and parking.
If you don’t ignore roads being empty ever, then you also have to not ignore train tracks (and stations) being empty ever, and that sounds like a much harder comparison to put numbers to.
For that matter, if you count the required buffer for cars, should you not count the required buffer between entire trains, not just between the train cars? I *think* trains might still “win” that in some sense, but still, it would significantly reduce their density.
Similarly about last mile roads and parking. If you include that for cars, then you need to include that for trains as well, and the cost of “last mile train tracks” is in a sense infinite, because you can’t actually make train systems that go as close to your house as cars do.
(The closest train to my current position is actually about half a mile away, so it technically covers about half of the last mile, but my car is about 20 times closer and goes more places. And I live in a high density city, good luck getting even close to that ratio in smaller towns and villages.)
The efficiency of a transport system isn't best judged by its downtimes, but by its peak times. Or at least, the downtimes aren't the point of a rail system - the peak times are. (Roads are designed to work best at downtimes.)
"If neither company had outstanding stock and therefore they both had a market capitalization of zero, how likely would you be to conclude that owning Ford would be 17 times worse than owning Tesla?"
on the topic of relative valuations, the couple comments in the article above (haven't read through the full commentary stack, no time at moment) haven't mentioned the most important point: Ford is in the ICE car business (9x% of revenues, 100%+ of profits) which is likely to be completely defunct or no more than a small sliver of the overall new car market within 15-20 years. Tesla is in the EV business which is the future (and its EV business is actually profitable). Ford is not doing a great job moving towards EVs, and their upcoming UAW contract will likely ensure it never does. Don't you think based on that alone Tesla deserves a huge valuation premium? (btw. i don't hold either stock, have never been able to convince myself Tesla isn't overvalued, but Ford is also potentially hugely overvalued / a melting ice cube)
The back and forth here on whether Musk is "smart" reminds me of the same debates about Chat-GPT. Some people claim that it/he is confronted with novel (or at least, novel to them) difficult problems and quickly providing correct answers. Others claim they lack real understanding, and just parrot what someone else came up with, make a statement that is too general to be useful, or just give a response that is wrong (but maybe you need some expertise to see why).
"Starlink's terms of service include a Mars clause: Users must agree that Mars is a free planet unbound by the authority or sovereignty of any Earth-bound government."
So who will govern Mars, then, and will the 'users' agree to be ruled by God-King Elon? What about if there are disputes over that, and people want to get appeals heard on Earth?
And of course, all an Earth government has to do in a dispute with Mars is shrug, say "okay, you don't want anything to do with us? have fun trying to be self-sufficient really fast when we're not sending any more supplies to you".
Who says the supplies would be coming from a government in the first place?
Earth governments could potentially block or impound supply shipments to Mars coming from the private sector, but as rockets can launch theoretically from anywhere and govs have poor track records of being able to impose reliable sanctions even on third world countries ... well.
If Mossad don't like you selling things you shouldn't, then they'll get you.
The very tangled and indeed nasty "Arms to Iraq" enquiry, where a company being prosecuted for selling arms to Iraq had the trial collapse because it was revealed the British government had been secretly helping them sell arms to Iraq:
And yet North Korea has MacBooks ... yeah you can imagine governements sanctioning Mars just because (the rationales used for sanctions on Earth wouldn't apply here, presumably). But that doesn't mean they can do it successfully.
Governments have been pretty good at making sure advanced rocket technology, infrastructure, and supply chains don't make it to third world countries. North Korea has put a significant fraction of its GDP into rocketry for decades, and can barely launch small satellites. Third-world countries building Mars transports is not going to be a thing for quite some time, and first-world governments can stop them from buying such vehicles from first-world contractors.
The "linear development" process that Fluffy Buffalo described is basically good for two things:
1. Spreading development across congressional districts to win support for the project, at least in theory - in practice, it matters a LOT more to have one well-positioned Senator or House member to drive a project forward. We have SLS in large part because Dick Shelby dominated in the Senate Appropriations Committee and make it a priority, and Europa Clipper because John Culberson went hard to bat for it in the House.
2. Avoiding a high profile failure, like a prototype exploding on the pad. Instead, you keep all the failures and explosions hidden inside smaller facilities far removed from the public eye, and (hopefully) by the time it reaches the launch pad it's near-certain to not fail (and if you have even an inkling that it might fail, you take it back from the pad).
Re: Twitter/X, I can't remember the source now, but supposedly Musk decided to buy the platform because it seemed stupid to him that even though he was the richest man on earth, he still couldn't say what he wanted. So if X loses money but becomes less restricted in terms of censoring people (or, less charitably, censoring Elon Musk), maybe he'd still consider that a win.
On the Boring Company, it's worth pointing out that there are manufacturers of mass-produced TBMs - these are ones in the 30cm to 3m diameter range (yes, they really are TBMs; they're the same technology as the big ones). These are used for water and gas pipes and sewers when they are laid in locations where it's not workable to just dig a trench, drop the pipe in, and cover it over (most typically under rivers, but also under existing large buildings, under airport runways and probably some other cases I'm not thinking about).
One reason why the claim that he could make the tunnels cheaper by making them smaller produced quite a lot of doubt was just how little price benefit there is from the competitive high-volume market for small-diameter TBMs (large TBMs are usually Germany's Herrenknecht in Europe, Japan's OGITEC in East Asia or CREG in China, there are about 20 manufacturers once you get down to 3m diameter). The Boring Company isn't touting a technical breakthrough in TBM technology the way that, say, SpaceX has recoverable rockets.
If they were going into the big-tunnel space to compete with what is essentially a geographically segmented monopoly, then I'd believe that they could come up with organisational innovations to make TBMs cheaper (much as Falcon 9 was noticeably cheaper even before it was recoverable, because it was competing with a hyper-conservative oligopoly culture) but if they want to hit orders of magnitude, I think they will need a technological breakthrough, and there's no indication that they're doing anything genuinely novel.
"the claim (still not proven, but plausible) that Community Notes wasn’t really a Musk project"
The code for it is actually open source so you can just see the history here: https://github.com/twitter/communitynotes (it started in 2021). The code has changed and has been improved post acquisition but the core idea and algorithm of what makes Community Notes special is still very much not an Elon thing.
"Community Notes"/Birdwatch became more prevalent when he acquired it but that's simply a coincidence of the project fully deploying on the frontend side slightly before and during the acquisition process, thus why people think he was behind it. AFAICT his main contribution was renaming it to Community Notes as he wanted to move away from bird-themed puns
Missed in the discussion of market caps is that Tesla is much more vertically integrated than Ford. So a true comparison would add the value of Ford to the value of it's various suppliers.
One fun fact I learned about Musk was that his maternal grandfather emigrated to SA from Saskatchewan, because he was worried Canada was becoming too socialist. He was a pilot and in his spare time would fly around looking for the lost city of the Kalahari.
Here is an abridged 2022 John Carmack interview about Elon. Carmack attributes Elon’s success to his level of commitment to go all-in on highly risky businesses where he could have gone bust.
I met him as I was starting to do Armadillo Aerospace. He came down with his right-hand propulsion guy and we talked about rockets and what can we do with this. It was specific things about how are our flight computers set up, what are different propellant options, what can happen with different ways of putting things together.
In some ways he was certainly the biggest player in the alt space community that was going on in the early 2000s. He was the most well-funded although his funding in the larger scheme of things compared to NASA was really tiny but it was a lot more than I had at the time.
But it was interesting, I had a point years later when I realized my financial resources at this point are basically what Elon's were when he went all-in on SpaceX and Tesla. I think in many corners he does not get the respect that he should about being a wealthy person that could just retire and [instead] he went all-in where he could have gone bust. There's plenty of people, you look at sad athletes or entertainers that had all the money in the world and blew it, and he could have been the business case example of that.
I have a great deal of admiration that he was willing to throw himself so completely into that, because in contrast with myself I was doing Armadillo Aerospace with this tightly bounded [limit], it was John's crazy money at the time that had a finite limit on it, it was never going to impact me or my family if it completely failed. I was still hedging my bets working at ID software at the time when he had been really all-in there and I have a huge amount of respect for that.
The other thing I get irritated with is people that say, “Oh Elon's just a business guy, he just was
gifted the money and he's just investing in all of this,” when he was really deeply involved in a lot of the decisions. Not all of them were perfect but he cared very much about engine material selection and propellant selection. For years he'd be telling me get off that hydrogen peroxide stuff, liquid oxygen is the only proper oxidizer for this.
The times that I've gone through the factories with him we're talking very detailed things about how this weld is made, how this sub assembly goes together, what are startup shutdown behaviors of the different things. He is really in there at a very detailed level and I think that he is the best modern example now of someone that can effectively micromanage some decisions on things on both Tesla and SpaceX to some degree where he cares enough about it.
I worry a lot that he’s stretched too thin with the Boring company and Neuralink and Twitter
and all the other possible things there. I know I’ve got limits on how much I can pay attention to and I have to box off different amounts of time.
I look back at my aerospace side of things, and it's like I did not go all-in on that. I did not commit myself at a level that it would have taken to be successful there. It's kind of a weird thing having a discussion while he's the richest man in the world right now. He operates on a level that is still very much in my wheelhouse on a technical side of things.
>> I understand the frustration... but my impression is that space exploration is one of the fields where very thorough, very systematic planning with very conservative change cycles is the most promising approach to get something that works at the first attempt - even if it takes longer and costs more than planned. Compare the JWST to the most recent "Starship" launch for illustration.
> This would sound plausible, except that Musk has succeeded by doing the opposite. I think this is why so many people are in love with Musk: he’s proven that valuing good ideas, moving fast, and not having bureaucracy can work, sort of, in a weird little bubble of his own creation. Yeah, the first Starship exploded, but most people predict Starship will eventually work, and when it does it will be a much more impressive feat of engineering than JWST or anything else created the “proper” way.
Weirdly, you seem to have misunderstood your own point of disagreement with Fluffy Buffalo. Musk hasn't succeeded at the only goal Fluffy Buffalo mentions - succeeding on the first try. You go on to claim that Musk is a counterexample to Fluffy Buffalo's ideas while conceding that he isn't. (?!)
The actual point here is that succeeding on the first try is not a valuable goal, not that planning isn't helpful in achieving it.
What isn't mentioned is how bureaucratic the organisations are that Musk's two most succesful companies compete with. Lack of good competition helps a lot. Explains his lack of success at Boring comopany, Twitter and that brain implant company.
Then what he did with Tesla is obviously impressive, but is it earth shattering? The company lost money for well over a decade. With negative margins of < -10% for most years while:
-Selling higher margin luxury cars.
-Underpaying and overworking employees.
-Delivering cars of subpar build quality.
-Providing poor after market service.
-Benefitting from billions in subsidies.
-Not even making the battery cells (the main tech) themselves
Other car companies looked at this and said to themselves, well we got unions to content with, and measures of quality control that are much higher, why on earth would we try to compete here and lose even more money?
Now that battery cells have come down enough in price and making EVs has become profitable, competition is flooding into the market.
Seems like Musks main skill here was keeping Tesla financially afloat and keeping the hype up, allowing him to sell a lot of stock and debt at high prices (exceptionally low interest rates helped a lot here as well). Being succesful at SpaceX was probably a major help to this.
Musk hasn't just beaten his competitors, he's beaten everyone else in Silicon Valley and around the world who could have tried putting themselves in a position to beat those same competitors. IE if it was easy, why didn't anyone else do it?
(cf Andrew Beal, Better Place, and all the other startups that tried to disrupt cars and rockets and failed)
Not the guy you replied to, and I haven't read all the comments and shit, but I've always seen Musk as a guy who has one particular superpower, which is an ability to raise practically infinite capital.
Even for twitter, which never really had a model for being profitable and he loaded up with debt to the point that it will definitely lose money, and which Musk has admitted he overpaid for and very clearly tried to back out of buying, he got a bunch of rich people to put in money.
Like, imagine you agree to buy an apartment building for way above market price, then very loudly try to wiggle out of it for months, then try to get other people to go in for a share, telling them how good a deal it is. Would anyone invest? Well if you were Elon Musk they would! Dude could sell people a 1% share of a bag containing $100, for $5.
If you and Musk both presented a plan for some new business model - producing electric planes or something - you'd have a really detailed proposal, a shitload of tech specs, a SWAT team of PhDs to nerdily explain all the nuances, and Musk would show up like "we're gonna cut costs 98% by hiring the Cat in the Hat! Here's a picture I drew in crayon" and he'd still raise like a trillion dollars to your 0.
And then they'd hemorrhage money for 15 years, crash 47 test planes, repeatedly break their own promises, etc, but at the end of the day there'd be a working electric plane.
The really dramatic/self-important description of the world of finance is that it's how society decides how to allocate resources to different endeavors.
So ... is it useful to allocate a shitload of resources to "whatever Elon Musk wants to do"?
Argument in favor: yes, look at this stuff he's done.
Argument against (assuming the stuff he's done is good): anyone who gets that much capital could do a lot of stuff. If you, Scott, got as much capital as Musk did, you'd probably totally transform how psychiatry works in the US - you'd get everything on your most fanciful wish list of "how I wish psychiatrists and adjacent people operated" and then some.
Argument in favor (assuming the above is true) it's still useful to have someone with the "raise infinite capital" cheat code on. The regular business world is so mired down in bureaucratic nonsense and conformism that randomly elevating one guy with slightly above replacement level ideas, to pursue those ideas, is still better than whatever that capital would otherwise be allocated to.
Perserverance. Since founded Tesla lost money with sizable negative profit margins for 16 years. And for 15 since Musk invested in it. And this is while running Tesla almost as a cult. Which probably lowered costs quite a bit. Which is not a luxury competitors had. (which is why I think also running SpaceX at the same time had a major synergistic effect here, helping Tesla survive).
Typically start ups start making a profit after 6-7 years and are shut down if they can't make it. For example Amazon lost money for 7 years. Google for 3-4 years, Facebook for 5 years.
And I don't think Tesla has beaten all competition. A lot of good EV's with long ranges are coming out now, that often have better build quality than Tesla. Now that it has become profitable to actually sell EV's. I have some friends who were in the market for an EV and they did not even seriously consider a Tesla (and not primarily because of Musk's politics).
good comment, and the first i've seen that touches on the "cult" aspect of Tesla. Tesla's retail investors are notoriously rabid and worship the ground Elon walks on. They provided a backstop to Tesla's market cap even when the company was generating losses quarter after quarter. They would 100% buy any promise Elon pushed and support it unconditionally, despite reality differing from the goal. Lots of examples here, but the most damning is Tesla's full self driving, which is going on like 8 years of 3 months maybe, 6 months definitely.
By far the largest reach of the Tesla influences is X, which I would argue controlling the narrtive of Tesla has existentially import. (it's possible Tesla has become large enough from a market cap perspective that only a "key person" risk event would cause a collapse)
BTW I am not arguing here that making Tesla a success was trivial. Just that it should be seen within context. I am pushing back against the "he did something that nobody thought was possible" narrative. The competition did think it was possible to make electric cars, just that it would not be profitable. So they waited until it did. As is happening now.
Although arguably Tesla did cause making EV's to become profitable sooner by probably a couple of years due to economies of scale in battery cell manufacturing.
Musk's unofficial marketing arm (with X being the platform with the largest reach/influence) is by far the most impactful driver of Tesla's ability to stay afload (via high market cap) for so many unprofitable years. How many company customer bases report best-in-class owner satisfaction with below average quaity and worst-in-class reliability? Musk's fanboys reached a critical mass and were able to keep Tesla's stock price elevated through the existentially precarious periods, ensuring Tesla could always raise more capital.
Also the birth of hedge fund manager / EV evangelist CNBC promoters (think Cathy Wood or Chamath) finding it easier to get an audience preaching faith in a company's future rather than any kind of logical or coherent argument based on typical company success metrics. Tesla is the posterchild for this sort of marketing
My view is the vast majority of discussion from the original book review to the following related discussion misses for the forest for the trees. Musk is almost entirely a story of access to retail capital providing a longer runway than basically any other founders. Tesla success means he can tap capital markets whenever he wants/needs based on his reputation of "never bet against musk" etc
The ledger of successes/failures or smart/dumb things Musk has done we can argue about, but the main story is Tesla. Without it nothing else is really possible.
Success/Smart
* UPenn and let's assume he was accepted to Stanford's graduate Physics program
* Zip2 - for sure benefitted from macro timing, but still he built something from nothing and got the payout from Compaq
* early investor in Tesla - sorta seems like a no-brainer since the founders already had a quick and sporty prototype, but EVs were seen as a losing bet at the time so this was definitely a source of Elon alpha
* SpaceX / Starlink and learning enough about rockets to be additive from an engineering perspective. Neither of these are novel ideas
* ability to consistently receive subsidies for all of his business ventures
Failures/Dumb
* Paypal/X.com period where he nearly killed the company by refusing to implement any compliance measure (chargebacks alone were totaling $2,300/hr as a result) and pushing an incentive program that the company could afford. Elon was ousted from CEO role and replaced with Thiel as a result
* SolarCity was quickly approaching bankruptcy in 2016 so Musk offered to merge SolarCity and Tesla, which was pretty clearly a musk bailout at the expense of Tesla shareholders. Very beneficial to the narrative of Musk to not have SC bankrupt
* Twitter - clownshow as has been well documented
* publicly committing securities fraud by faking a private buyout (taking Tesla private at $420)
* Neuralink / Boring Company - time sinks that are possible because of unlimited access to capital. Nothing these companies are doing is novel or profitable
* calling a cave rescue lead a pedo
* saying chess is a "too simple" after his feelings got hurt from a disagreement with Gary Kasparov over Ukraine
* using Tesla resources to build a home in Austin (ongoing criminal investigation)
Don't forget that he *also* broke security laws when acquiring his initial stake in Twitter in early 2022. (He filed the disclosure forms well after the required deadline and when he did finally disclose, he claimed on the forms that he had no intention of acquiring Twitter, immediately before he proceeded to do just that.)
`Lost money for a long time' would also seem to describe Amazon, or Facebook. Maybe even Google, although the dawn of Google is before I was an adult and paying attention.
Amazon already had positive operating margins in 2002. While being founded in July 1994. So 7 years of losses, give or take. Google was profitable almost right from the start.
Facebook was unprofitable for 5 years after founding.
Tesla was founded in 2003, Musk invested in it in 2004. They became profitable in 2019. So 15 years of losses give or take.
> AI experts, is this a big deal? Can other AI teams not access Twitter data through the public web? Is it a substantial amount of text compared to other corpuses? Is the structure (280-character blurbs written by morons) a limiting factor? Or is this a genuine treasure?
Not really an expert, but... at least before the API changes (and possibly still afterwards), Reddit comments were free and less-restricted than tweets. Their upvotes occasionally even correlate(d) with truthfulness/usefulness (see: why we google "best [thing] to buy reddit" instead of "best [thing] to buy").
Also potentially relevant: Twitter is no longer really accessible through the public web since Musk made it impossible to read tweets without an account, and I think at the same time also limited the amount a single account is allowed to access per day?
1) I've had to block FAR more people to keep my feed as consistent as it used to be, somewhere on the order of fifty a month when the number was previously less than ten. Part of it's the algorithm seemingly getting shittier. Another part of it is the top of every thread filling up with blue checks, who mysteriously tend to post the most asinine corporate boot-licking takes. A Twitter subscription service is not a BAD idea - in fact, this was in the works before Elon took over, to allow people to edit their posts, as far as I know. But one that allows people to boost their engagement just creates a bad incentive structure where the people dumb enough to piss away eight dollars on a free platform are promoted to the top. The best blue checks are people who bought it for the quality of life features, or the cynical businessmen looking to get an edge - both of which are in the minority. It's like Linkedin-lite, and Linkedin is the worst social media platform.
2) I get at least one cryptocurrency spam message a day under Elon's Twitter, that I didn't get before. At least half of the new accounts that follow me are bots with GPT generated bios and AI generated profile pictures. If his goal was to fix Twitter's bot problem, he's failed. He's partially not to blame for this - AI has expanded its capabilities far more than we expected in the last year - but at the same time, I don't see the same frequency of spam on any other social media platform.
3) There have been more major Twitter outages. This might just be anecdotal. But at the very least, the one where he pretended to have restricted the number of posts people can see per day to cover up a very loud and obvious engineering failure was entirely his fault. If you're trying to build up Twitter as something committed to free speech and transparency, it's not a good look to accidentally cripple it for close to 24 hours and then just lie and say you did it on purpose. If there are more incidents like this, more people will leave the platform, and because he lied about WHY it happened I haven't seen any evidence that he's taken steps to keep it from happening again other than reverting the changes.
There's more to it. Anecdotally, I feel like the environment has become more hostile. But if you want to boil down the mismanagement at Twitter and my current gripes with it to actual, quantifiable issues, that would be it. Twitter Blue cannot be tied to the promotion of tweets. If they want an organic platform and not a soulless collection of corporate robots like Linkedin, they need to decouple engagement from the amount of money you pay. They need now more than ever to have hundreds of people moderating the platform BY HAND, not with AI. I would not mind him laying off 90% of his engineers if it meant he brought in half as many moderators, but as it stands the platform is too vulnerable to spam. And if they don't want to lose users every time someone deploys shitty code and bricks the site for a day, they need to be actually transparent about what actually broke, instead of coming up with a stupid lie.
I remember reading somewhere that publicly diagnosing other (famous) people who aren't your patients is a violation of professional ethics. Is that not the case?
It's called the Goldwater rule (and primarily applies to political figures, though it's stated for any public figure). Presumably this heavily qualified "has some symptoms of X but not others, but I don't know what he doesn't share in public" is sufficiently careful for it.
Okay, and I suppose Musk opened the door on this a bit by discussing his mental health in public. Theyre not really diagnosing the person, they're diagnosing their online persona. It still doesn't sit well with me for some reason.
>I’m not sure whether this means that everyone else is an idiot with a pointless bureaucracy fetish, or that only a few very special people like Musk can make the non-bureaucratic version work.
You could ask Stockton Rush that question.
In other words, heavy survivorship bias going on here. We remember successes, not failures unless they're especially notable failures. Bureacracy is the institutional way of remembering failures.
That's very similar to my take. It's also how I think about government. To me, the best government is one lead by a singularly amazing leader who has significant power and control. That's also the worst government, depending on the person in control. And either one has a really hard time passing control to someone else without losing what makes it special or just totally falling apart.
"I’m not sure whether this means that everyone else is an idiot with a pointless bureaucracy fetish, or that only a few very special people like Musk can make the non-bureaucratic version work."
"Pointless" seems like going too far. Maybe everyone else is a lazy bureaucrat with a perfectly self-interested bureaucracy fetish!
Anecdotally I often parse bipolar but not bipolar as ADHD or something like it: tendency to excessive mood swings, extended "stuck on" or stuck off"", but in response to stuff exaggeratedly and spiralling, not actually random.
I don't care so much whether this is enough to be a medical condition or not, more whether "things you might expect to be under an adult's control but don't seem to be are harder to keep under control for the person, or whether the person doesn't care, or somewhere between". He seems to have SOME passionate intense work, and SOME giant pissy fits about things. But he doesn't seem to acknowledge that having angry meltdowns at people is a problem.
Re ability I expected him to be smart but not the best engineer at somewhere like SpaceX. I think that's probably still true. More like, all the OTHER things that make a driven leader good like breadth of knowledge, common sense, drive, vision, where is he on a spectrum between "good at things he knows the right amount about" and "petty and aggrandising at anything where he's even slightly out of his depth"
> This doesn’t even have to contradict Musk’s ADL claim! If Parag or whoever employed a thousand censors to keep the ADL happy, and the ADL becoming unhappy cuts Twitter profits by 60%, then there’s a strong business case for those censors!
I am less inclined to think of the ADL as causal than most others here, it seems like. I've worked with advertisers and they all said they were trained to be very careful of the brand presence, to the point where it looks panicy and puritanical to outsiders. I think if you took away the ADL specifically but left everything else the same, the world we see would not change. Here's why:
I used to work on the web site for a company that sold something bland and generically useful -- let's say refridgerators. We were beating the other refridgerator companies in our marketplaces, partially because our refridgerators were well made but mostly because our ad people were good at placing ads, targeting ads, and winning ad auctions without losing money. I did a lot of the SEO, and wrote a lot of the landing pages our ads pointed to, so I wound up talking to them a lot.
They were pretty chill to talk to, and clearly great their jobs, but in their work they were incredibly conservative. Any time our brand appeared next to anything even faintly controversial they went on the war path. Hate speech, sure, but also nudity, drugs, discussion of abortion; "anything that might make a suburban mom uncomfortable," as one of them summarized it once. If we placed an ad in a newspaper or something and it ran alongside a piece about a gang shooting, one of the ad folks would get on the phone and yell at the newspaper until they apologised, took it down, and refunded us.
It never made sense to me, but when I'd try to ask about it they'd mostly brush me off; this was how they were trained, it clearly worked, so I had to trust them. Finally, one night at the end of a company party I was drinking with the guy in charge of ads for north america, and I decided to rise the issue directly. "Why _don't_ we run ads against porn?" I asked. "After all, people who jerk off need refridgerators too!" "Yes," he patiently replied, "but then our brand is 'the refridgerator for jerk-offs'. We won't get enough sales to be worth that." Then he said a lot of stuff about brand awareness, but it wasn't phrased as a searing burn so I don't remember it precisely.
As far as I know the ADL never called us, but if they had our ad folks would have reacted right away. Not because they were scared of the ADL, but because they were scared of the same things the ADL is, plus a bunch more things besides. The ADL has probably organized some boycotts in their time, or tried to, but my guess is that most of their work is more like raising awareness, and the _modal_ ADL interaction is them calling a brand to say "Hey here's a picture of your ad running next to an ASCII swastika" and the brand replying "Thank you so much, we also care about that for self-interested profit reasons and will take action immediately," with no coercion even implied.
Now, separate from all that: are the advertisers actually right? Honestly I still don't believe it, and I could imagine a smart engineer seeing all this from the outside and assuming it was a lie meant to cover up something more sinister. But, at least in my corner of the industry, it wasn't.
> I thought there was a week or two when everyone threatened to switch to Mastodon, then found they didn’t like Mastodon and went back? So where did everyone go? Was it Mastodon after all? Facebook Threads? Blue Sky? Or did they all start learning to paint and spending time with their friends and families?
Personally I started spending more time bouldering and bicycling, and substituted Manifold Markets as my source of "what's going viral on Twitter right now" news.
In fact in some cases Manifold led to a better sense of what's actually going on in the world, such as in the room-temperature superconductor markets where people who read Twitter all day gave it way more credibility than was warranted, leading to a big profit opportunity.
I find it so interesting how resistant people seem to be to the idea that Elon Musk is very smart and capable and that has contributed positively to his success. It seems people are looking for any possible thing they can point to to say "see he's not that smart after all, he just got lucky". Maybe it's a response to the Musk-worshipping that was going on around 2018-2021, but it's a weird manifestation of a 'God of the Gaps' style argument.
Yes! Well-said! This is thing that bothered me about a number of comments too. "Look at his imperfect bathing average! Clearly a failure! Mentally ill too!" Etc.
I just don't see why the conclusion drawn from those incidents is "Musk must not be so smart despite all the evidence pointing to him being smart" rather than "It turns out very smart people are capable of doing things that seem very dumb". We have plenty of examples of the latter case being true. In fact in the case of Hyperloop, there must be plenty of very smart engineers/scientists who all did very well on their SATs and got into great grad school programs who devoted great efforts to working on Hyperloop
My immediate reaction is "Why are we debating Musk's intelligence?" Yeah, you have a higher IQ than Musk. Who cares; have you built a successful rocket company? A little more subtly, everybody knows you can build a Starship-class rocket, and more or less how to do it; we did it fifty years ago and the Soviets came close. The tricky part is actually doing it.
The point that comes out that is unusual is that he has the ability to absorb a lot of detailed knowledge and work with it and using that to both impress and manage the herd of extremely good and highly motivated people he's recruited to do the work.
I did make a joke (not on that post) about all the non-defective Elon Musk clones going off into a Galt's Gulch retreat. It seemed like a convenient explanation for his energy, ubiquity, and erratic behavior. :-)
I am well aware of the existence and implications of perchlorates in Martian soil, I have even put a fair degree of thought into how to put them to good use. But I don't find it necessary to mention them every time I talk about any aspect of Mars colonization, and I don't think that's a particularly good metric to determine the seriousness of a discussion.
If your model is that perchlorates make Mars colonization impractical without some new miracle, but the Space Nerds don't know about the perchlorates because they're just ignorant fanboys, then no. We know about the perchlorates, we don't think they make Mars colonization impractical, we've got reasons for that, but we've also got reasons to talk about a *whole lot* of things that aren't soil perchlorates.
If you're going to make a hermetically sealed colony that has no interaction with the outside atmosphere or soil, then Antarctica is back on the table.
> AI experts, is this a big deal? Can other AI teams not access Twitter data through the public web? Is it a substantial amount of text compared to other corpuses? Is the structure (280-character blurbs written by morons) a limiting factor? Or is this a genuine treasure?
More data is basically always better in the LLM game, but having high quality clean curated factually grounded data is even better. See for example Textbooks Are All You Need: https://arxiv.org/abs/2306.11644
With regards to the Vitamin C problem of building tunnels on earth- Rhubarb needs very little light to grow extraordinarily quickly. One cup of rhubarb is 16% DV Vitamin C. I think that the scurvy problem with Chicxulub 2 would be nil.
> For example, I’m surprised to hear this! I thought there was a week or two when everyone threatened to switch to Mastodon, then found they didn’t like Mastodon and went back? So where did everyone go? Was it Mastodon after all? Facebook Threads? Blue Sky? Or did they all start learning to paint and spending time with their friends and families?
Look how far down the replies you have to go before you see something besides explicitly anti-vaxxer gibberish. Blue check-dominance is atrociously bad for conversations that take place at IQs above room temperature and involve/allow many respondents.
Right now there's not a good alternative AFAICT, so mostly people are just spending less time there and moving to non-identical websites (e.g. Reddit, Facebook, Instagram).
I am interested in conversation. I am not interested in conversing with people who think that getting vaccinated for COVID-19 is bad, on that subject, because that position is extremely, mindbogglingly stupid and based in 0 evidence. I don't care how many "mind killing phrases" are in that sentence, I also don't want the top fifty responses to "belief in spherical Earth correlates with IQ" or "belief that the 2016 and 2020 US Presidential elections were legitimate correlate with IQ" to be "that just proves that IQ doesn't make you smart." I cannot have a useful conversation with such people, because I would first have to spend 18 months slowly dredging their brains of whatever conspiratorial sludge has currently polluted them so that we could have an evidence-based conversation.
2. Is a 4 percent increase in the percent of the labor force that has some form of disability (the highest in 15 years, and the largest 1-year increase in disability on record), in the exact period during and immediately following the vaccine rollout, "based in 0 evidence"? https://fred.stlouisfed.org/series/LNU01374597
3. In the Pfizer study that led to the emergency approval of the vaccine, the vaccine group had around 4x as many adverse events and 7% more deaths than the control group (actually, 21% more deaths, but Pfizer explicitly excluded several of the dead vaxxers from their results sectiuon, as they mention in the study). Notably, 20% of the vaccine group deaths were due to heart attack, vs 0% in the control group. Is this "based in 0 evidence"? What else would possibly constitute evidence? https://www.ncbi.nlm.nih.gov/pmc/articles/PMC8461570/
I know enough to know that VAERS has fundamentally junk data gathering methods and can only be used as a direction of possible investigation, not as evidence in itself, so I'm definitely not interested in digging through your other two points to find if they are equally cherrypicked.
70% of the global population has gotten vaccinated and all you can do is point to this mush. If it was actually killing and maiming people in any appreciable amount doctors - in the generic sense, not like, one specific doctor - would have noticed by now. I know you don't trust doctors, but that's the thing: I don't want to argue up to "actually, the medical establishment that has overseen decades of rising life expectancy is not wildly incompetent enough to miss that something administered to seven out of every ten humans is actually really bad." That's not a good use of my time. I won't learn anything. Maybe I could persuade you but I don't care about you.
> VAERS has fundamentally junk data gathering methods and can only be used as a direction of possible investigation
You've just argued that *the primary safety monitoring system* for the vaccine has "fundamentally junk data gathering methods." And you think this bodes well for the safety of the vaccine?
> I'm definitely not interested in digging through your other two points to find if they are equally cherrypicked.
Not surprised, pretty typical behavior for your type. Calls the official safety monitoring system's data "cherrypicked," and then proceeds to cover their ears and yell "LALALALALALA I CAN'T HEAR YOU!!!" when presented with evidence that goes against their beliefs. And then somehow still thinks that their views are based in "evidence".
> 70% of the global population has gotten vaccinated and all you can do is point to this mush. If it was actually killing and maiming people in any appreciable amount doctors - in the generic sense, not like, one specific doctor - would have noticed by now.
There were more deaths in 2021 after the rollout of the vaccine than in 2020 without the vaccine. In the US alone, millions are dead. Insurance companies reported a vastly higher death rate (up to 2x higher than baseline) *specifically for young people* in 2021 immediately following the vaccine rollout, *not* among older people, as would be predicted if the 2021 deaths were primarily due to COVID itself.
> is not wildly incompetent enough to miss that something administered to seven out of every ten humans is actually really bad
No, incompetence alone doesn't explain it. For most, it's a mix of the general hyperconformity and fear of standing out from the crowd that all humans have, along with fear of losing your medical license or your career being ruined if you disagree with the mainstream narrative due to backslash from people like you whenever evidence contrary to your viewpoint is presented. And then of course there are the genuinely evil and malicious at the top who just wanted people to die (or didn't care how many died as long as they sold more shots), but they are not the majority.
Since you are exactly the kind of person that would have just gone along with those atrocities - "Oh, I'm sure ze authorities are just helping ze Jews, Helga, zey are good people. They can't be doing anything bad, surely one of them would have spoken up about zis if it were true!" - I don't care about you either. Be sure to schedule your 7th booster. And don't worry about that weird cramp in your leg or that irregular heart fluttering you get, I'm sure it's nothing. I'm just here to inform some of the more "Trust the Science!!!" types that these "No Evidence" claims are propagandistic drivel.
No, because if XTwitter is the "everything app", then everybody who doesn't want to be a nobody has to pay for it. And if it's on the order of $10/month, people who don't want to be nobodies will eventually suck it up, stop whining about how it used to be free, and send Elon their $10/month. Which might be enough to make him a trillionaire.
Assuming he can accomplish the "everything app" part, which is extremely dubious.
>This would sound plausible, except that Musk has succeeded by doing the opposite. I think this is why so many people are in love with Musk: he’s proven that valuing good ideas, moving fast, and not having bureaucracy can work, sort of, in a weird little bubble of his own creation.
There's a much more important structural issue to understand when discussing why SpaceX moves so fast while aerospace incumbents move soooooo slow:
Incumbents built the bulk of their engineering processes around the "waterfall" development model, where requirements definition, architecture, design, implementation, testing, deployment, and maintenance are all distinct phases, gated by exhaustive documentation and formal review processes before proceeding on to the next step. They did this because it was state-of-the-art engineering methodology back in the 60's and 70's, and it was how NASA and the DoD wanted it done. The waterfall model affects all of the incumbents' business processes. It's fair to say that Boeing, NorGrumm, and LockMart are essentially big machines designed to execute waterfall engineering, in a symbiotic relationship with the DoD and NASA, which are big machines for accepting waterfall engineering.
SpaceX, on the other hand, is an iterative company. How much of that is simply because it was founded when iterative and agile engineering processes were state of the art is unclear. With an iterative engineering process, you work really hard at pairing down requirements to the absolute minimum needed to make progress, you keep your architecture fairly fluid, and design / implementation / test / deployment are unified, short, continuous cycles.
SpaceX has been a huge success with NASA and the DoD (now mostly, but not completely, Space Force) because both organizations have realized that the waterfall model simply won't scale going forward, and they made reforms to purchasing that kinda looked like a "waterfall lite" model, which allowed companies like SpaceX to implement their own processes, with a smaller number of checkpoints to make it compatible with government acquisition regulations. This is why the commercial services programs that got spun up in the late oh-oh's were so important.
The results have been spectacular, and no better example exists than the comparison between the SpaceX F9/Dragon 2 crew system and the Boeing Starliner. D2's been pumping out 2-3 crewed missions for the last four years, while Starliner is going back into the shop for its third set of major problems and still hasn't launched a crew.
It's tempting to conclude that Boeing is simply incompetent with stuff like this, and that's possibly part of the problem. But the more important issue is that Boeing simply isn't set up to move like SpaceX is, and reforms will cost billions of dollars and take the entire organization offline while they revamp engineering management.
Iterative development is just plain better. SpaceX uses it because it was best practice when they spun up. Boeing can't use it because their engineering organization can't adapt to it.
"Agile" is a common iterative software model. Iterative software is the reason that web software is now constantly updated, and that locally executed apps have largely transitioned to an auto-update model, where new features appear semi-magically.
Iterative hardware is more difficult. That SpaceX has succeeded in implementing an iterative methodology is an achievement not to be discounted.
PS: I realized I didn't really answer your question. Wikipedia has perfectly good descriptions. See "waterfall model", "iterative and incremental development", and "agile software development". Agile's a bit of a cult, albeit an extremely functional cult. I'm not sure you can apply the word "agile" to a hardware project, but you can certainly design hardware iteratively.
The original values from the 'agile manifesto' are (copied from Wikipedia):
Individuals and interactions over processes and tools
Working software over comprehensive documentation
Customer collaboration over contract negotiation
Responding to change over following a plan
Basically, work in small loops, often ~2 weeks, to get the smallest possible unit of work you can done and delivered rather than making small progress on a big plan and not delivering anything until the end. Delivery might mean delivered to the customer or just that it is merged into the main build (for software).
It's worth noting that there are a large number of very dogmatic branded implementations of agile such as Kanban, Scrum, Extreme Programming. Personally I don't believe strict adherence to any of them matters. The main value is that product management can find small bits of work, hand them off to engineers, and then not need to interfere for a couple weeks. At the end they can see how things look and easily adapt for the next cycle (or 'sprint').
I've worked for government contractors that use waterfall, startups that use agile, and large companies that say they use agile but are closer to waterfall. From a developer standpoint, agile means I spend more time accomplishing and less time making plans that will fall apart anyways.
Agile/iterative doesn't completely relieve you of the need to provide a reasonable architecture, but it does force you to architect stuff that's easy to rip out and modify when it (inevitably) turns out that what you thought were going to be the big issues aren't, and things you didn't think of at all turn out to be serious problems.
But there's no substitute for having a good guess as to what your product will grow up to be, and how your customers will use it. Products based on a bad guess fail. Products based on a good guess still morph beyond all recognition, but the kernel of the good guess can guide development through a lot of stuff that would otherwise kill the product.
I still haven't decided whether Musk's guess for Starship was good or bad. The architecture has been consistently driven to support high-scale missions to Mars and back, which I still find wildly improbable. But the things that requirement has forced on the project--ultra-heavy lift, on-orbit refueling, full reusability, extremely high launch cadence, and lots of earth-orbit-rendezvous--seem to coincidentally have built a system that's pretty good for cislunar and lunar surface applications, excellent for launching Starlinks (which hadn't even been imagined when Starship's architecture was first promulgated), and likely to enable a lot of other earth-orbit applications that nobody had imagined.
Manifestos aside, I'd say that the difference between Agile and Waterfall is organizational philosophy.
Under Waterfall, before performing any nitty-gritty work, your entire organization must produce a detailed designed document outlining every single component of your final product, and every single step on the way to getting there. The customer must sign off on this document before a single line of code is written (or a single bolt is tightened). Amending this document is possible, but very difficult (by design), requiring sign-offs by everyone involved. The advantage of the Waterfall model is predictability and accountability: the client knows exactly what he's getting, when he's getting it, and how it's going to perform.
Under Agile, teams iterate on short-term deliverables every couple weeks. The end product is understood in terms of vague generalities (i.e. "we want a word processor" or "we want to launch a satellite into orbit"), but these are attractors rather than firm commitments. The short-term deliverables *are* firm commitments; even though they could be changed, doing so without a good reason is discouraged. The advantage of Agile is rapid iteration. If you are developing some truly innovative product, then discovering mid-stream that your client asked for something physically impossible, nonsensical, or inconvenient for the users is not a problem: you can pivot toward something more reasonable over the next two weeks.
The disadvantage of Waterfall is lack of innovation. Innovation requires exploration, and exploration cannot be baked into the design document and chiseled into stone tablets once and for all.
The disadvantage of Agile is lack of accountability and lack of focus. Once you start exploring, it's possible to explore your way right off a cliff, and deliver some kind of a self-replicating food processor instead of a rocket.
There are ways to circumvent the disadvantages of both approaches, but ultimately you'd want to use Agile for innovation, and Waterfall for reliability.
Iterative development greatly increases the chances that you will fail catastrophically because half your people didn't fully understand how you changed the design midstream. It has its advantages, and is better in some contexts, but it is not "just plain better".
Software and hardware have different answers to that issue. In software, the development and configuration control tools are so good that "refactoring" isn't nearly as big a deal as it once was. Pervasive use of object-based design has helped a lot as well.
Hardware is a different story. I'd say that substantial re-architecting of hardware systems is well-nigh to impossible, but fiddling with the pieces-parts inside a good architecture is pretty easy.
For all their internal complexity, the high-level stuff in a rocket architecture isn't that complex. You have to make very, very careful decisions up front, and then you get to live with those almost forever. But there's a lot of room for incremental improvements in subsystems.
There's also lots of room to wreck your rocket by making a small incremental "improvement" that isn't, because the person making the change doesn't understand all the interactions.
If it absolutely, positively has to work, and if you're flying people or billion-dollar payloads it kind of really has to work, then once you start actually *building* the rocket you go into a solid "any proposed improvements will wait on the Mark II rocket" mode. And you don't start building the Mark II until you get everybody in a room to agree on exactly what you're going to build. SpaceX learned to do this on the Falcon 9, about the time they started getting contracts to launch national security payloads on Falcon 9, and it's not a coincidence that's the one actually reliable thing they've built.
Not as reliable as ULA's rockets, but close enough for government work.
I think NASA required a stable design 5 flights before Demo-2, the first crewed test. That would have been mission #79.
The last loss-of-payload accident, the Amos-6 pad explosion, would have been mission #29. So SpaceX ran 51 missions with an unfrozen design, during which time they handled 13 NASA missions and 3 DoD missions. So at least the billion-dollar payload issue wasn't really a problem.
My understanding is that SpaceX is still iterating on Block 5, albeit much more carefully and incrementally, and with NASA and DoD sign-offs on changes.
As for reliability, that's complicated. If you count just Falcon vs. Atlas V, SpaceX is more reliable--even with the 2 failures and 1 partial failure. That's a result of the sample size being so much larger, with the correspondingly narrower confidence interval. If you count all Delta and Atlas launches under ULA management vs. all SpaceX operational launches, then ULA is slightly more reliable.
The Air Force was requiring design stability and Aerospace-grade mission assurance for GPS launches, which came several years before commercial crew. And ULA's 142 operational launch attempts isn't *that* much smaller than the Falcon 9's 258. Certainly not enough to compensate for two failures vs. zero.
I don't know what to tell you. NROL-76 launched on an F9 FT (NRO birds are expensive, but this one only went to VLEO, which may mean that it was a cheaper asset), and USAF launched DISCOVR on a v1.1 and an X-37B (not a cheap asset) on a Block 4. You may be misconstruing the meaning of "design stability," or maybe GPS had funny requirements. There wasn't an F9-based GPS launch until Block 5 was deployed. And I have no idea what "aerospace-grade mission assurance" is.
Where'd you get 142 ULA launches from? I counted 111 Deltas and Atlases from December 1, 2006, which is when ULA was incorporated. Before that, it's Boeing for Delta and LockMart for Atlas.
If you go much further back in the Delta II/III/IV history, there are some failures. There's also a partial Atlas V failure of an NRO mission in 2007 (payloads went into low but viable orbits), which I missed the first time through.
Huh? OOP is no longer the hot new thing, and functional programming is expanding out of academia, but I still use object-oriented languages and APIs every day. Which part of the industry do you work in?
No one really does "functional" either. You just... figure out what you want to achieve, and then write the code that makes the computer achieve that thing. That's it. No contemplation of the true Platonic essence of an object and how it should relate to other objects in the hierarchy, no quibbling over a function that has a side effect or could be refactored as a quadruply-curried higher-order function - you just write the thing you want to do.
I think you are using a uselessly narrow definition of "practiced in industry". Sure, OOP as practiced in industry usually falls short of the strict forms discussed in academic papers. But OOP in its less strict, git'er'done form is *everywhere* - and, for the purposes of this discussion, that's enough to have had a big effect on the refactorability of code.
Um. What? Java (released mid 90s) is built around OO (admittedly a pretty simplistic and limited version of it). So are most mainstream languages today (e.g. Python and C++). C is probably the biggest exception. People have been using OO in industry since at least the late 80s with Smalltalk.
I'm old enough that pervasive use of OOP was a big mid-life paradigm shift, and I spent a fair amount of time trying to wrap my head around it. But today, the reason you don't see a lot of discussion of OOP is because it's pervasive. It's what all the cool kids learned in school, and they don't even think about it.
That's not to say that there aren't plenty of very poorly designed class libraries and interfaces. But that's no different than how coding has always been: 80% of the stuff gets done by 20% of the coders, who have to struggle mightily to unsnarl the messes made by their less talented colleagues.
Say what now ? If you've ever used the "." operator in Python, or iterated over a list, then you've been using OOP. I suppose OOP is not of much use in firmware programming, but there's more to the world than firmware...
Theory aside, I think it's fairly clear that software quality has degraded over the last decade or two; if, as you say, this coincides with a shift to agile development, this is at least some evidence against agile in software (though not overwhelming evidence, there are other potential explanations).
I think it's fairer to say that, for applications where software quality isn't that important, quality has declined. For mission-critical applications, quality--and functionality--have improved considerably.
I've been retired for a while, so my knowledge of the tools isn't super-duper these days. My impression is that integrated development/test environments are so much better than they were, say, 15 years ago, that you can get arbitrarily dense test coverage over your entire code base. So it's really a question of how much you want to invest in building test suites, and how hard you want to ride your coders on unit testing.
But yeah, for stupid websites that nobody cares about, quality has plummeted. It'll be interesting to see what happens as generative AI grows into that space. I have no clue what'll happen.
There is a wasteland of failure for small rocket startups that were high risk iterative models.
It will be much more difficult for SpaceX to stay the way it is once success has been achieved. Making dramatic iterative changes to a rocket with 230 consecutive successful launches and a low relative cost doesn't have much upside. The long story short is there is no guarantee that they just won't become another Boeing because that is the natural trajectory for things like this. Fast scrappy successful startups turn into fat lazy engorged monoliths. Google. Apple.
SpaceX has been quite public about the F9 design being largely frozen. That's a perfectly reasonable thing to do with your cash cow. It's especially reasonable when the cash cow generates a ton of paperwork to NASA and USSF any time you breathe on it.
The same is not the case for Starship, where SpaceX is dreaming up weird stuff all the time and trying it out. A lot of it never makes it off the pad, but the pad is a major--possibly even the dominant--component of the launch system.
I don't disagree that SpaceX will eventually get slow, bloated, and stupid. But it hasn't happened so far. I wish they had a viable competitor to stave off the sclerosis.
It's a lot easier to iterate faster on software, especially when almost perfect abstraction barriers are the norm. But I agree that Boeing seems to be iterating much more slowly than is ideal. One big potential reason is that when you have interactions between different organizations you often have contracts specifying what needs to be delivered which are very hard to revise. There's a barrier like that between SpaceX and NASA, but otherwise SpaceX gets to keep revising within those bounds. Boeing is the prime contractor among a web of outsourcing and so this sort of rigidity reaches much further down with them. You also have politics developing withing organizations, per the Iron Law of Bureaucracy, which can make different departments interface like separate legal organizations.
The difference is primarily that Boeing is owned by stockholders who mostly don’t care about the long run and sure don’t like talk about “bet the company”…hence the 737Max.
I would argue that Boeing doesn't iterate **at all**. Yes, Boeing will produce elaborations of previous designs, but they all go through the full waterfall process before anybody starts bending metal or writing code.
A fair--and important--point, however, that Boeing's outsourcing is a huge encumbrance to moving fast and breaking things. All of Musk's hardware businesses start out with a minimum of outsourcing, then get more and more vertical over time. He wants control over the entire supply chain, so he can fiddle with any portion of it at will.
As a person who has experience in building stuff with multidisciplinary teams (SW, HW, RF and MECH), with hard problems in each discipline, I want to push back hard on Agile having anything to do with SpaceX being better or faster.
When people talk about agile usually mean a mix of two things:
- The rituals: Standups, retrospectives, and so on. Not focusing on this as it's just ways of trying to make people talk with each other and pressuring each other to perform.
- Iterative design: The great discovery of SW is that iterations are for free. You can build something minimal and then continue adding features forever. When a SW company builds an app, it's going to start extracting revenue from the minimal product (or your VC), and is going to continue iterating on it for a long time, instead of having a fixed delivery date with fixed requirements, you just prioritize which feature goes next based on costumer feedback.
The problem of the rest of the disciplines is that iterations are not for free. They consume money and time. Waterfall project management is just the assumption that the money and time are not infinite, and if at some point you want to integrate the stuff that each discipline has made, you need to have your ducks in a row in a certain point in time.
SW people are used to working just with other SW people, and have absolutely no idea how hard is to make a guy with a phD in theoretical magnetometrics communicate with a mechanical guy who is used to hammer away his problems: they could as well be speaking different languages. You just can't assume that if you let those people iterate in their own, the end result will be something manufacturable that does what it's supposed to do, fits where is supposed to be placed, and it's compliant with the relevant standards.
So the solution is to set official reviews in place, plan for a number of iterations, set processes in place that define the interfaces on the disciplines... And you will be looking at something that is quite similar to what Boing and NASA do. The problem is that these dinosaurs set their processes in a time when CAD was in it's infancy, drawings where actual paper drawings, reviews could take months and they have not incentives to change them.
Now, there is quite a lot of low hanging fruit in the CAD and simulation space. I don't know about the field of rocketry, but in my neck of the woods, most of the enterprise tools were built in the 90s, are an absolute pain in the ass to use, integrate with each other horribly and have terrible version control.
I can imagine that a company with a monomaniatical focus on reducing iteration time could make a review that takes months in those dinosaur companies in hours. In the company I work for, the first times we where doing cross RF-HW-MECH simulations it would take us days to make sure that what every discipline had in it's CAD was the same as the rest, now we are down to minutes by automating the integration between the tools, but it was not easy and goes against what a lot of senior people are used to.
Mind you, this has nothing to do with Agile vs Waterfall. I have a gantt chart always open where I can see what me and my team needs to deliver and when, there are several iterations planned, some of them including cross-discipline dependencies, and if the alternative is a bunch of stickers on a wall, you'll pry it from my dead hands.
Personally, I would say that the biggest thing that Elon has done for making his company special is hiring very smart and driven people, not let them be encumbered by legacy processes, and hammer them continuously in a very specific direction. Smart people, if you don't let them diverge, will end up figuring out the best way of doing things.
Dunno... In my mind if you have delivery dates, fixed external requirements that don't move through the project, and a gantt chart, it's just not agile. Of course, Agile proponents have this Not-True-Scotsman thing going on where everything that works is Agile, and if it doesn't you are just not using it correctly...
it IS agile, but only because agile is a big spectrum.
In the context of this discussion: Every thing that is not classic waterfall model has a non-zero amount of agile in it.
You have classic waterfall on one end with fixed deadlines and a single iteration. And you have stuff like extreme programming on the other end. Scrum is somewhere closer to XP (i.e. it does have fixed dates for delivery, but it's open what will be delivered). And spiral model or V modell or "improved waterfall" are closer to the classic waterfall modell. (I am sure you know all this, but I am explaining it to 3rd parties, who might stumble upon this comment chain)
To a first-order approximation, SpaceX **doesn't** have delivery dates. Instead, they have a very smart but fundamentally unreasonable leader whose personal multiplier (the actual-delivery-date-to-initial-delivery-estimate ratio) is north of 2 (not uncommon amongst engineers). If Elon had a manager, they'd be able to trim his estimates back before they hit the SpaceX Twitter account. But he doesn't.
So everybody has to scramble before Elon loses patience and fires them. I'm sure that makes the culture fairly painful, but also rewarding, because stuff gets done that otherwise wouldn't--or couldn't.
Musk joked a while back that the unofficial company motto was, "Making the impossible merely late."
I hope I was clear that "iterative" and "agile" aren't the same thing. All agile is iterative, but not all iterative is agile.
There's no question that the iteration cycle for hardware is longer than that for software. However, in SpaceX's case, there are three features that enable their iteration:
1) The first, somewhat reductively, is that they don't use waterfall very much, if at all. They don't sweat the requirements. Indeed Musk has a standard "all requirements are dumb" mantra. He even goes so far as to say that you should especially mistrust requirements from a smart person that you otherwise trust. They do the absolute minimum to get something off the pad.
2) They're incredibly software-rich. They build their own tools. They even build stuff like CFD suites in-house. I don't know what they use for CAD, but if there are tweaks that improve their process and productivity, they'll invest in them.
3) They run incredibly hardware-rich. This is largely a function of Musk being cash-rich, with lots of like-minded investors, so they're not afraid to blow stuff up. Being private helps as well, since they don't have to write down the stuff they blow up in public accounting.
Hiring well obviously helps. But SpaceX is big, and big companies can easily attenuate the efforts of their smartest employees. Process is always the backbone of every successful project. They have some secret sauce here.
The Isaacson bio makes reference to Elon’s interest in the mid 20th century Southern CA defense industry, and the idea to use a tent to rapidly expand manufacturing space as they did once at Tesla.
That story comes from Lockheed Skunkworks division. What seems to distinguish Skunkworks, when you look closely at their original operating principles, is an early version of iterative vs waterfall manufacturing.
Because of the tent and the Isaacson reference, I strongly suspect this was a primary influence for Elon, along with iterative being the de facto method in software.
For hardware, I also suspect “iterative” design is more about rapid prototyping into a stable production design, versus true continuous delivery. Still a very big advantage over waterfall.
Another thing I think gets overlooked is that SpaceX is at least as much a software company as a hardware one.
Back in the Olden Times, lots of aerodynamic and structural regimes simply weren't available to the engineer, because they weren't dynamically stable. You had to make your hardware dumb, because the compute power and software sophistication to stabilize dynamically unstable systems was impossible. That started to change in the late 70's, but the hardware was still pathetic.
When SpaceX spun up, they could essentially throw arbitrary amounts of compute power at problems for almost no mass penalty. Ultra-fast networks and sub-millisecond minor cycles let you design hardware a lot more flexibly--and cover up a lot of design flaws that would otherwise send you back to the (literal) drawing board--than was previously possible.
Again, it's much easier for a company founded in the oh-oh's to capitalize on this change than it is for a company that stopped doing methodology innovation before most of its current employees were born.
Great points. Also, out my depth here, but I wonder what they have in place to create a CICD-like toolset for their overall design and engineering pipeline.
I'm curious about the state of the art here. Does anyone know what proprietary tools they have in that realm?
I don't know the answer to that. I'd kinda guess that continuous delivery would be a little dicey for flight software, but they also produce a huge number of tools in-house, and I'd bet that they use a CICD model for those. There's also of course a huge amount of launch processing and ground segment software, which probably needs fairly tight control, but not as tight as what flies on the vehicles.
I'm too lazy to dig it up, but there was a really interesting presentation by the guy who was in charge of SpaceX's in-house computational fluid dynamics system. They've done some really innovative things to tailor the size of the CFD grid to what they have available for compute resources. Rather than a fixed grid of points on the vehicle, they have an algorithm that makes the gird sparser or denser on various parts, based on how the CFD solution appears to be shaping up. And it's all completely custom, proprietary code.
Something that hasn't gotten a lot of discussion is his ongoing cervical radiculopathy problems. He had a total disc replacement a few years back, which, by his account, failed and had to be redone, presumably as a more standard fusion. (The Blofeld-like scarves and neckerchiefs he's often seen wearing are, I believe, an attempt to cover up the scars.)
I've had a similar problem, and I can attest that the pain is excruciating and chronic, and the only thing that really helps are opioids. My problem occurred just before cervical TDRs became a widespread technique in the US, and I hung on, using ice packs and Norco, until I was able to get a TDR. My surgery was successful, and I was fortunately able to wean myself off the opioids without too much trouble, but I wouldn't be at all surprised if Musk had to rely on them. Even if you're careful, getting the balance right between enduring the pain and not being gorked out on the meds isn't easy.
Musk has always been a bit erratic, but things got markedly worse about four years ago. That was, I believe, close to the time when his surgeries were occurring. Correlation is obviously not causation, but the timing kinda matches up.
I'd speculate that mental <--> physical issues tend to feed one another.
"It is from this obsessive identification with the body, I believe, that many physical and emotional problems arise. In meditation, as we learn at deeper and deeper levels that the basis of our personality is not physical but spiritual, such problems fall away. Often we do not even have to confront them. We simply go deeper, move away from the tenements of consciousness where they arise.
…
You can leave psychosomatic problems hanging in the closet and find another home in a much safer neighborhood; they will never be the wiser. This approach is very different from the conventional wisdom of the modern world. It cannot work without meditation. For it is much more than a change in lifestyle; it is a transformation of thought-style.”
Elon Musk was noticeably in a grumpier than usual mood during the 3 part Starbase tour he did with Tim Dodd (The Everyday Astronaut). I'm having trouble finding it now, but somewhere Tim flashed up a tweet/message from Elon where he apologized and said it was due to back pain.
He also tweeted about it how he smashed his "c5-c6 disk" during Sumo wrestling
Yes, and that's why he was putting off the Zuckerberg MMA match (and probably would have lost). It's an interesting anecdote because you have to wonder if it's related and thus an example of the sort of behavior that contributes to all the mood-disorder comorbidities: you don't see other comparable men like Richard Branson or Larry Ellison or Jeff Bezos getting injured doing things like... [checks notes]... "trying to throw an actual sumo wrestler who weighs twice or thrice what they do by sumo wrestling". (Men their age are usually a *little* more aware of the fact that at that age, injuries, especially joint injuries, no longer go away.) But it is just one of countless anecdotes about Musk's risktaking.
Yeah, agreed! I actually was not aware of the sumo wrestling thing, just came across it while searching Twitter this morning.
Would be interesting to try to track down when he started taking sleeping pills also. Back pain and mental health issues both make it harder to sleep. He's tweeted a couple times how he takes sleeping pills occasionally, mentioning Ambien on at least one occasion. Prescription sleep medications can trigger depression in some people as well as memory and cognitive issues.
rebuttal to tg56's comment mistakenly comparing Tesla's balance sheet vs Ford's by citing long-term debt. tldr: the long term debt is a function of Ford having a financing arm and is more than offset by financial assets and equity cushion. Details in link below
I do not think that the Boring Company is going to build effective public transit.
Musk does not seem to want to build transit that involves sharing a vehicle with strangers. He's said this explicitly: "there's like a bunch of random strangers, one of who might be a serial killer."[1] It's also what he has designed. The Vegas Loop involves individual cars. Hyperloop was supposed to be composed of 'pods'.
This is a major problem. Putting multiple people in a vehicle allows you to get much higher densities.[2] By insisting on individual cars, you're accepting a 5x (relative to bus) - 50x (relative to large subways and commuter rail) reduction in the capacity of the transit line.
The choice to use cars is not because it's a small diameter tunnel. The London Tube has 12 foot tunnels. They use unusually small trains for a subway, but are still able to move a lot more people than cars could.
The Boring Company can probably reduce the cost of tunneling in the United States. The US builds transit at 10x the cost of Italy or Spain.[3] Plausibly Musk can even improve on global best practices. But because of his choice of what to put in the tunnels, he's starting with an order-of-magnitude handicap.
> By insisting on individual cars, you're accepting a 5x (relative to bus) - 50x (relative to large subways and commuter rail) reduction in the capacity of the transit line.
Only compared to a transit line that consists of a constant end-to-end stream of buses/trains though. This almost never happens; a typical transit line has a train and then a gap of several kilometres and then another train behind it. A constant stream of private cars has a much higher passenger density than a train every N minutes, especially if it doesn't need to stop at every station.
It's also a helluva lot more pleasant, and I'm not a believer in making people suffer for the sake of efficiency anyway.
A constant stream of passenger cars along one lane of a highway has similar passenger density as a train every hour. At this frequency, it's not worth building a tunnel for a train either.
As an example, there are two ways to travel from Oakland to San Francisco: the Bay Bridge and the Transbay Tube. The bridge has 10 lanes for cars going towards San Francisco, while the tube has one "lane" for trains going towards San Francisco, which run every 2.5 minutes (maybe 4 km apart). Peak traffic occurs during morning rush hour, when the bridge carries 14,000 people per hour and the tube carries 27,000 people per hour. [1]
I like pleasant private transportation too. I just think that it's worth recognizing that this is what the Boring Company is working on. They're not going to revolutionize mass public transit by putting electric cars in a tunnel.
Fine, now I've sat down and done the maths. A jam-packed subway train (1500 people) every three minutes has a capacity of 30,000 people in an hour. You'd need 8.3 people going past every second to manage that with a single lane of traffic, which is implausible even under generous assumptions. So you're right on that one.
But of course if tunnels were cheap, you wouldn't have a single chokepoint like that. People don't want to go from West Oakland to Embarcadero, they want to go from Alameda to the Marina, or Berkeley to SFO, or Richmond to Richmond. You'd have a whole bunch of tunnels crossing the Bay in a whole bunch of different places so that people can get from where they are to where they want to be, efficiently.
That would only make sense if the cars themselves took up no space and had no need to keep any distance from each other in case one of them suddenly stops; otherwise, you'd have every single one of them crashing into the car in front of them in a massive chain reaction.
All the same limitations that apply to trains also apply to cars (it seems strange to me that anyone would not see this), except that people can sit or stand directly next to each other in far higher density in trains or buses than they can while sitting and driving in cars.
Public transportation is also much safer than cars. Musk being afraid of the potential serial killer next to him seems comical, considering the prevalence of serial killers. He should be much more afraid of his own capabilities to consistently drive safely and the capabilities of every driver on the road. You don't need a crazed murderer to die on the road; you only need a moment of inattention or a lapse in judgment.
Driving in a car is also at most moderately more comfortable than taking a train (or not at all if you actually have to drive yourself), unless you have a deep disdain or disgust for other people, which might be the case for a member of the ivory tower bourgeoisie. However, that's not something to be proud of; on the contrary.
I also feel that much of this perspective comes from an American standpoint. Maybe you (and Musk) would not dislike public transportation as much if you had actually experienced good public transportation. Where I live, nearly no one I know owns a car; there's no point. Public transportation is most of the time more comfortable, vastly cheaper, and safer. And often, it's even faster because finding a parking spot in the city sucks.
Re: going to Mars, since Elon has said the most entertaining outcome is the most likely, then I am putting my money on a massive asteroid wiping out the first SpaceX colony there about 15 years in.
The only one of those kids' names that would be normal to anyone with an understanding of American pop culture would be the one that was changed (to another normal name). 9% is, uh, abysmal as a batting average.
Also, that painting is something else. I hope that comment regarding calling Elon Musk "daddy" is some effort at trolling.
Should have written that a ton of furries have migrated from twitter to bluesky: i thibk about 10% of blueskye is furries, and one of the biggest accounts on blueskye, braeburned, is a furry porn artist that has 10% of the site following him
Many left leaning or progressive minded people have made alternate accounts on other websites but havnt left twitter yet, because they are uncertain if its actually gonna collapse
RE: sclmlw's comment on the Boring Company and other comments about size/cost
The Boring Company's car tunnels are about 14ft outer diameter (it says 12ft inner diameter on their website, assuming 1ft liner thickness, I think 14ft is fair). A typical subway tunnel might be 15ft-21ft outer diameter. Not much difference in diameter, but the cost approximately goes up with the cross sectional area, because of the volume excavated. So comparing 14ft to 20ft that's a 2x difference in volume of excavation. Also a larger diameter tunnel will require a thicker tunnel liner. It's reasonable to claim that the Boring Company could build tunnels maybe 1.5x to 2.5x cheaper than a typical subway tunnel.*
Score some points for the Boring Company - it is significantly cheaper. However it's not the 10x cost savings they're aiming for according to their website, that will need genuine technological innovation. Also I would counter that they can build the tunnel for 50% of the cost, but using cars, you're not even close to 50% of the passengers per hour (The Chaostician did a great example in another comment).
*(For an apples to apples comparison, this comparing 2 14ft tunnels to the more traditional "twin bore" subway system with 2 15-20ft tunnels, one in each direction of travel. There's also a modern "single bore" approach with a single 35ft-50ft diameter tunnel. This is more of an apples to oranges comparison because the tunnelling costs significantly more. But you save money by putting station platforms and misc parts of the station inside the tunnel itself. Plausibly this is the method west coast US people are hearing about in the news, because the Seattle Alaskan Way Viaduct and the BART to San Jose projects use single bore?)
RE: Safety
Safety wasn't a big part of my original comment because I don't have much experience there. I said they're going light on ventilation, exit walkways, and fire suppression systems (sprinklers). I was wrong about fire suppression. Looking at Youtube videos, they have a sprinkler system visible at the top of the existing Vegas Loop tunnel. Cutting out the exit walkways makes sense for a car tunnel - the bottom of the tunnel is already flat, instead of a subway tunnel with rails and other random crap like sump pumps and cables to trip over. My biggest gripe is the ventilation.
The ventilation of a tunnel has 3 purposes:
- Ventilate exhaust fumes (scratch this one for electric car tunnels)
- In case of fire, blow the fire/smoke in 1 direction, instead of letting it spread in 2 directions. Preferably use reversible fans to blow the fire/smoke to the direction with fewer passengers, or whichever direction the fire department desires.
- In case of fire, cycle fresh oxygen in the tunnel so passengers and firefighters don't suffocate (Fire consumes the oxygen. In principle you could NOT ventilate and try to smother the fire with lack of oxygen, but in practice I believe it's better to circulate air and send firefighters in there)
This is why in the bay area there's very large fans visible at the entrances to the Caldecott tunnel. IIRC they were retrofitted after a gasoline tanker caught fire in the tunnel.
Electric car tunnels still need a plan to deal with fire. (Google "tesla catches fire"). It says they have a fire safety system and a ventilation system on their website, but I didn't see any ventilation on a quick scan of Vegas Loop Youtube videos. I don't want to badmouth safety on a project I'm not familiar with, maybe they have hidden ventilation, or presumably if they're skipping ventilation they have a detailed fire safety plan ready to go with the Vegas fire department. Maybe electric car fires are a slow burn compared to gasoline car fires, and there's more time to send firefighters down there?
The Seattle Alaskan Way Viaduct replacement tunnel is a double-decker car tunnel, with no stops. You can see a cross-section on Wikipedia, and it seems like a decent way to handle 4 lanes of traffic, including large trucks? There were some major delays, including hitting steel that the tunneling company knew was there, and this might have made the news.
The Seattle light rail tunnels are a twin bore pair of 18-21 foot tunnels that handle 1 light rail track each (although the 18 foot part can also handle city buses). The stations are dug out separately.
I'm a small hedge fund manager, race car driver (I own a Model S Plaid as a daily driver), and ex-software engineer. I've been following Musk for professional reasons for about 5 years now. While he demand tremendous respect in public markets, I've never got the impression he's super smart. Some anecdotes that come to mind:
His tweet on Tesla polar moments (https://twitter.com/elonmusk/status/1042842449008197632) is nonsensical. Outside of racing, no polar moment is not extremely important, and there's no way a Model 3's is better than a lightweight, mid-engined car.
When he announced the Tesla Semi, I analyzed his claims using data from the DOE SuperTruck competition (analyzing range is fairly easy if you have Crr and CdA figures). I found them to be dependent on rapid drops in battery prices, which I think has been shown to be accurate.
The Hyperloop whitepaper. It's all hand-waving nonsense, glossing over the important questions that would determine if the idea is viable. Can the air bearings support its weight? What does it even weigh? What happens if the tube leaks, or someone shoots a hole in it?
FSD. I have a subscription to it for market research purposes. I don't think this is Musk being dumb as much as it is fraud. Tesla really needed that FSD cash flow early in its life, so he said whatever he needed to. Now the company is walking back some of the earlier claims.
I followed his Twitter acquisition from the moment he suggested it on Twitter (thanks for that Elon). He bought at a really dumb time, when tech companies were valued absurdly and interest rates were rising. I can't completely fault him for this, as it's not his job (as it is mine) to track asset bubbles. However he did not seem to understand the reason social media companies cater to the left is because of advertising revenue? How could anyone causally following the business not know this?
Making Mars a backup Earth? You don't just need a colony, you need millions, probably billions of people to produce the technology needed to survive on Mars. This is basic economics. Ergo Mars is much more of a biological and terraforming problem than a rocketry problem.
He wanted x.com (or PayPal; I can’t recall the timeline) to switch from a LAMP stack to Windows in the 90s? My eyes can only roll so much. Good thing he was fired.
I've just never seen him say anything revelatory, and I learn things from objectively dumb people all the time. I pay close attention to Elon because he moves markets, not because he teaches me anything.
Going over his big successes:
Landing rockets: Well I know nothing about rocketry, but this seems really awesome. I don't know how much of this was Elon, and how much Gwynne or other engineers. Anyone know who originally had the idea?
Electric cars: Anyone who knew anything about cars knew BEVs were better if the batteries got good enough. He should be given some points here.
Starlink: This is more of a derivative of cheap launch costs. Good on him for realizing it.
Twitter: Assuming it succeeds (my guess is it will), that is some points in his favor. However Twitter was the worst-run big tech company on the planet. This would be a business success, but not evidence of super-genius.
I see his successes more as evidence of how terrible our society is at utilizing intelligence. There are lots of people out there smarter than Musk, but they lack the motivation, charisma, cult following, and narcissism to command the sort of capital he does. Musk can try all sorts of dumb shit, fail a dozen ways, and somehow still command respect and status. Most people are too excessively afraid of failure to do the things he does.
I think he’s smart enough to realize when the mainstream heard has it wrong, narcissistic enough to believe only he can fix things, and charismatic enough to build a cult of personality around this.
I wonder if his success is due to the existence of so many rich, low-neuroticism tech bros? He can do and say all sorts of stupid shit that most people would (as the media does) focus on, yet his fans merely shrug off. They know his successes are far more important, while neurotic wordcels focus on his many flaws. If this is the case, he might not have succeeded at any other point in history.
If years trading public markets have taught me anything, it’s that people are quick to form cults around their favorite stocks and CEOs. A ticker (more a person in this case) is simultaneously a flag to rally around and a bet that might make them fabulously wealthy. Whatever else he is, Elon is the greatest stock pumper of all time.
I agree with basically everything you've said. Just a small remark:
> Ergo Mars is much more of a biological and terraforming problem than a rocketry problem.
The biological and terraforming problems can only be approached once the rocketry problem has been solved. Getting to Mars safely and efficiently is a necessary precondition for making it inhabitable.
Of course, a much smarter path would be to colonize Earth's orbit first. Unlike Mars, orbital habitats actually provide an economic benefit in the form of a location advantage not found anywhere on Earth: microgravity.
Eventually yes, I agree. However surely there's a lot of theoretical work to do before we start sending rockets over? Maybe we're at the point where we need run tests on the red planet itself?
e.g. if nuking the poles is the answer, we ought to do that when there aren't any humans on the planet. If there are people there, they might actually slow progress down.
I tend to think the sending rockets to mars obsession has more to do with his narcissism than anything else.
But for now, those tests are pitifully inadequate because of the expense of getting anything at all to Mars; MOXIE had to be substantially descoped to fit the mission.
>But out of all of that, my biggest take-away was that Teslas..... just aren't very good? Their structures up to the Model 3 are quite inefficient and don't have great rigidity. The dimensional variation is shocking (far beyond even SBU, IYKYK). The hang-on parts are generally relatively poorly performing on their own. They can't touch our structural or powertrain durability tests. Rate and handling is bad, ergonomics fails to meets package targets, NVH and sound quality are poor, and we pay JD Power far too much to find out just how bad the quality numbers are (hilariously bad). I don't think it's an exaggeration to say that most other OEMs can't make a Tesla, because our systems and processes prevent us from releasing something that half-baked.
This is basically what I'd expect from Tesla trying to build a whole auto manufacturing ecosystem from the ground up - obviously it's not going to be as good as the incumbent one. That's presumably why you almost never see a brand new car company launch and successfully compete. Tesla managed it because they had a killer feature that no one else had - a functioning electric car!
>So where did everyone go? Was it Mastodon after all? Facebook Threads? Blue Sky? Or did they all start learning to paint and spending time with their friends and families?
I, for one, started reading actual published novels again, which I've barely done for the last 5ish years.
Yes, you’re right. The bio repeatedly describes Musk as completely insensitive to emotions — quite the opposite of the borderline type.
There does seem to be something fundamentally disorganized about his personality, though. Particularly the highly unstable relationships (divorcing Riley twice etc). I just don’t buy the bipolar story because of the rapidity of the shifts.
I’d be very curious to hear a real psychiatrist weigh in.
Physician here. Here is the grand hypothesis coming up in the medical community at the moment: chronic pain is best understood as the endpoint of an evolution within the nervous system that is increasingly called "central sensitization syndrome." (CSS) This syndrome is caused by a chronic dissociative or avoidant stress response, usually having it's origin in early childhood adversity. It's common in people who have a "push through" or type A approach to life. And you run into that type of personality in day laborers as well as fortune 500 ceos. Frequently in people with chronic pain there are other manifestations of CSS, such as IBS, chronic pain, TMD, etc. Every new stress, be it a viral syndrome, a grief event, a difficult move of house, what-have-you, can cause the central sensitivity to further evolve. Elon is an absolutely classic chronic pain patient. The fact he had a laminectomy a few years ago with ongoing chronic pain is evidence that maybe the laminectomy was not the right treatment. In general, the right treatment for chronic pain is addressing the "push through" and avoidant stress response originating in early trauma.
" A complex system that works is invariably found to have evolved from a simple system that worked. A complex system designed from scratch never works and cannot be patched up to make it work." -- John Gall (1975)
> Do we have evidence that Musk has been fatigued and felt worthless and just wanted to lie around in bed and not cared about Mars or anything for two straight weeks?
As a matter of fact, Scott, we *do* have accounts of *exactly that*, Musk physically lying around catatonic (unable even to make it to a bed), which I highlighted yesterday: https://www.reddit.com/r/slatestarcodex/comments/16heyx9/book_review_elon_musk/k11o283/
Isaacson 2023, Elon Musk § "Are you bipolar?"
"Devastated by the breakup with Amber Heard and the news that his father had a child with the woman he had raised as his stepdaughter, Musk went through periods when he oscillated between depression, stupor, giddiness, and manic energy. He would fall into foul moods that led to almost catatonic trances and depressive paralysis. Then, as if a switch flipped, he would become giddy and replay old Monty Python skits of silly walks and wacky debates, breaking into his stuttering laugh. Professionally and emotionally, the summer of 2017 through the fall of 2018 would be the most hellacious period of his life, even worse than the crises of 2008. “That was the time of most concentrated pain I’ve ever had,” he says. “Eighteen months of unrelenting insanity. It was mind-bogglingly painful.”
At one point in late 2017, he was scheduled to be on a Tesla earnings call with Wall Street analysts. Jon McNeill, who was then Tesla’s president, found him lying on the floor of the conference room with the lights off. McNeill went over and lay down next to him in the corner. “Hey, pal,” McNeill said. “We’ve got an earnings call to do.”
“I can’t do it,” Musk said.
“You have to,” McNeill replied.
It took McNeill a half-hour to get him moving. “He came from a comatose state to a place where we could actually get him in the chair, get other people in the room, get him through his opening statement, and then cover for him,” McNeill recalls. Once it was over, Musk said, “I’ve got to lay down, I’ve got to shut off the lights. I just need some time alone.” McNeill said the same scene played out five or six times, including once when he had to lie on the conference room floor next to Musk to get his approval for a new website design.
Around that time, Musk was asked by a user on Twitter if he was bipolar. “Yeah,” he answered. But he added that he had not been medically diagnosed. “Bad feelings correlate to bad events, so maybe the real problem is getting carried away for what I sign up for.” One day, when they were sitting in the Tesla conference room after one of Musk’s spells, McNeill asked him directly whether he was bipolar. When Musk said probably yes, McNeill pushed his chair back from the table and turned to talk to Musk eye to eye. “Look, I have a relative who is bipolar,” McNeill said. “I’ve had close experience with this. If you get good treatment and your meds dialed right, you can get back to who you are. The world needs you.” It was a healthy conversation, McNeill says, and Musk seemed to have a clear desire to get out of his messed-up headspace.
But it didn’t happen. His way of dealing with his mental problems, he says when I ask, “is just take the pain and make sure you really care about what you’re doing.”"
Read the new biography on the plane yesterday.
I'm surprised that nobody has raised the possibility of a personality disorder, perhaps some version of borderline personality.
Some reasons to think in this direction:
1. Musk may have cycles, but they are very rapid. He can flip on a switch. Borderline is/was often misdiagnosed as Bipolar, but these are weeks+ cycles. I've not seen evidence, for example, that Musk has ever had a six month+ major depressive episode. At the same time, diagnosing Borderlines as Bipolar is common, in part because it's a more glamorous disorder.
2. The list of symptoms of borderline matches pretty well, including "Unstable and chaotic interpersonal relationships"; "Impulsive or reckless behaviors"; "Chronic feelings of emptiness"; "Inappropriate, intense anger that can be difficult to control"; "Transient, stress-related paranoid or severe dissociative symptoms" (see, e.g., Twitter behavior) https://en.wikipedia.org/wiki/Borderline_personality_disorder#Signs_and_symptoms
3. Pretty extreme forms of child abuse and abandonment — both high risk factors for borderline. Although the most significant risk factor, child sexual abuse, is absent, it is notable that his truly awful father had leanings in that direction.
The main difference, of course, is that Musk is extremely high-functioning. Another difference is that Borderline symptoms tend to subside in later life, but Musk shows no sign of regularizing his behavior.
My impression is that people with BPD are very concerned with abandonment and cycle in their relationships between trying to be close and provoking conflicts due to hypersensitive perception that the other person is critical or unloving. They are manipulative in a desperate way (not devious like a psychopath) for example threatening to commit suicide to get other people to take care of them. They often feel terrible and have difficulty regulating emotions so they are prone to take drugs or engage in activities like cutting. This doesn’t match Elon Musk very well.
1. @gwern: Appeciate your comment but this is completely the opposite of how he behaves in public and has run Twitter. How is constantly tweeting all day instead of focusing on his long term vision a evidence of a great undisturbed undivided focus and vision? He keeps putting out these I work 18-20 hours a day, sleep on the Twitter floors, sleep on the factor floors PR stuff and people keep lapping it up (I am sure he has done that occassionally but not nearly to the level of his PR would suggest.)
Why is the person who is visibly constantly running into expensive rakes is the "most focused person in the world"?. He foolishingly bid for Twitter at $44B, then tried to get out the deal repeatedly, tried to lie to the Delaware chancellery courts which would have been a losing proposition, was then put into a position to buy Twitter at a wildly overvalued $44B price and after having bought Twitter, then lied to the world that he did it for noble free speech and public square platform reasons.
This is how he normally operates. Him stepping on rakes and him tagging along without fatal blows is considered evidence of his unrelenting focus rather than the obvious occam's razor observation that Most Business Mistakes are not Fatal as Arnold Kling put it (https://arnoldkling.substack.com/p/most-business-mistakes-are-not-fatal)
2. @SSC original post:
No, the user experience is not the same on X after firing 80-90% of the staff. It is noticably worse.
What is the evidence that he is running Twitter well? Also a social media company is not measured purely on "user experience". There is no moat for social media companies other than network effects. When you bring your company into serious disrupte you risk your company's moat so company is not on firm reputational and network effects ground as it was pre-Elon. He has made Twitter not "cool" anymore.
Twitter is by Elon own admission down by 90% in value, so how is it "doing well" and how could SSC possibly not mention this or even factor it into SSC's prediction that X will be a "success"? Also what is success? Is it just not folding or becoming the $3T company his cult has been saying it has the potential to be which would be a richer company than Apple (and yes, they bandy about that $3T figure. Just search on YouTube and you'll find a lot of videos with that exact figure).
Also there is nothing in the post about the dozens of lawsuits and the downright shameful non-payment of bills to vendors, non-payment of rents at various of its locations. The richest man in the world is stiffing vendors and landlords and this doesn't even get a mention in the post about how Elon is running Twitter? He is on track to pay a lot of fines for stiffing vendors and ex-X employees. That is not is X "doing well".
Would have liked more discussion on if and how $TSLA is overvalued which is his primary claim to the title of the richest man on Earth
See the section marked "Updates" at the end.
Thanks. I made a mistake. You do not in fact find a lot of claims with that $3T figure. I confabulated that figure from the news that Apple crossed the $3T market cap news I had read. The figure the Musk cults/fans bandy about is a Trillion dollars. ("Twitter will become a trillion dollar company" is the popular claim).
> Appeciate your comment but this is completely the opposite of how he behaves in public and has run Twitter. How is constantly tweeting an evidence of a great undisturbed undivided focus and vision?
I don't understand your point. Yes, of course physical catatonia is the opposite of 'how he behaves in public'; when you have catatonia, how are you going out in public? (The Tesla CEO is practically physically dragging him into the earnings call in that anecdote.) That's my point. Objections based on 'well, I don't see him during his public performances being depressed!' are self-refuting.
As for 'constantly tweeting', how do you know? Tweets are extremely easy, can be handled by assistants, and he periodically announces he's going to stop tweeting for a while - not that you're exactly firing up data analytics software to try to infer his latent cycles from timing during day and overall amount such that you would observe any absences... (Probably a good project for someone, let me know if anyone does it.)
> He keeps putting out these I work 18-20 hours a day, sleep on the Twitter floors, sleep on the factor floors PR stuff and people keep lapping it up (I am sure he has done that occassionally but not nearly to the level of his PR would suggest.)
I'm sure he does do all that quite a lot, although I don't know what level his PR suggests in any falsifiable quantitative sense... Those are easy when you're in a hypomanic phase, and so much less impressive than they look. You only slept 4 hours last night, Musk? Yeah, you and every other bipolar entrepreneur, big deal. (I spent an hour or two today reading through accounts on Hacker News today of bipolar techie types, usually with formal diagnoses, to check my impression, and yeah, Musk looks even more typical in like of them - remarkable only in quantity, not quality.)
"He would fall into foul moods that led to almost catatonic trances and depressive paralysis. Then, as if a switch flipped, he would become giddy and replay old Monty Python skits of silly walks and wacky debates, breaking into his stuttering laugh."
This isn't how I expect bipolar to work. There is no "switch flipping" (except very occasionally when a manic episode follows directly after a depressive one). A patient will be depressed for weeks or months, then gradually come out of it, and after weeks or months of coming out of it, get back to normal. Being "moody" in the sense of having mood swings is kind of the opposite of bipolar; I would associate it more with borderline or PTSD.
Throughout late 2017, Musk was continuing to sound as "manic" as ever - for example, committing to build hyperloops from NYC to DC and saying the government had given approval (it hadn't).
I am absolutely willing to believe Musk has extremely bad times, mood swings, and horrible lows, but you are going to have to do something other than keep repeating that if you want me to believe it's bipolar.
But isn't it starting to sound like a mood-based pathology? At the point that someone can be fairly described as having "hyperthymic temperament, hypomanic signs, and moderate to severe MDD" and where they present with pretty severe depressive symptoms (based on the excerpt gwern provided, although I want to note that if you're rousable with a little encouragement you might be in the vicinity of catatonia but you aren't there yet) you (well, I) tend to think the same sort of thing has gone wrong for them that goes wrong in bipolar. Akin to the idea that some instances of MDD have more in common with BD than other forms of depression, etiologically speaking--are "on the bipolar spectrum," but the pt doesn't manifest enough manic symptoms to meet the dx threshold of hypomania (I've seen this--their depressive symptoms are more severe and melancholic, they have bipolar-like temporal patterns of wellness and unwellness, they have assorted soft signs, they don't respond well to antidepressent monotherapy, but they don't have proper hypomanic episodes, or they are too short and infrequent to catch). Musk could have such a high mood baseline (which would be uncommon, as dysthymic baselines predominate) that his discrete up-episodes are harder to detect, and that his "moments of normality" within the depression look like hypomania because that's his normal. Hypomania is easy enough to miss even when it's a big deviation from the norm. (I also think it's plausible that the rapid switching was brought on by drug use, legal or illegal.) I would also like to see something more pathognomonic of a classic manic episode to (mentally) slap a dx on him, but my suspicions are definitely raised.
Of course the point of the dx is prognostication and treatment, so the real question is--what kinds of pathological behavior would you expect from Musk in the future, and what kinds of treatments might he respond to?
I was actually going to comment in the other direction, on the indefensible grounds that I'm a fairly good...mental health phenotype detector/pattern matcher? And that Musk just doesn't strike me as bipolar, the way that, say, RDJ, Eminem, and Kanye do (all of whom I clocked before their diagnoses were public). But my phenotype-detector is obviously more prone to Type II errors.
ETA like a month later: I realized shortly after I wrote this wall of text that every reported symptom that led to bipolar suspicion can also be explained by stimulant abuse. It's impossible to get a good diagnostic picture while abuse is ongoing, and stimulant abuse makes easy hash of phenomena like quasi-catatonia, wild mood swings, and risk-indifference.
Is there any reason, apart from the classification schemes of 19th century psychiatrists, why mania needs to be accompanied by depression? Can't someone just have the mania without the depression? You can certainly have depression without mania.
What if Musk just has the light side of the force without the dark side?
My understanding is that bipolar can be diagnosed in the absence of depressive episodes; it's called "unipolar mania" and isn't very common.
But Isaacson's account certainly makes it sound like Musk experiences depression.
Yes it does happen, but supposedly it's a lot less common.
To someone like me, i.e., not a professional, Musk's behavior seems more like a positive trait (capability to work very hard for extended periods of time) that in some people perhaps is strong and unmanageable enough to turn into bipolar. He's hardly unique in this!
The fact that he was very sad about Amber Heard leaving him (apparently without even a goodbye gift in bed) doesn't really strike me as pathological either.
What's with the obsession of pathologizing everyone and everything? Why can't Elon Musk just be a normal person who gets distraught when his girlfriend abandons him and his father has a semi-incestuous relationship with his stepdaughter?
My pre-1995 SAT scores slightly exceeded Musk’s (Wow! That surprises me), but I don’t think my IQ is 136-140. I was IQ tested as a 9-year-old and had an IQ of 128.
Childhood IQ tests really aren't reliable. They are the ones most likely to be influenced by your environment.
Just for reference, I got a 1480 pre-1995 (IIRC 760 math, 720 verbal) and tested at 136 in high school. At least back then, the verbal part of the SAT was mostly about spending long hours studying words you've never heard before (nice gatekeeping!). The IQ test seemed to be more well-rounded. I nailed some areas like spatial relations (I grew up playing with Tangrams) and repeating a long number backwards, and sucked at some others like story problems. I shudder to think how badly I'd so at some of that stuff now.
"At least back then, the verbal part of the SAT was mostly about spending long hours studying words you've never heard before (nice gatekeeping!)."
It was actually about being a smart person who read a reasonable amount across a variety of fields and learned those words naturally. But I guess some people could or had to game it.
The scores kind of surprised me too: SpaceX engineers were said to be pretty elite, such that I'd expect many of them to score higher (adjusted for the drift in meaning), and we read about Musk impressing them as an engineering polymath. The idea that what's exceptional was the combo of talent and focus/drive, that makes sense, but still I think I'd chalk this up more to the SAT scores being noisy in this range.
Remember that IQ is only moderately good at predicting any particular talent; the example I usually use is that Kasparov has tested IQ 135. He is a "merely" 99th-percentile smart person who happens to be amazing at chess.
Chess success and IQ are only weakly correlated with the strongest correlation being for kids. That's because raw calculations are a small part of a game compared to memorization (especially of opening lines) and pattern recognition. The trend is becoming even stronger with the raise in popularity of faster game formats like rapid (usually 10 minutes/player for an entire game) and bullet (usually 1 minute/player for an entire game) that don't allow a player to calculate for a long time a single move.
One useful management trick is to hire people who are smarter than you.
I was 114 as a child and averaged 130 as an adult. It's not a hard and fast number.
What does it matter? If Musk had an IQ of 50, would his accomplishments be any the less impressive? Surely IQ itself is only a proxy, which we care about only because it lets us make (probabilistic) predictions about someones likelihood of succeeding at [given task]. But in this case we don't need proxies, we have thirty years of actual record to judge on.
Yes. The lower Musk's IQ, the more impressive it becomes that he made a company that's better than THE ENTIRE WORLD at rocketry.
On the Twitter point, speaking anecdotally from personal experience, I’ve seen a significant amount of followers and correspondents drop off Twitter post-Elon, and it has not slowed down. I’m mostly the film Twitter space, and a lot of folks moved to Mastodon, there were a lot of complaints then about Mastodon, and then people started to migrate to Bluesky, which seems to provoke much less technical misery. Twitter Blue remains intensely unpopular.
Bluesky seems like a weird choice, considering how long Jack Dorsey has been at the tops of Twitter, all the while its shittification kept going on (and also the Trup era !).
But I guess these people are more concerned about Twitter being evil because of Elon, rather than because of being a platform ?
(Also, while Bluesky is a protocol, Jack Dorsey still doesn't particularly seem to be willing to make amends, considering how Bluesky *Social* is proprietary.)
I think that while some folks are leaving out of a personal distaste for Musk, a lot of it is because of the increased chaos and decreased functionality since he took over. Personally, I didn’t bother to make a Bluesky account until this week, when Musk started threatening to charge for Twitter access. The constant arbitrary rule changes, interesting people and entities leaving, and increased bugginess have made Twitter a genuinely worse experience than it was before (which is saying something).
As for why Bluesky is drawing folks, I think largely because the functionality is VERY similar and easy to learn, albeit more rudimentary.
> P.S. the interface is so slow and laggy, holy cow
Maybe this has changed relatively recently in Teslas, as all my experience in them comes in the past 2-3 years, but the idea that somehow the Big 3 carmakers are *ahead* of Tesla in infotainment systems is completely crazy to me. Every Big 3 car I've ever been in has a UI straight from Nokia in 2004 with processing power to match. They are only remotely usable if you turn them into dumb glass by letting your phone take over everything, and even then "slow and laggy" is usually a great description of how they react.
Overall, I appreciated reading the comment, but the whole time I felt like I was reading one of those reviews of Apple products from a decade or so ago comparing how many gigahertz are in the CPU and whether or not it has a serial port to plug in a mouse, when that's not what customers care about *at all*. I don't know what "rigidity" means or what "panel alignment" even is.
I think he meant "Substack interface".
Hm, that's possible, but a couple paragraphs above:
> It really makes you question the customer sometimes, because if we put out a touchscreen that failed like that, we'd rightly be ridiculed. CEOs have lost their jobs over far less.
So just substitute that quote for the one I picked if so.
Yeah, ambiguous. My prior is based on the knowledge that the touchscreen failures were hardware-related, and substack's interface is slow and glitchy.
The biggest reason no other automaker had such a huge and beautiful touchscreen was/has been that no such screens were available that were automotive-qualified. Musk seems to have a tendency to ignore such nuisance obstacles (sarcasm), with predictable results.
Fwiw, the "predictable result" was that Tesla had to replace a whole bunch of faulty touch screens, which for our Tesla they did in our driveway at no cost to us, and the upside is that you have a massive screen that's actually big enough for the camera and Google Maps to be useful - smaller screens mean squinting and missing key details
Predictable nonetheless. The screens were failing, as predicted. Ye olde Toyota findeth such outrages not acceptable :)
I did!
We really need to have an Open Thread about the ACT substack experience. I don't know if it substack or the layer on top to make it more like the old SSC interface. But it is slow! And laggy! I read most everything on an ipad but I can't read ACT on an ipad because it implodes. Even on a PC, ACT lags. Its weird that no one talks about this more.
At this point I have abandoned reading Scott’s Posts on Substack and only read them directly in my Mail app. I use the built in iOS Mail app instead of Gmail because it doesn’t truncate long emails.
I was also confused by the rigidity comment. The "panel alignment" stuff I _think_ refers to thinks like the gaps between the doors and the body etc, which Tesla is pretty famously bad about (or at least, was a few years ago, apparently they have gotten better). Basically the "fit and finish" stuff.
But the rigidity comment I didn't understand. I'm curious what it is important for. It can't be for safety/structural integrity because my understanding is that Teslas outperform other vehicles on tihs by so far that they basically broke the scale when they first came out.
But overall, I think my response to all of that is: Consumers were willing to deal with all of that because everything else about the car was so much better.
I have a new electric car (not a Tesla), and while there are definitely things I would like to improve about it, I don't think I'd ever go back to an ICE vehicle again, and I'd be willing to accept a fair number of tradeoffs, including bad cosmetics/fit and finish, for the advantages I get from an EV.
In other words: reading that comment felt like watching Goodhart's law in action. The engineers at the big company had a ton of metrics they needed to meet around quality etc., but meeting those metrics meant that they couldn't move fast on the far more important things that consumers (or at least one very particular segment of consumers) cared about.
In a fully mature market, yeah, I'd like an EV that didn't have to make all those tradeoffs. But if my only options for getting a vehicle now are "take the tradeoffs or wait 5 years", I'll take the tradeoffs every time.
And yeah, even though I hate all touchscreen interfaces (and, other than overall price, was probably my biggest strike against Teslas when I was shopping), you can't argue that their touchscreen interface isn't better than everyone else's. All of them suck, but Tesla's sucks the least.
Goodhart's Law is a great reference point there. I'm sure the tensile strength of some of those Tesla parts is lower than spec 713(a)(b) for a GM car or whatever, but as a customer, what I'm interfacing with is the vastly-superior Tesla charging port. And I can buy a Tesla on their app today, for the price they say it is, and drive it around charging it at the big, reliable Tesla charger network.
You're also interfacing with the rest of the Tesla, including NVH, rate and handling / ride quality (what body stiffness contributes to, FYI), ergonomics, perceived touch / sound quality of the interfaces, wear (think seat and interior carpet), etc. As well as the lack of buttons, offset gauge information, weird stalks, etc.
In $current_year, the Model 3 is a lot better than the Model Ss of the past, but you're fundamentally paying German OEM prices for a much lesser vehicle with the key killer app being an EV powertrain. It's been interesting seeing who draws their line where over what - and will be very interesting now that the major OEMs are getting serious about EVs.
> You're also interfacing with the rest of the Tesla, including ...
Right, but I think this is where Car People go wrong, thinking that normal customers rank these things anywhere close to where they do. I care vastly more about the size and UI design of the infotainment screen than I do the perceived touch/sound quality of the interfaces. A lot of this stuff veers into mechanical keyboard or automatic watch territory, I think.
This is where things get difficult. As a consumer I have no visibility into things like "body rigidity", component reliability, etc. Teslas look great and test-drive amazing. It's only when you're in a crash 6 years later suddenly these boring automotive qualification issues make a difference.
Well, and the average vehicle on US roads is ~12 years old. When Teslas start hitting the used market in appreciable numbers we'll get to learn a lot about legacy auto requirements.
Except that Teslas seem to get best in class safety scores from every safety rating agency...
When you ask customers to rate where they consider these features, they always rate quite low compared to the basics (cost, fuel economy, safety, brand perception/signal). But there's a very sharp cliff - if the turn signal stalk feels like cheap crap (or the blinker sound sounds cheap) during a test drive, or if the door handle feels particularly bad, this can impact a cross-shop. Most cross-shops aren't really genuine (most customers have some idea of which car / brand they want to buy already, and are looking for justification), but 1. don't add fuel to the fire with bad design and 2. surprise and delight with better than expected design.
For the longest time Tesla was somewhat immune to this - if you wanted a particular type of EV, it was them or nothing. We'll see how the next few years pan out as actual competitors enter the sector.
I think that the general direction of this thread ("automakers prioritize many things that are customer-invisible and that in an EV rather than an ICE don't matter to the terminal use-case") is one that makes sense overall but I would definitely *not* say that NVH is only for Car People - if anything, it's the opposite. Car People like hearing the "engine note." I just want to hear my audiobook and/or music.
Goodhart in action again I think. To this normal person (nor a Car Person) the Tesla handles so much better than any ICE car that it's no contest. And it also handles vastly better than the electric cars from the legacy manufacturers (I test drove half a dozen). Maybe in five years time this will be different, but right now, it's no contest.
Maybe this is off-topic but can I ask why you have such a strong preference for EVs?
It's a relatively small number of things:
1: where I live, the difference between charging my vehicle and the price I would be paying for fuel saves me between $100-$130/month. This means that when calculating fuel + monthly payment, I could not have gotten a new ICE vehicle that met my requirements for cheaper than I got the EV, although that probably wouldn't have been the case if I'd been willing to buy used, since the used EV market is basically nonexistent at the moment. And once I'm done paying it off, the savings continue.
2: Relatedly, never having to visit a gas station is very nice. It's a minor thing, but I _never_ have to do it anymore.
3: I haven't owned this vehicle very long, but so far maintenance is considerably easier. Tire rotations are about it. Admittedly this one could change as I've got less than a year of data. _Theoretically_ EVs should have less maintenance, but that doesn't stop bad engineering from creating problems.
I will admit that at least part of my rosy outlook is that this vehicle is a significant upgrade from my previous vehicle which was a beater I bought in grad school that didn't even have power windows or AC. Just about anything I bought would be a significant upgrade, completely aside from ICE/EV status. But having fuel just basically never be a consideration for 95+% of the driving we do is great. And I've even taken it on a couple 10+ hour drives that involved multiple fast charging stops, and haven't really had any issues, so even for road trips, it's not anywhere close to enough of a downgrade from ICE to negate the benefits in day-to-day driving.
I have similar affection for EVs, with one minor caveat- routine maintenance is much easier, but maintenance for things that actually go terribly wrong is much more difficult (and mechanics tend to be out of their depth with EVs, especially in rural areas). You can't fix the stator-rotor by eyeballing it, it's difficult or impossible to replace a single part of the battery pack rather than the whole thing (despite the fact that usually only a small area is the problem child when things start going wrong) and many EVs have much more complex electronics than midline ICE vehicles. However, if you have the money to burn on full part replacement and live in an area with high EV use, these problems mostly go away. And they'll probably go away entirely in the next couple of decades as EVs become the majority of the automobile market.
That's fair. For my specific vehicle, according to estimated replacement costs of the battery, it will take ~6 years of my fuel savings to break even. Even if I only start counting from when the vehicle is done being payed off, the battery is likely to last longer than that unless I get pretty unlucky. Plus I think it comes with a 10 year warranty. So at least for my particular case, even if I have to pay for a replacement battery after 10 years, I'm still coming out ahead. But that's all pretty speculative.
Ah yes, I have recently discussed something like this with an artist with a 380Mm car with regular trips to other continents : no car after ~2002 is an option, because they have not so much electronics, but locked down computers that no mechanic outside of Europe knows how to fix and/or has the proprietary computer suitcase required for diagnostics.
Since EVs are much simpler, I wonder if we'll see again repairable cars that are EVs, sooner than repairable ICEs ?
Unlikely; you need computer-controlled battery management to make EVs practical. Those computers are simple to make, simple to install, and simple to program to brick your car if touched by a mechanic who isn't paying Danegeld to the automaker.
Yeah I just bought a brand new Subaru and the interface is trash-tier laggy, before that I had an Audi and it was at best ok.
I think this highlights the software vs. hardware concept of quality.
The Big 3 interfaces are ugly, lack functionality, and are slow but they are basically bomb-proof. Unlike the Tesla interface it will almost never go into a reboot loop or have some other major issue that requires a repair. The repair/failure stats are probably on par with other parts of the car, like how you expect 1 in x0,000 transmissions to fail during warranty or whatever. I have a Honda and the interface isn't great but I have done some weird stuff (disconnect battery while running diagnostics, mash random buttons) and it always recovers unlike my other computers.
Tesla however has a nice UI but a higher failure rate and more bugs. However, the bugs can be fixed and many failures can be fixed remotely. Legacy carmakers can't do that so they have to have highly reliable computers and interfaces.
Tesla has software-style high quality, where things are easy and pleasant to use. Big 3 have hardware-style high quality, where absolute failure rates requiring repair or replacement are low.
Additionally - NHTSA will have whatever body part of you they'd like if they find out you have a UI failure that gets anywhere near something they care about. Losing speed indication, for example, is frowned upon with similar intensity as a hood fly-up or loss of propulsion.
How Tesla has managed to outrun this is kind of interesting - their high software churn rate basically outruns the US recall system. A lot of what they change and update is safety-adjacent, but they outrun ODI with what you'd call (at any other OEM) "silent recalls". Whether this is better (debatably faster fixes) or worse (the recall system doesn't hold them correct to account on software issues and the high churn rate may introduce new failure modes) is an exercise for the reader.
My Bolt’s tendency to have infotainment boot-up fail and refuse to reboot itself for ~5 minutes, causing me to not have a backup camera when I turn it on, begs to differ with the idea GM’s software, at least, is bomb proof.
In the Tesla, there’s a steering-wheel button combination to reboot now, which I’ve used a couple times in a few years of ownership. Annoying, but much better than 5 minutes of can’t-even-turn-off-the-radio-which-insists-on-turning-on-when-I-don’t-want-it.
Overall the Bolt’s rate of infotainment-failed is about 5x/month of ownership in my experience.
Yeah, that comment read to me like the analog of a comment from a programmer about how Mac OS is so much worse than Linux. Like, yeah, maybe there are specific metrics on which Teslas are worse, and maybe those metrics are what the legacy automakers have optimized heavily for, but those metrics have roughly zero overlap with what J Random Consumer cares about. [Speaking as a random consumer who bought a model Y six months ago after comparing to half a dozen legacy manufacturers, because it simply blew all the offerings from the legacy manufacturers out of the water. I hadn't intended to buy a Tesla going in, but in the `head to head comparison' it was simply a walkover].
I'm late to the party, but re. The Boring Company, it's the most-visionary and most-important of his projects. I've been advocating just such a project for, I dunno, 20 years now.
Boring tunnels underground, evacuating them of air, and running vehicles through them, is the only way to achieve hypersonic travel on Earth. The theoretical limits on speed then are just the ability of the human body to tolerate g-forces. It's the obvious, necessary, and /final/ transportation technology for Earth; it can never be surpassed by anything short of teleportation.
Also, the first person to build a trans-American hypersonic tunnel system gets to choose where to build the transfer points. Buy a bunch of prairie in Kansas and build your west coast / east coast tunnel interchange there. A megalopolis will spring up over the next decade, and you'll have made up to a trillion dollar profit on those land purchases.
Entirely ignoring the physical problems -- the economic aspects of a hypothetical "hypersonic commute" ought to terrify people. In particular, the same people currently willing to pay half or more of their wage for the privilege of living in SV. Suppose the vacuum tunnel is built, now you can pay that same half+ of your wage for tickets on this train. (Not to mention, tickets to the late Concorde are likely to look like a bargain in comparison.) And now you get to "enjoy" all of the down-sides of space travel, right here on Earth: e.g. microscopic rock falls from the tunnel, now you're perforated as if from a meteorite strike, and breathing vacuum.
Why is the idea of moving meat every day hundreds (or, as contemplated, thousands) of miles so that it can drink bad coffee, watch Powerpoints, and push buttons, still seen as anything but insanity?
Because fundamentally, people want to be in places with other people, and transporting meat quickly is the only alternative to packing all those people in one place?
The massive rents people accept for SV should be a sign that we really could benefit from better transportation.
Much of the time the people "want" to be in places with others simply because this is the only way to get paid. There are alternatives to moving them in the flesh, but for so long as nothing (e.g. COVID) is forcing the issue, employers would rather force the meat to move at its own expense, or fight in bidding wars for cockroach flats.
Picture if employers were required to justify in-person presence requirements for office work to the same standard that they are currently required to do for IQ testing. (i.e. to an effectively-impossible standard, in front of the proverbial "rock with the word "No" written on it".)
As for better transportation, it seems that Moloch tends to eat most of the "win" in the long run. Compare the experience of having a car today to having one in e.g. 1930s USA.
If a for-profit teleporter were to be built between SV and Topeka, Kansas, the workers will be paying the teleporter operator something similar to the current difference in cost of living between the two locales, while rent-seekers quickly bid up the "new" land. And if the teleporter becomes gratis, the commuters will simply find their wages cut. Rather like the "cost of living adjustment" megacorps presently do (where you can move from e.g. Washington DC office of $company to the one in Taipei, and your wage will drop precisely in half.)
Plenty of people are in the bay area not because it's the only place they get money, but also because they have friends and family there, or like the culture there, or so on. There are many reasons to want to travel, commute for work is just one of them.
Plenty of people also actually prefer working face to face, and not alone at home in front a computer screen - it would be nice if this could be up to individual preference instead of company-wide directives, but having the option is still good.
I disagree with Moloch eating all or most of the win. I live in switzerland, my experience of transportation has been fairly comfortable, if expensive, to date. Making everything even faster would be strictly better, it would mean the "city" as measured by time to reach the center could expand tenfold. Removing transportation would be horrific and would crater the livability of the place.
It is kind of cheating if you're comparing the ease of transportation in a time when fewer people existed and fewer people traveled. I'm also not sure cars are the best example of transportation that scales well.
If a for-profit teleporter were to be built between SV and Kansas, the teleporter would hopefully be state-subsidized rather quickly, with 100 other teleporters built to connect SV to all over the country, so the rent-seekers would bid up land all over the nation. Isn't this what already happened with railway in all the nations where it's significant?
As for megacorps paying less, currently the money is being wasted towards landlords for uncomfortable housing. I don't know whether or how much of that money workers would be able to benefit from if housing became overall cheaper, but we would probably all benefit of it at least indirectly, there aren't many wastes worse than the rent currently going on.
It isn't clear that concentrated "Meccas" like SV would continue to exist at all if transportation could be anything near "a network of free teleporters". If workers could disperse to arbitrary geographies, why would employers refrain from doing so? (Or, for that matter, friends/family/cultural orgs, etc.)
See also Clifford D. Simak's "City".
Maybe! In fact, I almost hope so. That sounds quite utopic to me.
That said, given that in reality we probably won't have teleporters, we'll still have "hubs" or stations between which you can go *very* fast over long distances, and then use more humble means to get to your specific place. So you'll still have these hubs, in one form or another.
Yes it is clear. The lower transport costs become, the more significant tiny differences in transport costs become. If in place A costs are $100, and in place B costs are $100.10, the difference is immaterial. Lower the cost to $0.3 and $0.4, the same difference as before is now significant.
Costs are not only money; they also include annoyance, or lack of fortuitous meetings with others who also happen to be travelling in the same places.
As transport costs decrease, we should expect more agglomeration effects, industries converging on particular cities, unless other costs increase enough to stop that happening as may well be the case with SF and its poo problem.
Switzerland keeps Moloch at bay through hard work by millions of people to maintain a dominant culture that values solving coordination problems over individuals achieving personal utility maxima.
> Compare the experience of having a car today to having one in e.g. 1930s USA.
It's quiet, comfortable, air-conditioned, ~never breaks, streams the entirety of human knowledge to you in audio form, uses satellites in space to pinpoint your location and signal for distress if you crash, and is an order of magnitude cheaper?
Who said anything about commute ? The SUBsonic hyperloop would be somewhere between high speed rail and (supersonic) air flight : as you say only rich people would be able to afford it for commute.
SUBsonic hyperloop would be *much* safer than space travel. Air flight is very unsafe too (losing propulsion in flight !!), and yet somehow got popular.
Air travel is one of the safest modes of travel, and most commercial planes are designed to be able to fly with partial loss of propulsion (e.g. loss of 1 out of 2 engines) and to have good control even with total loss of propulsion. Once they're up to cruising height they're in a very safe situation.
The only mode of transport with better safety than planes is the one I insist upon but everyone disregards, elevators!
By the way, what's the point of subsonic hyperloop? that's just a high speed train on a gigantic vacuum tube, at that point you might as well lose the myriad of engineering issues that the vacuum tube brings, you're getting very little bang for you buck anyway.
According to Musk, compared to the alternatives (driving, train, flying), the hyperloop would be :
Safer
Faster
Lower cost
More convenient
Immune to weather
Sustainably self-powering
Resistant to Earthquakes
Not disruptive to those along the route
(Supersonic hyperloop is a contradiction in terms. EDIT : https://www.astralcodexten.com/p/highlights-from-the-comments-on-elon/comment/40389290 )
A big cost driver for transportation tunnels are the emergency and safety aspects:
* how do you get everybody safely out in case of a fire/power outage/obstruction
* how the emergency services can reach people or equipment in the middle of it all
I question this. We have had the capability for supersonic travel for decades, yet it has never taken off (aside from Concorde). Maglev trains, though much faster than current trains, have also failed to see any application outside niche demonstrations. I'm not sure why exactly - perhaps there's just not enough demand for high speed travel? But if these two technologies have not been widely adopted, I don't see why hypersonic tunnels would be.
The fastest that a Concorde could take you from New York City to San Francisco, ignoring time spent going through airport security and sitting on the runway, would be about 2 hours. The fastest a vacuum tunnel could take you, with a continuous acceleration of 3g, would be twelve and a half minutes.
A sad consequence of this is that it would separate urban and rural people even further. Every major American city would be closer to every other major American city than to any rural city.
I don't think twelve minutes at 3g is at all plausible for a commercial transportation system, for several reasons.
My take is that the Concorde ultimately disappeared primarily because it couldn't reach the volume of passengers and routes required to start driving economies of scale and demand for better designs. With the first-generation design, it was only able to compete for first-class passengers, and even for those it couldn't pull in much more money, if any at all, than subsonic first-class seats.
The operating cost premium of Concorde flights made selling economy-class seats infeasible, especially since there wasn't really enough room in the fuselage to put nicer seats and differentiate the classes. So a transatlantic round-trip Concorde ticket cost (according to Wikipedia) about $15,000 in today's dollars, and they were competing with first-class subsonic flights, which go for a similar price.
So you're rich or you work for a company with deep pockets, and for the same cost, you cut your travel time in half or a little bit more but also get a less comfortable seat. Meh? It's certainly a nice option to have, but most of the reason people hate long plane rides is that they're uncomfortable, and they're not nearly so bad in international first class. And you're only spending a few hours on the routes the Concorde ran on in the first place. (Nowadays it'd be even worse, since you can get wifi in your subsonic first-class seat and keep working on the plane, so you don't even lose much time for the company that's paying for your first-class ticket.) So while most people would take a supersonic flight if the cost is equal, the benefit is marginal, so you can't really charge much more for a supersonic flight than for any other flight...
...especially because you need to attract as many people as possible – you have to fill an entire 100-seat plane with first-class passengers! Which is the second problem, because there just aren't many routes that have that much routine demand from people willing to shell out $15K per flight (oh, and by the way they also have to be overwater so you don't have everyone angry at you for making obnoxious sonic booms everywhere, so NYC–LA, London–Dubai, etc., are out).
British Airways at least was able to hit the sweet spot and make a nice profit on operations, contrary to popular belief – but only on a handful of routes like London–NYC. With the current market consisting of a handful of planes on a handful of airlines, it's nearly impossible to make money on a better plane design unless you do so much better you can massively expand the market (and the improvement required here in one go would be huge, given the gap between first-class and economy fares). So nobody tried, and the technology stagnated.
A combination of rising maintenance costs because the design was aging and a nasty accident in 2000 didn't help the Concorde, but ultimately at that point it seems the main problem was that no airline saw any way they could end up in the black keeping it versus focusing on subsonic first class – unless the tech got cheaper, it was a lot of extra bother to run these flights without a whole lot of benefit.
I have nowhere near enough domain knowledge to guess whether early hypersonic tunnels would have larger or smaller operational and amortized design/construction costs than the Concorde.
Transportation via rocketship would be faster than a vac-train, beyond a certain distance.
A very large distance, if you count the hours that all of the overhead that comes with flying with a rocket brings. This video goes into some of the details: https://www.youtube.com/watch?v=02pFTSMbevY
And let's not even mention the fuel tab. And the fact that historically, something like 1% of rockets go boom. And all the other reasons why this is, and will forever be, sci-fi.
Fuel costs are likely to be negligible by the end of the century (converting CO2 via abundant solar energy) and rockets as reliable as 1970s aircraft. The time scope given was infinite, but even looking to 75 years from now it seems possible.
Sound is a big unknown, there hasn't been much pressure to develop quieter launches yet. Underground launches, active sound suppression technologies, these are potential mitigants.
To be clear I really don't think this will happen in the next 50 years, but in 100 years I think it will be possible. Maybe the economics will not work out, maybe people will not put up with the sound, but it will for sure be easier than a 5000km vac-train.
It wouldn't be faster at the theoretical limits, because the rocket would have to go a longer distance, and pass through the atmosphere on the way up and the way down. And it would require much, much more energy. So it's strictly inferior, in a theoretical sense, although it's technically easier (since we've been able to do that for 60 years now).
The point of the tunnel is to let you create a vacuum, which is the only advantage space has over airplanes. You could of course use a rocket motor to go through the tunnel, but there's no need for that, since less-flammable technologies can give you higher acceleration already.
Theoretically it's possible to maintain a tunnel at a perfect vacuum. Realistically I haven't seen anything to show that that's going to work. The current upper limit for vac-trains is around 1000km/h. Going faster just gets exponentially more difficult.
Yes, yes, of course things we've already done are easier than things we haven't done. I'm talking about Musk's /vision/. Travel by evacuated tunnel will be very hard, and might not work within Musk's lifetime; but it is the theoretical best mode of travel. It is the logical endpoint of Earth transportation technology.
The hyperloop would be subsonic, not supersonic.
The whole point of the concept is to prevent the formation of a supersonic air cushion in front of the vehicle that would then only keep growing as more and more air piles up in front of it, prevented by the speed of sound from going around fast enough.
"Supersonic" with a pressure of 1 atm (and some reasonable range of temperature) means over 1230 km/h. Hyperloop max speed would be 1220 km/h, I wonder if it's a coincidence ?
(BTW, "hypersonic" means more than 5 times the speed of sound.)
Of course, hyperloop is supposed to work at only 0.001 atm, which is equivalent to that of the stratopause which is at ~50 km (between the stratosphere and the mesosphere, interestingly with a temperature around 0°C), at which point the speed of sound is... still 1190 km/h ! Yeah, not a coincidence, still, it's weird that hyperloop *is* effectively ("barely") supersonic at that speed, maybe they have forgotten to adjust for air pressure in these early calculations ?? (Again, breaking the sound barrier is something to absolutely avoid, consider also the damage from repeated supersonic shocks !)
Your typical commercial flight operates instead around 10km, with a typical temperature of -50°C, and a pressure of 0.240 atm, which corresponds to a speed of sound of already only 1080 km/h.
Huh, I wouldn't have guessed that the speed of sound was so much more sensitive to temperature than pressure !!
https://aerospaceweb.org/question/atmosphere/q0112.shtml
P.S.: It's also planning to limit forces to 0.5g.
Wait, you mean the tunnels aren't supposed to be evacuated of air? I guess at some point I assumed that's where he was going eventually. How embarassing. :P
I'm less-excited about them then, but they'd still be great.
I would call 0.001 atmospheres to be "pretty evacuated"...
P.S.: You cannot achieve total vacuum, you can only try to approach it as much as you can...
Also, Hyperloop could not work without any air at all : it uses some of that air to generate the air cushion on which the pods glide !
Current plans don’t use an air cushion.
https://www.hyperlooptt.com/technology/
Re the boring company, it's worth noting that (unlike rockets or EVs), building tunnels for an order of magnitude (or more) less than is typical in US infrastructure projects is already regularly done in many other countries (including countries with similar or higher labour costs) - and that's for full-sized train tunnels, not just small-diameter tunnels for cars. That the Boring Company has failed to even match that for their smaller tunnels would imply they're probably not doing a great job on the engineering side.
Edit: I looked up some numbers and this is at least somewhat wrong, looks like they actually did match the low end of european costs (at least, assuming we can take their claimed costs at face value).
I thought most US-elsewhere differences in tunneling costs were political.
It's confusing (much like healthcare costs). Broadly speaking with rail tunnels there's two political issues - unreasonable requirements (e.g. asking for overly large tunnel diameters and big stations, or requiring mined stations instead of allowing cut and cover) and paperwork costs (e.g. getting into legal disputes with contractors or overhiring planning consultants).
But on the other hand, American (and latam) construction costs for roads are also an order of magnitude higher, and they don't seem to have higher technical requirements (the legal requirements might still be an issue, but I'm more confused about that). I don't know of any in-depth study for this, but it seems like there's a general cost premium for US infrastructure projects that's a bit hard to explain.
Transit Costs Project is the closest you're going to get for an in-depth study. Run by NYU.
"I’m not sure the Boring Company is interesting enough for this to matter."
Has anyone pointed out that of *course* the Boring company is not interesting; it is, of course, boring. That's straightforward nominative determinism.
I always thought he named it that on purpose as a pun; it's a company where they bore through the earth, and also it's just a less interesting company.
> Or did they all start learning to paint and spending time with their friends and families?
This coincided with Zvi starting to release novel-length posts on AI every week, so I just switched my twitter time to those (and presumably everyone else has too).
My social media engagement statistics plummeted at the time Zvi started the weekly AI roundups. Much better use of time than snarking for someone else's AI training funnel.
"then of his top-300,000-most-intelligent-Americans cohort, he would be in the top 30 most intense, or alternatively, for his top-30,000-most-intense-Americans cohort, he would be in the top 30 most intelligent."
This rings true, in some approximation of the specific numbers. Over some decades I've met or known or been related to a few people who I'd place that high on one or the other of those lists, but, pondering it a bit now, no one who I could list that high on _both_. That _combination_ seems exceedingly rare.
It also reminds me of a sports quote: "Michael Jordan is what happens when a sport's greatest athlete and its greatest competitor happen to be the same person."
Same here. I've known some very intelligent people, and some very intense people, but none who are strongly both.
I'm surprised that the lies about Musk getting into Stanford's PhD program are still being repeated as evidence of his intelligence. This twitter thread goes through the evidence: https://twitter.com/capitolhunters/status/1593307541932474368
In summary, Musk claims that he graduated UPenn in 1995, was accepted to Stanford's physics program, attended for two days, and then dropped out. But none of this is quite true. Musk did attend UPenn, but wasn't awarded a degree in 1995, but rather in 1997, two years after he supposedly graduated- from the eminently bribeable Wharton school. Coursework that Musk posted reveals that he was taking sophomore classes in his junior year. And while UPenn now claims Musk got a physics degree from them as well, the 1997 graduation announcements don't include his name (https://pbs.twimg.com/media/FhyREt6XoAAbh2p?format=jpg&name=large). The implication is fairly obvious: Musk struggled in school, didn't manage to graduate from UPenn in 1995, went to SF to work at tech startups, and then after he'd raised some money and there was the potential for embarrassment (and legal issues) if it became clear that he didn't have a degree, he arranged with UPenn to be given a diploma two years later. Musk has made vague allusions to unfinished classes as a reason for this mixup, but at the time he was on a student visa- it is not believable that someone would "forget" that they hadn't attained the degree that would keep them from being an illegal immigrant.
Meanwhile, the story about getting into Stanford's physics grad program is total bunk. You do not get to start a graduate phd program without an undergraduate degree- if Musk was ever accepted (doubtful, if he wasn't a good student), his acceptance would have been rescinded the spring prior to his starting semester, when he failed to graduate. Musk has been inconsistent about what department he was accepted to (MSE, and Applied Physics have both been claimed), claims that he was accepted to work on a topic ("advanced capacitors and batteries") that the professor he claims he was hired by did not work on at the time, and when subpoenaed, Stanford's graduate admissions office was unable to locate any records relating to Musk (https://pbs.twimg.com/media/FhySnhNXgAQnFfW?format=png&name=medium). Plus, Musk also has claimed that he moved to SF specifically to go into the tech business, something that couldn't be true if he moved to start a 5 year PhD program.
The synthesis of all this is that Musk had an unexceptional academic trajectory, for someone getting into tech in the 90s. He transferred into an elite university, gamed a bit too much, struggled in his classes, didn't quite manage to graduate, but moved to SF at the right time and made loads of money at a tech startup anyway. Nothing about this would be objectionable, but Musk has told stories about his academic career to cultivate a persona of being a scientific genius, and it's just not true.
Vance has a section on this in the book:
"At first, I, too, felt like there were a lot of oddities surrounding Musk’s academic record, particularly the Stanford days. But, as I dug in, there were solid explanations for all of the inconsistencies and plenty of evidence to undermine the cases of Musk’s detractors.
During the course of my reporting, for example, I found evidence that contradicted O’Reilly’s timeline of events. Peter Nicholson, the banker whom Musk had worked for in Canada, took a stroll with Musk along the boardwalk in Toronto before Musk left for Stanford and chatted about the incarnations of something like Zip2. Musk had already started writing some of the early software to support the idea he’d outlined to Kimbal. “He was agonizing whether to do a PhD at Stanford or take this piece of software he’d made in his spare time and make a business out of it,” Nicholson said. “He called the thing the Virtual City Navigator. I told him there was this crazy Internet thing going on, and that people will pay big money for damn near anything. This software was a golden opportunity. He could do a PhD anytime.” Kimbal and other members of Musk’s family have similar memories.
Musk, speaking at length for the first time on the subject, denied everything alleged by O’Reilly and does not even recall meeting the man. “He’s a total scumbag,” Musk said. “O’Reilly is like a failed physicist who became a serial litigate. And I told the guy, 'Look, I’m not going to settle an unjust case. So it’s just like don’t even try.’ But he still kept at it. His case was tossed out twice on demur, which means that basically even if all the facts in his case were true, he would still lose.
“He’d tried his best to like torture me through my friends and personally [by filing the lawsuit]. And then we’ve got summary judgment. He lost the summary judgment. He appealed summary judgment, then several months later lost the appeal and I was like, 'Okay, fuck it. Let’s file for fees.’ And we were awarded fees from when he appealed. And that’s when we sent the sheriff after him and he claimed that he had no money basically. Whether he did or didn’t I don’t know. He certainly claimed he had no money. So we were like either we’ve got to like impound his car or tap his wife’s income. Those didn’t seem like great choices. So, we decided that he doesn’t have to pay back the money he owes me, so long as he doesn’t sue anyone else on frivolous grounds. And, in fact, late last year or early this year [2014], he tried to do just that thing. But, whoever he sued was aware of the nature of my judgment and contacted the lawyer I used, who then told O’Reilly, 'Look, you need to drop the case against these guys or everyone’s going to ask for the money. It’s kind of pointless to sue them on frivolous grounds because you’re going to have fork over the winnings to Elon.’ It’s like go do something productive with your life.”
As for his academic records, Musk produced a document for me dated June 22, 2009, that came from Judith Haccou, the director of graduate admissions in the office of the registrar at Stanford University. It read, “As per special request from my colleagues in the School of Engineering, I have searched Stanford’s admission data base and acknowledge that you applied and were admitted to the graduate program in Material Science Engineering in 1995. Since you did not enroll, Stanford is not able to issue you an official certification document.”
Musk also had an explanation for the weird timing on his degrees from Penn. “I had a History and an English credit that I agreed with Penn that I would do at Stanford,” he said. “Then I put Stanford on deferment. Later, Penn’s requirements changed so that you don’t need the English and History credit. So then they awarded me the degree in ’97 when it was clear I was not going to go to grad school, and their requirement was no longer there.
“I finished everything that was needed for a Wharton degree in ’94. They’d actually mailed me a Wharton degree. I decided to spend another year and finished the physics degree, but then there was that History and English credit thing. I was only reminded about the History and English thing when I tried to get an H-1B visa and called the school to get a copy of my graduation certificate, and they said I hadn’t graduated. Then they looked into the new requirements, and said it was fine.”"
I don't think that biography quite assuages my doubts- yes, O'Reilly lost lawsuits, but as Musk himself points out, he lost on summary judgement, before any factual allegations had to be proven. It wouldn't have mattered if Musk hadn't actually gotten into Stanford, or if he'd been given a degree on the sly a couple years late- for all legal purposes, Musk is a graduate of UPenn. In a later lawsuit where similar facts were brought up, Musk quietly settled.
But that doesn't explain the problems in the narrative Musk tells. He says "They’d actually mailed me a Wharton degree. I decided to spend another year and finished the physics degree", but Musk didn't get either degree until 1997, according to UPenn (https://www.plainsite.org/documents/tbdmox/2019-email-from-the-university-of-pennsylvania-confirming-elon-musks-physics-degree/). And the diplomas have issues, as well- unlike his economics diploma (https://mediaproxy.snopes.com/width/600/https://media.snopes.com/2022/12/musk-econ-penn.png), which notes that it is a "Bachelors of Science in Economics", the physics diploma (https://mediaproxy.snopes.com/width/600/https://media.snopes.com/2022/12/musk-physics-penn.png) doesn't list a department at all. This could be due to how different schools at Penn design diplomas, and irritatingly, I haven't been able to find a picture of a different Penn arts school diploma to compare because all the google results for "UPenn physics bachelors diploma" are Musk's diploma. But it raises an eyebrow. As does, of course, the fact that UPenn still has general education requirements (https://www.college.upenn.edu/gen-ed)- maybe they changed between 1995 and 1997, but as of right now it certainly looks like you need to take english and history classes to graduate.
And the PhD program part of the story still has issues, as well. I haven't heard of any PhD program admitting students on the promise that they'll finish their bachelors *after* they get the PhD. Maybe Musk was given provisional admission and didn't end up attending after he didn't graduate. But it strains credulity that someone on a student immigrant visa, without a degree and therefore unable to transition to a professional visa, would just drop out of a PhD program and not realize that they hadn't attained a bachelors degree. The immigrants I know are all very aware of the requirements around degrees and jobs, because screwing something like that up can get you deported and banned from returning to the US.
">>>yes, O'Reilly lost lawsuits, but as Musk himself points out, he lost on summary judgement, before any factual allegations had to be proven. "
This is misleading and best, straight up incorrect at worst, and possibly conflating a motion to dismiss with a motion for summary judgment. Summary judgment is generally based on the factual record as fully developed (at a minimum, in relevant part, but often for the whole case) -- there are no more facts that are going to come in. If you win on summary judgment it means that you have established that *no reasonable juror* could find in the other side's favor because there is *no issue of material fact* that could support their position, and thus the issue shouldn't proceed to trial and you are entitled to judgment as a matter of law.
Isn't there a lot of separately implausible things that all need to be true for this to make sense, compared to the relatively simple hypothesis that someone with the means, motive and opportunity to embellish his academic record and then cover it up, did so?
Weird, to me it seems extremely implausible that these prestigious schools were all conspiring with Musk to fake his degrees/acceptance compared to the simple hypothesis that he's telling the truth.
I could very well be wrong, I was vaguely aware of there being allegations against the authenticity of Musk's academic merits before, but this is the first time I've taken a deeper look at the arguments for and against this claim.
But at least going by the comment from wax above (which I haven't verified, but the comment provides plenty of sources, and nothing in it triggers my bullshit detector), there are just too many odd things that happen to be just right for Musk's claim to hold up: Penn just happening to drop the requirement for the courses he was missing a year or two later (and then happening to reinstate these requirements at some later point) Stanford somehow accepting him into the PhD program before his bachelor was technically completed... to me it just reads too much like one of those contrived sitcom plots where, through some implausible sequence of perfectly innocent reasons, two characters just happen to look like they're engaged in something risqué the moment a third character walks in. Which, I am sure, must have happened to some people at some point in real life, but far less often than people get caught in what looks like a risqué situation because it is.
Schools can and do change graduation requirements as frequently as each academic year, and graduate schools frequently accept undergraduate applicants before they have completed their bachelor's on the condition that they complete their bachelor's (and usually maintain a certain GPA). These are both very basic things that happen.
2023 Musk, sure, elite colleges give out honorary degrees and other weird stuff to particularly rich, famous, or powerful people. 1995-7 Musk? There's very little reason for them to know him or care at that point. Why would they want to conspire with some 25-year-old about faking a degree?
Sounds like those underwater pyramids are definitely Atlanean in origin.
(i) I know someone who has a PhD from Stanford, and remembers Musk (very briefly) being in his PhD program. (ii) what does it matter anyway? That he got into a PhD program at Stanford is not even in the top-ten most impressive things about Elon Musk.
I'm a Physics professor. I know another Physics professor who has a PhD from Stanford, and who tells me that he remembers Elon Musk as being in his PhD program (same year), up until Musk dropped out. I consider this dispositive. Musk was admitted, and he showed up (admittedly he didn't stick around for very long).
Musk supposedly dropped out after only two days, he must have made quite an impression ?!?
(Which I guess is plausible, going by other accounts of Musk's random brilliance..?)
Yes. Apparently Musk tried to convince my friend to drop out with him and join him in founding [some startup] but failed to persuade. But apparently he gave a very intense sales pitch.
Ok, now I'm getting confused : was that a graduate (=masters) program, or a doctoral program ?!?
I'm guessing that Stanford is a bit weird in that they combine both into a super-program lasting 5 years ?
Having a Master’s before getting a PhD is the usual process but there are schools and programs that accept applicants into PhD programs that do not have a Masters.
> AI experts, is this a big deal? Can other AI teams not access Twitter data through the public web? Is it a substantial amount of text compared to other corpuses? Is the structure (280-character blurbs written by morons) a limiting factor? Or is this a genuine treasure?
My guess: the main thing Twitter will be good for is having a good source of links to high quality articles. This was what Reddit was most useful for wrt training GPTs back when its API was free [1]. Highly upvoted articles linked on Reddit were assumed to be good, and so included in the training runs. Otherwise you just get a bunch of SEO garbage. Similarly, highly liked tweets linking to stuff by verified users likely won't be SEO garbage, so will be useful for building datasets with.
[1] https://arxiv.org/pdf/2101.00027.pdf#page=4
Twitter overall has value as training data for LLMs but it's not crucial. I wouldn't consider links to be very valuable. (I do have relevant experience).
Why wouldn't you think links are so valuable? I wouldn't be that surprised if Twitter links to more quality content than it generates (likely it generates a lot, but much is also likely bots).
Would you want to limit your web training data to only linked articles? Clearly not, as this would significantly reduce the size of your training dataset and adversely affect the resulting model.
So what's left is using Twitter links as some type of a quality signal. But it's more likely these links point to stuff that's controversial, political, or appeals to a wide audience. This is the kind of content that's easy to find, so it's not what you're missing in your training data. And that's assuming you're even using ratings of content pieces to train your LLM in the first place.
What's been done recently is using a known, good LLM to rate and filter training data for a smaller LLM but that's quite different. You could likely hand-craft a rating system that would be better than any link signal from Twitter, simply based on the domain name where the content appears.
Maybe, but I’m skeptical. We do filter datasets based on social media engagement. I’m guessing you think The Pile shouldn’t have gone with their Reddit link approach?
These crawled links accounted for only 62.77 GiB of The Pile. At 4 chars/token, that's 15.7B tokens. Newer open datasets, which don't use this strategy, reach 6.3T tokens (https://arxiv.org/pdf/2309.09400.pdf), making this just 0.25% - a rounding error.
I've seen no evidence that this strategy of using social media links improves the model compared to adding more data from top domains, or that it's even a decent indicator of high quality diverse training data.
Good points! You have changed my mind.
My understanding was that Deleted Tweets, Direct Messages, and GroupChats were high value for Twitter, would that content also be valuable from the perspective of LLM training? It is data that isn’t publicly available.
"Moon Moth writes...[kid's names]"
Can you imagine the drama?:
"he has had 11 children by three mothers. ... Do the mothers get on? “Not with each other,” Isaacson jokes. And sometimes not with Musk:”
-- ‘He is driven by demons’_ biographer Walter Isaacson on Elon Musk _ Financial Times, 9/11/23
Personally I stopped using Twitter when they started requiring Twitter Blue to use Tweetdeck. The old Twitter interface was good, the new one sucks, but hey, at least there's Tweetdeck! Except, oops, now you have to pay for that. So I just stopped reading Twitter.
Re the AI potential from the Twitter archive
" Is the structure (280-character blurbs written by morons) a limiting factor? "
I think this question answers itself quite succinctly
🤓☺️
I think the Twitter corpus contains some alpha that isn't captured by other corpora, even if one ignores the temporal aspect. Reddit probably has a more valuable corpus right now but they are smaller and seem to have antagonized many of their their core users, just like X.
Elon seems like a jerk, which really bothers me on a personal level and influences how I view him.
I'm very non-confrontational (even cowardly, at least with regard to social interaction). I wouldn't be able to handle working for him even if that was somehow an option for me. However, I get a lot of vicarious satisfaction from watching someone tech-inclined who's both bold and powerful enough to tell the whole world to shove it, even when I think the world is right and he's wrong.
I apply Stafford Beer's "the purpose of a system is what it does".
Musk has done two unique (and to me very valuable) things.
First, he changed the mindset of many western governments, and China's government, from "electric cars are a cute idea but totally impractical" in 2010, to, in 2022, "you may not sell anything BUT electric cars as of [a date in the 2030s]." In twelve years!
Second, he is in the process of creating an in-principle uncensorable world-wide communications network that cannot be stifled by government action, or by the absence of government (say in a rural village in Afghanistan or Eritrea.)
The rest is noise.
I suppose you're right in places where there is no government but I don't think you can operate Starlink in China or Russia for long without censoring.
The Great Firewall has no roof.
That presumes you can get the equipment and have a way to pay for the subscription. I think it's far more likely that they will offer the service with the censorship in order to make money - at least that's what Twitter/X is doing.
<speculation>Perhaps X is just playing nice with the expectation that enough dissidents will "obtain" Starlink terminals that the content will be effectively uncensored.</speculation>
I actually give Musk less credit for the EV adoption piece. 2 points. 1) incumbent electric cars were fairly boring and lame before tesla, but tesla's existence (pre-musk) is proof that the market would figure out the appeal of EVs eventually 2) the price of lithium-ion batteries is a massive driver in EV affordability, and the price dropping is purely the result of mining/refining/assembly infrastructure maturing
> If Parag or whoever employed a thousand censors to keep the ADL happy, and the ADL becoming unhappy cuts Twitter profits by 60%, then there’s a strong business case for those censors!
Wait. Zoom out from Musk for a moment. Surely there is something about this state of affairs about which we should be very unhappy. What you have described is essentially an extortion racket, not unlike the ones that the mob sometimes ran. Saying "there's a strong business case" for paying the extortion money doesn't seem like something we want to normalize. Why is OK for outside groups to impose such costs? What if there were an analogous, competing group making demands in the opposite direction? Would that be just fine too?
I agree this is bad; I'm not trying to make a moral judgment here, just a trying-to-figure-out-why-a-business-works-the-way-it-does judgment.
Oh, OK. It sounded like you were holding it to be actually neutral. I agree that, descriptively, this is how big tech has come to work.
But I also think that you can steelman Musk's actions as a higher-level judgment that the business landscape itself has become corrupt and needs to change, even if that costs Twitter in the short run. A bit analogous to SpaceX pursuing a different business model from previous spaceflight contractors.
Yeah, if you trace back all the references, my original claim was "Parag must have been really dumb to hire 10x more people than he needs", and this was what I was updating on.
Re: hiring 10x more people than necessary, it may be something discussed in https://danluu.com/in-house/ [0], where, tl;dr Twitter had several world class experts at (simplifying the systems involved) operating systems and programming languages who are nominally not directly working in the product, but regularly can have their expertise called on to solve "unsolvable" bugs. When I think of the number of people lost, I would assume that there are lots of examples of this during the earlier Twitter purges. And their disappearance won't be "missed" until the next large "unsolvable" outage comes and lo and behold, all of the expertise has left the company.
[0] Note that while the blog post makes many confident claims about cost savings and counterfactual a regarding incidents, I have not actually seen their reasoning at work and I have a low grade suspicion that the blog author is far less careful about finding alternate explanations compared to their apparently high confidence level. This is a vibe and should not be interpreted as well reasoned.
As a software engineer, I find it highly implausible that the experts of old Twitter had some magic secret sauce that Elon would find difficult to replace via new hires.
> tl;dr Twitter had several world class experts at (simplifying the systems involved) operating systems and programming languages who are nominally not directly working in the product, but regularly can have their expertise called on to solve "unsolvable" bugs.
In reality this practice is more about prestige and recruiting than unsolvable bugs. You get these CS celebrities at big companies, they do a bunch of research and sometimes you can apply it, and they give talks with your logo on the slide deck which helps you recruit the programmers the next few levels down from them.
For companies like Google and Facebook that are throwing off billions of otherwise-unspendable dollars, this is worth it. At Twitter maybe not.
Why are you assuming that these high level technical experts got fired, rather than various low level, and especially non-technical employees (that also weren't even at Twitter a few years before that) ?
There was a lot of talk previously on Twitter about how technical talent got hollowed out, and how it was sure to fail around the World Cup because it was a period of peak usage. Of course that didn't happen, but if it was a called shot it'd be for the more recent spate of problems and it was just called too early.
Also things like https://gizmodo.com/elon-musk-twitter-layoffs-engineering-spreadsheet-1849767712 (link to spreadsheet: https://docs.google.com/spreadsheets/u/0/d/1OoD3pVxFl718fnxs_cEaOIbcj-45-qNKZFnE1jQW86M/htmlview?pru=AAABhHfrL8U*W5b4OE8SZeHN3325cn-rdg)
Also there is a certain common dynamic to lay offs / hiring where if a particularly competent member, or a member with high status within engineerin, but maybe not management leaves that can also cause a chain reaction of team members who lose faith in management then leave, or whose job can no longer be performed without the star member.
Obviously not great evidence, which is why I presented this as speculation but hardly unsupported. Do you have a model that fits more of the facts (layoffs leading to instability) or do you dispute that some details aren't facts? (Twitter is still stable, crucial engineers were re-hired and so on)
Elon obliges the censors too. Not the ADL/The Left/EU/mainstream media block but actors like China, Turkey and India and he pretends to not censor and pretends doesn't face many of the tough choices as the previous Parag regime (https://twitter.com/kenroth/status/1657635913478946816 , https://web.archive.org/web/20230523161245/https://twitter.com/elonmusk/status/1657422401754259461)
Also I think the pre-Elon regime was worse on censorship and I like that he tells the ADL/The Left to get lost and has largely not engaged in censorship of the Right.
- I expect him to acquiese to the Chinese censors too since Walter Isaacson reports in his book that Elon Musk told Bari Weiss last year "that Twitter would indeed have to be careful about the words it used regarding China, because Tesla's business could be threatened. China's repression of the Uyghurs, he said, had two sides." (https://twitter.com/MattGertz/status/1701968224001081480, https://twitter.com/lynaldencontact/status/1644492940394344450)
He may simply feel more invested in the US, and like free speech is a battle that's winnable there.
Wait how is it an extortion racket? Is ADL asking that Twitter pay them? There really is a lot more antisemitic stuff on Twitter these days, I don't see why it's unreasonable for that to make advertisers more reluctant.
There are many examples of ADL making demands regarding political disputes that have nothing to do with antisemitism.
Its only an extortion racket if you think that the things that ADL points to are things that otherwise noone would care about or have no impact on the business, which seems unlikely. You could equally say that money spent on health and safety is extortion money for insurance companies/health department/etc
Many of the things about which the ADL makes demands are not things like antisemitism but rather disputed political matters.
The things the ADL has been complaining about Twitter allowing/condoning seem like normal, common-or-garden antisemitism. The sort of things an advertiser might be worried about even if there wasn't an organization devoted to pointing it out to them.
The ADL define anti-semitism as, amongst many other over-broad things, being critical of the ADL itself, so their definition of the term is deceptive and manipulative. Read their report and see for yourself:
https://www.adl.org/resources/blog/threads-hate-how-twitters-content-moderation-misses-mark
If you look at the first tweet they call out as evidence of Musk-enabled Jew hate, it's literally someone tweeting the ADL's own words back at them. They claimed he took their words out of context (he didn't), and that he was to blame for the ADL disliking any replies, including indirect replies.
Did Trump literally say the words "expelling warmongers, driving out globalists, casting out communists, and throwing off those who hate our country", or did he not ? (If not, did he says something very similar ?)
(My suspicion is that this is yet another example of the Twitter character limits causing damage.)
Because the alternative is "the ADL is not allowed to talk about content it finds distasteful on Twitter" and that's not how free speech works. The ADL is not putting a gun to anyone's head, it's saying "we don't like this content " and advertisers are voluntarily saying "we agree, it's bad for business for us to be associated with that content."
>What if there were an analogous, competing group making demands in the opposite direction?
They'd be protected by the same right to free speech that protects the ADL, and advertisers would have to decide which group they find more convincing on if they should pull their ads or not. The marketplace of ideas in its most literal form.
The thing any conversation about Musk inevitably reveals about the VAST majority of people who otherwise consider themselves intelligent is that they are unbelievably susceptible to manipulation and propaganda, and that there are many many midwits who love to look down their noses at someone that is deemed 'smart'
Fewer comments like this in the future, please.
Banned.
I’m very conflicted about Musk, because it seems to have it’s own particular version of Gell-mann’s amnesia.
When he talks about something you’re really good at, it’s hard to not see he is talking BS, but when he talks about other stuff, he really sounds like an expert. We should be discounting his views in everything else, but instead most people inflate his opinions of what they know is wrong.
A lot of people here know a lot about AI, but if someone expressed Musk musings on AI around here, there would be much less incentive to try to “read” deep insights.
It’s also interesting as someone who has worked in manufacturing that a lot of his “bright” ideas seem to be the kind that get you in trouble *on the long run*.
For instance, while the aluminium frame story makes him sound super smart if you stop where the book does, it takes on another look once you realize Tesla has had to settle lots of lawsuits related to the aluminium frame. The reason for not using aluminium is that you *really* want your car frame to be a single BIG piece, not several pieces welded together, especially in an EV where your frame protects the battery in case of impact. By welding you create stress points that tend to shear on crashes, which is one of the reasons Teslas are notorious for going in flames even after relatively minor crashes.
I am an aerospace engineer with experience on rockets. Musk knows what he's talking about on rockets.
That’s what I keep hearing, but I’ve not seen a single instance where he showed “knowing rockets” past what I’d expect for the average KSP player.
In the meantime, he okayed the idea of point to point rocket travel, which is so full of holes that IMO it puts into question how much he knows.
Point to point rocket travel has been done:
https://www.youtube.com/watch?v=wv9n9Casp1o
The points were close together for DCX, but putting them farther apart is just a matter of logistics, not physics.
It’s not a problem of “being done”, but “how”.
Musk proposed using Starship for P2P, one of (if not *the*) loudest rockets there is. If planes are inconvenient because of how far they need to be from urban centers and how much space they require, heavy rockets are hardly a good solution.
Then there’s the obvious risk factor. Even the best rockets aren’t reliable enough for commercial passenger travel. We’d need orders of magnitude improvements on reliability for the idea to be feasible.
Looked with any detail, the concept is an obvious non starter.
is the idea here that within-the-hour transport to anywhere on Earth's surface isn't useful, because you might have to take an hour-long train ride to get to the rocket-port? or that it's completely impossible to make a rocket reliable enough to routinely carry paying passengers on?
these are implementation details, and the idea that they make the whole idea an "obvious non-starter" is risible.
When someone is off orders of magnitude, is not “an implementation detail”, to say that is to fail at basic engineering. Moving something from ~1% failure rate to ~0.01% is a gigantic task, because you’re into heavy diminishing returns zone. And that’s assuming it can be done, which it might not.
By that logic making hypersonic planes is just “an implementation detail”...
There’s LOTs more issues by the way:
- Rockets don’t take off on bad weather, but more importantly *they can’t do vertical landing on bad weather either*.
- Fueling a rocket takes hours, and it’s done just before launch for safety reasons. Changing this requires compensating with even better failure rates.
- The economics barely work out even if you *could* reuse the rocket several times a day, AFAIK SpaceX quickest turnaround for reuse is still measured in weeks.
- Depending on location, a suitable launchpad/landing pad is far enough that lots of routes are pointless (optimistically: 10m boarding boat + 2h boat + 10m boarding rocket + 1h + 10m + 2h second boat= 5h30m i.e. NY to LA).
- A lot of people would simply be unable to handle the g load, reducing the g load necessarily increases flight time heavily.
- At least 10x the emissions/passenger trip as a regular flight.
I could keep going. This is just a *bad* idea, there’s simply no way around it, and wishing the end result (travel anywhere in 1h) does nothing to change the physics of it.
The concept art has launch pads built some distance offshore and reachable by ferry https://www.youtube.com/watch?v=zqE-ultsWt0
Has anyone done the maths on how loud it would actually be on shore at that estimated distance, and how that compares to, say, a helicopter flying overhead?
"It has worked on a much smaller scale, for a much shorter distance, under experimental conditions, with almost non-existant concerns for safety, economic feasibility, or public and environmental impact"
is an extremely weak rebuttal in general, and in this case in particular. If you really are an "engineer" (as opposed to, say, a mathematician or a theoretical physicist), you should already know that. It's like claiming that economically viable fusion power is easy because I can build a fusor in my garage.
The claim being rebutted is "he okayed the idea of point to point rocket travel, which is so full of holes that IMO it puts into question how much he knows". All that's needed to rebut that is a simple existence proof, which Delta Clipper Experimental provided before some of the people in this thread were born.
Would point to point rocket travel be economically viable for passengers? Dunno. It could wind up like the Concorde, technically feasible but economically a failure. But that's a question requiring detailed analysis and probably building prototypes. Almost certainly an intercontinental passenger transport would be a different design than a Mars craft.
Would there be lots of detail to work out? Oh, Hell yeah. Would the ride be rough enough to filter out some passengers and cargo? Depends on the vehicle design.
Would some adventurers want to ride it even if it cost ten times what a 747 does? Yeah, but not enough to pay for it. That's why you do business case analysis--something I have done as an engineer.
> All that's needed to rebut that is a simple existence proof, which Delta Clipper Experimental provided before some of the people in this thread were born.
No, it didn't, because in the context of this thread, "point to point rocket travel" clearly doesn't refer to a short hop of a couple 100m by an experienced test pilot in a vehicle with room for exactly zero passengers, but to Musk's plans (or concept art?) for commercial, airliner-like transport of regular people. Nobody doubts that SpaceX can launch and land rockets vertically – but that's not what Xavi meant by "idea which is full of holes". The DC-X tests only demonstrated that rockets can launch and land vertically, but that's only a miniscule part of what's needed to show that the idea isn't "full of holes".
To add to that, the space shuttle regularly landed at a different place to where it launched, and could in principle have been used for point to point transport. There are major logistical issues, however, as well as national security issues (I doubt Starship will ever be allowed anywhere near China, for example). Personally I doubt Starship will be used for point-to-point transport, but its not hard to imagine that something else could be.
Again, it’s not about “what”, but “how”.
I’m perfectly willing to concede you can design a craft for point to point suborbital transfer, but it’s clear that that vehicle isn’t Starship.
To propose Starship is a viable craft for that purpose is IMO bonkers. Even if SpaceX can iron out the details, Starship will by design be very different from what you want from an “space airliner”, which it’s not surprising if it was designed as a craft to get heavy payloads to Mars.
Yeah, how can you prove that your world-crossing rocket isn't a nuclear-armed ICBM, *especially* with a ballistic flight profile like this ?
"but I’ve not seen a single instance where he showed “knowing rockets” past what I’d expect for the average KSP player."
Where have you been looking? Because anyone talking about rockets to the general public, is going to cap their exposition at roughly the KSP/Scott Manly level if they want to actually communicate. KSP may be a silly-looking bit of entertainment, but it actually encodes rocketry knowledge at about the highest level an audience without engineering degrees can handle.
As someone with a couple of aerospace engineering degrees, Elon knows his stuff. He's not in the top tier of rocket scientists, but he's good enough to work with and effectively manage the top tier.
I’m not expecting him to start trying to explain real orbital mechanics, but I’d expect him to have insights that require good knowledge. I’ve found 0 evidence of that, including his SpaceX presentations.
I’m not saying “KSP is bad”, I’ve spent my fair share of time on it. I’m saying “if that’s your bar for genius level rocketry, your bar is far too low”.
Again, all I get is people who say “he knows his stuff”, but I’m still waiting for actual *evidence* of that claim, and so far the evidence I got says he doesn’t fundamentally understand a lot of the physics involved.
You've moved the bar from your claim that for experts "it’s hard to not see he is talking BS". An expert just told you otherwise. Even if Musk really doesn't know his stuff, this still defeats that claim. Clearly it's not such obvious BS to others who know this stuff.
I think your reading of my point is a particularly uncharitable one. You’re translating “I see experts claim Musk points are BS” to “all experts disagree with Musk”, thus one counter example is enough to disprove the point. That’s a ludicrous bar to pass, there’s still experts who e.g. think cold fusion is possible.
In my view, if someone is claimed to be a genius, I’d expect experts to disagree on the margins, but I’d also expect a broad agreement of the basics. That’s not what I see with Musk.
I see some experts agreeing with Musk, but then I see vocal critics who say he doesn’t understand something fundamental about their particular area of expertise. As an example:
- Pesenti (among others) on AI. Some are more diplomatic than others.
- Nicolelis on how the brain works.
- Theo on how internet advertising works.
- Bryan Cantrill on how software development works.
I am sorry for the tone of my last comment, which I think was overly confrontational.
Still, the book review and the comment section of full of quotes from people with expertise saying Musk does seem very knowledgeable in their field. Some more examples here (some overlap with the book review): https://www.reddit.com/r/SpaceXLounge/comments/k1e0ta/evidence_that_musk_is_the_chief_engineer_of_spacex/
I'm not assuming that you think every expert disagrees with Musk. I'm just objecting to the idea that there's Gell-Mann amnesia for Musk or that it's obvious on average to experts that Musk is talking BS.
I'm not sure what all the people in your examples have said about Musk, but Pesenti was objecting to Musk saying in a Lex Fridman interview, "I think we're missing a few key ideas for artificial general intelligence. But it's gonna be upon us very quickly"
1. Pesenti objects to Musk even using the term AGI, which he says isn't a thing. But it was Lex Fridman, an AI researcher, who asked Musk if he thinks AGI would will come from the current approaches or if some new ideas need to be invented. Musk was just answering the question as it was posed.
2. Pesenti is confident human-level AI won't happen within his lifetime. This is itself a controversial opinion among AI researchers. And Musk only vaguely says he thinks general AI will come soon, but he gives no indication of whether he thinks soon is within 10 years or more like 40 years. There's nothing objectionable about Musk's statement here and plenty of AI experts have said similar things.
3. Pesenti is upset that Musk is concerned about the dangers of an AGI taking over. A lot of AI researchers are concerned about this.
So I don't think Pesenti counts. He's jumping on a vague answer Musk gave in a casual podcast that plenty of AI experts would agree with.
It would be genuinely bizarre if he didn't know about rockets at this point though, right?
I did not understand why this quote was a big deal:
"When I met Elon it was apparent to me that although he had a scientific mind and he understood scientific principles, he did not know anything about rockets. Nothing. That was in 2001. By 2007 he knew everything about rockets"
That is literally **6 years** of extremely devoted and focused time. How could an intelligent person NOT become incredibly knowledgeable in that time period?
Because it was also the early years of Tesla and SolarCity ?
Elon at Tesla was 2004.
SolarCity was 2006.
He started on rockets in 2001 based on the quote. Again, this is not really noteworthy. An intelligent nerd knew many things about their domain of interest after years of hyper focus? Common.
My point is that also working on (at least) two other businesses invalidates your "**6 years** of extremely devoted and focused time" bit.
So 3 years + 3 more years of less focused time? "Invalidated" seems strong here. I think the essence of my point stands obviously.
Are you aware that they have slowly been progressing towards a "single big piece"? They already apparently have some of the largest presses in the industry, and are soon to be rolling out even larger ones that will essentially form the entire body from a single piece instead of the 2-3 it is currently made of.
https://www.theverge.com/2023/9/14/23873345/tesla-gigapress-gigacasting-manufacturing-breakthrough
So we should give Tesla a pass for knowingly making dangerous cars because they’re now doing better?
Even if (and it’s a big if) other carmakers are being excessively cautious, I think it’s bonkers to consider someone a genius because he’s aggressively cost cutting by not caring about safety while designing a car.
Teslas routinely ace road-safety tests. A welded frame may well be less safe, comparatively, than a single-piece forged one. But it's obvious that early-generation Teslas were not actually "unsafe" on any objective scale.
*Some* Teslas have aced *some* tests.
They also have a track record of reporting false results of their safety ratings (and getting called on that) covering flaws by having customers sign NDAs, and there’s leaked code that it’s a strong indicator that they have “special modes” for safety testing.
I am all ears to hear how a special mode for safety testing could influence the results of crash tests over the automobile frame.
Do you have any evidence whatsoever to actually suggest that Teslas are unsafe? Or is it all just insinuation and speculation?
Just off the top of my head, since the setup of crash tests is known in advance, you could adjust the airbag deployment. You could also tweak the auto braking to be way more sensitive, because test setups don't have floating debris or incoming sun rays that will cause phantom braking. Once you know the details of a situation, there's lots of stuff to optimize around.
There's lots of evidence, but of course looking for "Tesla" and "Safety" pulls out Tesla's press releases and articles that suspiciously tout the info in the press release. For starters, the obligatory wiki link:
https://en.wikipedia.org/wiki/Criticism_of_Tesla,_Inc.#Safety_issues
Here you can find info on troubling behaviors like the stealth recalls, the misleading safety ratings, etc.. There's also some of the notable Tesla issues. Note that this issues weren't addressed by Tesla through active monitoring or feedback response, the recalls were either safety regulatory body initiated or initiated in response to a lawsuit.
Just couting NTHSA recalls, Tesla apparently takes the cake:
https://www.autoweek.com/news/industry-news/a43625242/tesla-is-the-most-recalled-car-brand/
And that's with less models and less cars on the road that most of the other brands.
Finally, while the data is now old, this is informative of a lot of my views regarding safety testing in the automotive industry (I've worked in safety related electronics):
https://danluu.com/car-safety/
Briefly: instead of focusing on a particular rating, you look at how well existing cars test on *new tests* (tests that did not exist when the car was designed, in this case driver-side small overlap). The idea of this kind of study is to see if the automaker was optimizing for safety or optimizing for good ratings.
My conclusion with all of this data, is that Tesla wants high safety ratings, but as a company it puts safety far back in the priorities list. Their approach to safety sounds similar to the Web approach to development, ship fast, let the users find and report the bugs, fix what you can through updates. The stakes are a bit higher with a car than a 404 page though.
This perception also meshes well with other data points, like the high number of OSHA related complaints Tesla has, and the concerning number of accidents on the factory floor.
Safety of EVs has more to do with having a lower center of gravity, more weight from battery backs, and that battery packs add to the strucural integrity of the car. I haven't seen the forged frame touted as beneficial for safety, rather that it's theoretically cheaper to manufacture. The downside is that a single-piece frame can lead to higher rates of unrepairable structural damages for minor-moderate collisions.
All auto manufacturing has moved more towards having vehicles crumple to improve safety over the last 10-20 years, such that anything above a minor collision causing frame damage essentially totals a car. There's a reason Toyota is planning on doing large-piece casting as well: https://www.reuters.com/technology/why-are-other-automakers-chasing-teslas-gigacasting-2023-06-14/
A) Please provide evidence both that Teslas are significantly unsafe and that this is due to the frame compromises made by choosing aluminum and
B) this proves that Tesla (and likely Musk) knew exactly the thing you claimed they didn't know. I was showing you that your example of a supposed lack of basic knowledge was simply wrong.
They may have chosen a tradeoff that you think is a bad one, but they clearly knew about the welding difficulty and that a single piece would be preferable, which directly contradicts your Gell-Mann amnesia argument.
A would be very hard, even if Tesla was open about their safety incidents (which they are not, as I explained in another comment on the thread.
B requires me to show what Musk knew at that time which I'm not sure is even knowable..
What I can provide you is evidence of stuff like welds on structural elements that would be considered very deficient in any other manufacturers, and any mechanical engineer can tell you they'll compromise the results of any safety ratings the car might have. Safety ratings are measured in part by crashing actual units, which means the results are only valid as long as the manufactured units and the tested units are similar enough, i.e. there's a strict quality control that ensures that the cars rolling off the plant are almost identical to the ones tested.
Terrible welds on Model S structural elements (Spanish, the mechanic in question is repeatedly asking Tesla if the part really complies with their QA, and wants written confirmation before installing it):
https://www.youtube.com/watch?v=3dZq1ZOWHSk
Image of the full frame:
https://www.youtube.com/watch?v=714BjeAsaIg&t=0s
This video also shows another failure caused by the aluminium alloy, the mechanic explains that driving the car over a regular bump that the customer has in their access road has damaged the frame. One would expect the screw or the front wishbone to get damaged first. This kind of damage can easily impact stability on the highway. It also explains that as a consequence from that extra slack, the screws were continuously hitting the battery.
The aluminium frame has other implications, the main one being that the variability on the frame seems to be the underlying cause of misalignment issues. Some of these misalignments are catastrophic in terms of safety:
https://www.youtube.com/watch?v=HJ1tofzX6Bk
In this video, a misaligned rear door is rendered permanently closed, even in the case of emergency.
There's also a LOT of videos, articles, etc... showing quality issues, including this very funny (but still concerning IMO) finding from a Model Y owner:
https://www.extremetech.com/cars/314871-tesla-model-y-owners-confused-to-find-wood-bracing-car-components
Gotta admit I’m really surprised by the number of people here who have a success metric other than “but does it work?”
It’s like when you beat a guy who is a Karate expert by using Jiu Jitsu and he explains how everything you did was invalid.
The question is does it work at what cost. The case against Musk isn't that he has acheived nothing but that he's acheived things by a combination of luck, throwing money at the problem and marketing, not any particular organizational genius
if you could completely revolutionize orbital-lift rockets by "marketing and throwing money at the problem", Boeing et al. would have done that a long time ago.
Musk would belong on any list of genius CEOs just for SpaceX alone, leaving aside everything else.
"Luck, throwing money at the problem and marketing" could plausibly make someone rich, but I don't think it's plausible that someone could become *as rich as Musk is* without some serious talent.
Not saying you’re stating that argument but the next logical question would be “… did it work more than once?”
There are Hard problems that no one currently has a solution to. Musk has solved *more than one* of these Hard problems (rockets and electric cars), helped solve electronic transactions (which isn't a Hard problem, but still impressive), is casually working on tunnel efficiency with the Boring company, and is working on Twitter.
Since no one solved electric cars or rockets before, despite many billions of dollars on both, it appears that Musk may be able to identify real solutions to problems that other very smart and very capable people failed to solve. It seems obvious that solving these kinds of problems necessarily involves doing things differently than previously tried. That Musk has tried different things that have been successful is very strong evidence in favor of his ability to do it, rather than evidence that he managed to luck into solving multiple problems that tens of thousands of people spent billions of dollars trying to solve but didn't.
This is irrelevant to the discussion, but I wonder about the derivation of IQ from SAT scores. Is it really as simple as saying, well he's probably 1 in a 1,000 good at the SAT, so he's probably at least that intelligent?
You're right, I messed up here. Here's Emil Kierkegaard (from https://emilkirkegaard.dk/en/2022/04/iqs-by-university-degrees/)
"The correlation between SAT and IQ in a representative sample is about 0.80 if there were no measurement error and no restriction of range. The fact that this correlation is not 1.00 means is that you cannot take 99th centile SAT and convert to 99th centile IQ. You have to apply some regression towards the mean. You have to use the z scores for this as centiles are non-linear transformations. 99th centile SAT in the table is about 1450 (rounded). If we compute it using the mean of 1010 and SD of 200, we get 1475, which is pretty close. In z scores, this is 2.33. If we want to get the IQ for this, it is 1.86 (2.33 * 0.8), which is 128. In the same fashion, a perfect SAT of 1600 is 99.8th centile, and about 2.95 z, which is then 2.36 z on IQ, which is about 135. This is a lot less impressive than sometimes thought and presented on various websites because these websites implicitly assume a correlation of 1.00 between IQ and SAT scales."
When you correct for this it lowers my estimate of his IQ about 5 points, which I've edited into the post.
What's the correlation between one IQ score test and another IQ score test for the same person using different questions on a different day? I'd be surprised if it were much greater than 0.8.
Two old studies:
https://onlinelibrary.wiley.com/doi/abs/10.1002/1097-4679(197904)35:2%3C352::AID-JCLP2270350226%3E3.0.CO;2-2
"76 male college students on two occasions with a retest interval of either 1 week, 1 month, 2 months, or 4 months"
".94 to .74 for the Full Scale IQ"
https://psycnet.apa.org/record/1979-27632-001
"50 psychiatric patients (mean age 44 yrs)"
"The interval between test and retest averaged almost 2 yrs. All test–retest correlations were .90 or over."
Even scores on an IQ test would be expected to regress towards the mean on a different IQ test though, so that's not unique to the SAT.
It seems like the fair thing to do is either regress all test scores based on their correlations with other IQ tests/g-loading, whether they're officially IQ tests or not, which would confusingly mean no one's "IQ" is actually what they scored on an IQ test.
Or the simpler option of treating scores on all decent tests as more or less equivalent using centiles/z scores as you originally did.
In the case of Elon we have a strong prior that he's not of average intelligence so it doesn't make much sense to regress him towards the mean of the general public. Perhaps we should even expect him to do better on an IQ test battery with spatial tests.
I'm not sure how to think about this; maybe @emilkirkegaard or @cremieuxrecueil can weigh in. My guess would be that all IQ tests are correlated about 0.9 with each other and g, but that IQ = score on an IQ test and so it's by definition correct, and the SAT is correlated more like 0.8 and is not exactly an IQ test so it's not by definition equivalent to your IQ. I could be totally wrong about that though.
I think the right way to perform the Bayesian update is to treat the IQ/SAT miscorrelation as noise, correct it by regressing it, and then once we have an effective IQ score, combine it (at that stage) with our prior about how smart he is. Otherwise we risk double-counting.
"all IQ tests are correlated about 0.9 with each other and g"
The big batteries maybe.
"IQ = score on an IQ test and so it's by definition correct"
By definition yes, but it's still slightly misleading(and what could be worse than that?) for SAT scores to be subjected to regressions that IQ tests are spared from simply because they're called IQ tests. Hath not a SAT load?
But I guess it doesn't matter that much, even a simple "SAT score divided by 10" IQ conversion probably worked ok for the old SAT.
If you would always apply regression to the mean, wouldn't that equate to always lowering the IQ measurement? I'm confused why this number should always go down, rather than potentially up. If it only correlates at 0.80, that doesn't imply that it's always higher, only that they're not the same thing.
Am I missing something here?
Below average scores would regress upwards.
So scores below 100 always move up and scores above 100 always go down? That still doesn't make sense to me. A mismeasurement can go up or down, and there's no reason to think that a mismeasurement of a certain kind can only go one way or another.
"a mismeasurement of a certain kind can only go one way or another"
No one has claimed that, it's about most likely expected value. You should assume a confidence interval exists for any estimate.
Twitter: If Musk's eventual goal is to transform the company to an Everything App, it's probably premature to judge the results. This would be someone buying a property to flip, then claiming it was a disaster when the house is stripped down to the studs. It might be too early to tell whether this will turn out.
"Musk was an early investor in Tesla because the founders approached him": This seems to argue the opposite what you did in your review. Sure, Tesla approached Musk, but having recently exited PayPal flush with cash, I'm sure lots of people in SV approached Musk. His reply to Tesla seems to have been, "sure, and I've been funding a guy to do cutting-edge battery research!" which doesn't seem like a 'right-place-right-time' situation, so much as 'fortune favors the prepared'.
I take your point that it's unfair to attribute a single motivation to something as complicated as a person, but it's probably also unfair to say someone like Musk is motivated by the whims of the moment. It's clear with many of his ventures they are things he thought about for a long time. The book makes it seem like he was interested in space and rockets, but didn't really believe he could build them until they talked to the Russians and he did some back-of-the-envelope calculations. In the same way, it seems like he was interested in electric cars, but didn't believe he could make them happen until Tesla approached him.
In both cases, once he perceived his vision was 'possible', he devoted large amounts of time to making them reality. You don't build Tesla or SpaceX on a whim.
"the claim (still not proven, but plausible) that Community Notes wasn’t really a Musk project"
This one seems pretty easily provable, since there's documentation and screenshots that show Birdwatch (as Community Notes was called) and how it worked long before the acquisition, e.g. https://blog.twitter.com/en_us/topics/product/2021/introducing-birdwatch-a-community-based-approach-to-misinformation
I can also say firsthand that I was a Birdwatch beta tester since ~mid 2021 and the feature worked more or less the same way that Community Notes works now. It's great, but it certainly wasn't Musk's project. Credit to him for doubling down on it, but not pioneering it.
By "doubling down on it", do you mean he didn't cancel it, or is there some way he improved it or made it happen more?
Hard to say from the outside, but from what I've observed:
* He's called out the feature a few times and drawn attention to it being important to Twitter's future
* When I started using the beta, the notes were actually only visible to other Birdwatch testers - they went visible to everyone right around the time of the acquisition (Oct '22), so presumably not Elon's direct doing but adjacent to it
* Anecdotally, it feels like it's gained more visibility in the UI and more people have been invited into the program as notes writers, which presumably he would have signed off on
"I’m not sure whether this means that everyone else is an idiot with a pointless bureaucracy fetish, or that only a few very special people like Musk can make the non-bureaucratic version work."
It's the latter, though I'm sure Musk has access to a unique (temporally speaking) pool of disaffected talent. One mismatch that marketeering types sometimes seem to have with the general populace is that, yeah, the market CAN produce things with amazing quality and superior efficiency, but there is often a risk of total failure where you get nothing. When you apply this dynamic to things people rely on and pay taxes for (or sometimes, just the latter), they get understandably leery. And the more a successful firm begins to be relied on, the more they start to stagnate.
> It's the latter, though I'm sure Musk has access to a unique (temporally speaking) pool of disaffected talent
Yeah, Musk seems to have pulled off a more extreme version of what a lot of startups in Silicon Valley try to do of selling (mostly young) people on a dream and a mission, that makes them willing to work long hours at low pay, or put in extraordinary effort, that they wouldn't at a company that said "come work 9-5 and make progressive technical improvements".
Whether you see that as exploitative or just good leadership probably depends on the case in question. Anyone who has worked in industries like this will have seen young enthusiastic people get burned out by believing in something that turned out to be just spin, and not getting compensated for the real value of what they were providing. But there are also all the goto examples of startups that have done well
> It is basically impossible to become a world-class software developer if you start after you’ve achieved career success in another industry.
Does anyone have thoughts on why this would be? I can’t think of a counter example to contradict this statement. But it doesn’t seem like there’s any fundamental reason why you couldn’t become world class starting a bit later. Is it just that it’s tough to put in the hours once you’re in your 30’s? You're more susceptible to carpal tunnel and neck pain and it just physically gets more difficult to put in the same number of hours.
Troll answer which might also be correct: https://www.astralcodexten.com/p/critical-periods-for-language-much
I’m not sure I agree with the premise to begin with, it’s tough to be a world class developer to begin with, so it’s not unexpected that it’s hard to find counter examples, but I’m pretty sure there are some.
The only thing I can think of as an explanation would be that it’s hard to “think like a computer” if you aren’t use to it, and it’s an essential skill, but I doubt it’s unlearnable past your 30s.
"World-class" is such a vague term that it's neither true nor false. Does it mean "capable of becoming a staff-level engineer at a FAANG" or does it mean "capable of being Don Knuth"?
The biggest problem is motivation; people who become software engineers later in life are usually the people who didn't really want to become software engineers, they just fell into it after their initial cooler career choice didn't work out. So they're going through the motions to earn a paycheque. If they had the personality and drive to become really amazing software engineers then they would have done that in the first place.
Knuth is a talented mathematician who likes algorithms. I'm not sure I would think of this as the standard of "world-class developer": it's special but in a slightly different direction.
I was 32 when Hofstadter's "Gödel, Escher, Bach" inspired me to learn coding. I was not a world class developer but had a pretty successful 30 year career.
This depends how you define "world class developer". If you define it as someone who's primarily on the software engineering side of things then the transition is rare because almost no one switches from being successful in one field to being an IC in another field. If you just mean someone who's good at programming required at his job, plenty of people have e.g. learned astrophysics and then become good at programming astrophysics simulators or whatever.
Having seen the unusably poor quality of some academic simulators, gonna have to quibble with that last claim. Academics usually seem to learn just enough programming to get the computer to spit out numbers that support their pre-existing beliefs, and then stop. Does the program actually work? Do what it claims to do? Run on anything other than the researcher's laptop? Not considered relevant.
Computing changes a lot faster than most other industries do, which means that you have to be continuously learning to a greater extent than in other fields. And that knowledge compounds. Yes, some stuff I learned 30 years ago is now obsolete, but a surprising amount isn't. The knowledge I'm learning today doesn't replace yesterday's, it augments or builds on it.
So if you don't start young, it's very hard to catch up. You can learn software development alright but you'll never be world class because all your time will be going into learning the new stuff, not the older stuff that's still important.
Good point. Hash tables will always be around. And people will still be fighting about functional programming versus OOP for decades too.
I'm not sure about the original author's intention but it seems plausible to me. Software requires an absurd degree of detail orientation and willingness to dive down a dozen abstractions, far away from the problem you're nominally trying to solve. There's so much detail in between processors, memory, operating systems, programming languages, system architectures, if you've achieved success in some other industry it does seem pretty unlikely you'll put in the work to master these concepts, all of which you kind of need to be considered a "world-class developer" imo.
On the Twitter/X user experience: I have a lot of right leaning friends in the indie author space. There's been more of those people coming back to the site. Those who stayed on report a better experience--more interaction, less shadowbanning, quicker fixes when the mods screw up. I suspect the opposite is happening in left-leaning twitter spaces, and the site is staying even on users on net.
That would only work if there is an equal number of right and left leaning users to draw on. Since the most "online" demographics tend to be young educated urban people, who are more left leaning on average, I wouldn't expect that to be the case. Though it may be one of those things where what matters is actually a small number of highly engaged users, and Musk is able to make more money from the smaller pool of right wing whales.
Is there a plan to milk the whales, though? A hundred dollars a year isn't exactly whale territory.
I was scrolling up the comments from the bottom, and for a moment I thought your comment was about literally milking literal whales. :-)
There's probably way more "right leaning" people available given that left and right in pre-Musk twitter terms were heavily tilted towards the extreme left being defined as normal. Musk himself was like that, right? He tweeted once that famous meme where someone starts out being center-left and then stays still as others dash way further to the left, meaning he ends up on the right. These two camps aren't a 50/50 split by any means.
Also given that one of the chief complaints about Facebook is the preponderance of older people, it doesn't seem like lack of right wing people would be a problem, especially as in many parts of the world older and more conservative people are richer.
My impression is that normies tend to not have strong detailed political opinions, instead mostly going with the flow. Sure, if you ask them out of the blue, the lefty ones wouldn't articulate the hottest woke talking points, but neither would they be bothered when encountering those in the wild, being on the right side of history and so on.
After all that I’ve read about Mr Musk there is nothing I would buy, nothing I would say nor, anything that would make me give him any more of my time.
Musk is a Tigger. A personality type which leads Rabbits to become a bit too keen on Unbouncing him.
One day he will inevitably say that he knows how to climb trees and get stuck at the top of one with Roo on his back. It's something of a worry that with AI humanity could be playing the role of the baby kangaroo.
> I refuse to believe that going to Mars isn’t 100x more expensive than figuring out ways to solve these problems on Earth.
Geothermal-powered vertical farming (like done in Iceland) is already sufficient for this, I think.
From a quoted comment:
> It will use Tesla cars, driven by Tesla employees. In my view this is basically an underground Uber system, but it will probably have more expensive fares to regain the capital costs of building the tunnel (Boring Company is paying for the tunnels, and casinos are paying for the stations, they do NOT have funding from City of Las Vegas AFAIK). But this expensive Uber system is exciting??
It is known that the goal is not to have Tesla employees drive, but rather to use self-driving cars. This should be much easier than achieving general self-driving capabilities.
The capital costs of building the tunnels appear to be very low compared to those of subways in the U.S. I can see the system becoming profitable.
If anyone thinks that a public transport system with private vehicles is ridiculously inefficient, think of the number of times you've been on a whole damn train thundering down the line at less than ten percent of its passenger capacity. (For me it's a lot, anyway.)
Which is still *massively* more people per square foot than any highway.
Basic computation: a car with one person going at 30mph requires about 30 meters of buffer, plus 3 meters if the car. In terms of space usage that's worse than a single person per train car (or an entire New York subway train with six people in it). And that's for a road at full capacity *and also* ignoring roads being empty ever *and also* ignoring the massive space costs of last mile roads and parking.
If you don’t ignore roads being empty ever, then you also have to not ignore train tracks (and stations) being empty ever, and that sounds like a much harder comparison to put numbers to.
For that matter, if you count the required buffer for cars, should you not count the required buffer between entire trains, not just between the train cars? I *think* trains might still “win” that in some sense, but still, it would significantly reduce their density.
Similarly about last mile roads and parking. If you include that for cars, then you need to include that for trains as well, and the cost of “last mile train tracks” is in a sense infinite, because you can’t actually make train systems that go as close to your house as cars do.
(The closest train to my current position is actually about half a mile away, so it technically covers about half of the last mile, but my car is about 20 times closer and goes more places. And I live in a high density city, good luck getting even close to that ratio in smaller towns and villages.)
The efficiency of a transport system isn't best judged by its downtimes, but by its peak times. Or at least, the downtimes aren't the point of a rail system - the peak times are. (Roads are designed to work best at downtimes.)
"If neither company had outstanding stock and therefore they both had a market capitalization of zero, how likely would you be to conclude that owning Ford would be 17 times worse than owning Tesla?"
on the topic of relative valuations, the couple comments in the article above (haven't read through the full commentary stack, no time at moment) haven't mentioned the most important point: Ford is in the ICE car business (9x% of revenues, 100%+ of profits) which is likely to be completely defunct or no more than a small sliver of the overall new car market within 15-20 years. Tesla is in the EV business which is the future (and its EV business is actually profitable). Ford is not doing a great job moving towards EVs, and their upcoming UAW contract will likely ensure it never does. Don't you think based on that alone Tesla deserves a huge valuation premium? (btw. i don't hold either stock, have never been able to convince myself Tesla isn't overvalued, but Ford is also potentially hugely overvalued / a melting ice cube)
The back and forth here on whether Musk is "smart" reminds me of the same debates about Chat-GPT. Some people claim that it/he is confronted with novel (or at least, novel to them) difficult problems and quickly providing correct answers. Others claim they lack real understanding, and just parrot what someone else came up with, make a statement that is too general to be useful, or just give a response that is wrong (but maybe you need some expertise to see why).
The comment about Community Notes is accurate — started as Birdwatch Inn 2021 and was fairly well established before Musk took over
"Starlink's terms of service include a Mars clause: Users must agree that Mars is a free planet unbound by the authority or sovereignty of any Earth-bound government."
So who will govern Mars, then, and will the 'users' agree to be ruled by God-King Elon? What about if there are disputes over that, and people want to get appeals heard on Earth?
And of course, all an Earth government has to do in a dispute with Mars is shrug, say "okay, you don't want anything to do with us? have fun trying to be self-sufficient really fast when we're not sending any more supplies to you".
Who says the supplies would be coming from a government in the first place?
Earth governments could potentially block or impound supply shipments to Mars coming from the private sector, but as rockets can launch theoretically from anywhere and govs have poor track records of being able to impose reliable sanctions even on third world countries ... well.
"Who says the supplies would be coming from a government in the first place?"
Ah yes, because governments look so kindly on private companies exporting to regimes that are considered unfriendly.
The "Supergun" affair:
https://en.wikipedia.org/wiki/Supergun_affair
https://www.bbc.com/future/article/20160317-the-man-who-tried-to-make-a-supergun-for-saddam-hussein
If Mossad don't like you selling things you shouldn't, then they'll get you.
The very tangled and indeed nasty "Arms to Iraq" enquiry, where a company being prosecuted for selling arms to Iraq had the trial collapse because it was revealed the British government had been secretly helping them sell arms to Iraq:
https://www.theguardian.com/world/defence-and-security-blog/2012/nov/09/arms-iraq-saddam-hussein
And yet North Korea has MacBooks ... yeah you can imagine governements sanctioning Mars just because (the rationales used for sanctions on Earth wouldn't apply here, presumably). But that doesn't mean they can do it successfully.
Governments have been pretty good at making sure advanced rocket technology, infrastructure, and supply chains don't make it to third world countries. North Korea has put a significant fraction of its GDP into rocketry for decades, and can barely launch small satellites. Third-world countries building Mars transports is not going to be a thing for quite some time, and first-world governments can stop them from buying such vehicles from first-world contractors.
The "linear development" process that Fluffy Buffalo described is basically good for two things:
1. Spreading development across congressional districts to win support for the project, at least in theory - in practice, it matters a LOT more to have one well-positioned Senator or House member to drive a project forward. We have SLS in large part because Dick Shelby dominated in the Senate Appropriations Committee and make it a priority, and Europa Clipper because John Culberson went hard to bat for it in the House.
2. Avoiding a high profile failure, like a prototype exploding on the pad. Instead, you keep all the failures and explosions hidden inside smaller facilities far removed from the public eye, and (hopefully) by the time it reaches the launch pad it's near-certain to not fail (and if you have even an inkling that it might fail, you take it back from the pad).
Insightful. Thanks.
Re: Twitter/X, I can't remember the source now, but supposedly Musk decided to buy the platform because it seemed stupid to him that even though he was the richest man on earth, he still couldn't say what he wanted. So if X loses money but becomes less restricted in terms of censoring people (or, less charitably, censoring Elon Musk), maybe he'd still consider that a win.
On the Boring Company, it's worth pointing out that there are manufacturers of mass-produced TBMs - these are ones in the 30cm to 3m diameter range (yes, they really are TBMs; they're the same technology as the big ones). These are used for water and gas pipes and sewers when they are laid in locations where it's not workable to just dig a trench, drop the pipe in, and cover it over (most typically under rivers, but also under existing large buildings, under airport runways and probably some other cases I'm not thinking about).
One reason why the claim that he could make the tunnels cheaper by making them smaller produced quite a lot of doubt was just how little price benefit there is from the competitive high-volume market for small-diameter TBMs (large TBMs are usually Germany's Herrenknecht in Europe, Japan's OGITEC in East Asia or CREG in China, there are about 20 manufacturers once you get down to 3m diameter). The Boring Company isn't touting a technical breakthrough in TBM technology the way that, say, SpaceX has recoverable rockets.
If they were going into the big-tunnel space to compete with what is essentially a geographically segmented monopoly, then I'd believe that they could come up with organisational innovations to make TBMs cheaper (much as Falcon 9 was noticeably cheaper even before it was recoverable, because it was competing with a hyper-conservative oligopoly culture) but if they want to hit orders of magnitude, I think they will need a technological breakthrough, and there's no indication that they're doing anything genuinely novel.
"the claim (still not proven, but plausible) that Community Notes wasn’t really a Musk project"
The code for it is actually open source so you can just see the history here: https://github.com/twitter/communitynotes (it started in 2021). The code has changed and has been improved post acquisition but the core idea and algorithm of what makes Community Notes special is still very much not an Elon thing.
"Community Notes"/Birdwatch became more prevalent when he acquired it but that's simply a coincidence of the project fully deploying on the frontend side slightly before and during the acquisition process, thus why people think he was behind it. AFAICT his main contribution was renaming it to Community Notes as he wanted to move away from bird-themed puns
Missed in the discussion of market caps is that Tesla is much more vertically integrated than Ford. So a true comparison would add the value of Ford to the value of it's various suppliers.
One fun fact I learned about Musk was that his maternal grandfather emigrated to SA from Saskatchewan, because he was worried Canada was becoming too socialist. He was a pilot and in his spare time would fly around looking for the lost city of the Kalahari.
To save other people looking it up, the move from Canada to South Africa was in 1950.
Here is an abridged 2022 John Carmack interview about Elon. Carmack attributes Elon’s success to his level of commitment to go all-in on highly risky businesses where he could have gone bust.
https://www.youtube.com/watch?v=IQro0rkg2DE
I met him as I was starting to do Armadillo Aerospace. He came down with his right-hand propulsion guy and we talked about rockets and what can we do with this. It was specific things about how are our flight computers set up, what are different propellant options, what can happen with different ways of putting things together.
In some ways he was certainly the biggest player in the alt space community that was going on in the early 2000s. He was the most well-funded although his funding in the larger scheme of things compared to NASA was really tiny but it was a lot more than I had at the time.
But it was interesting, I had a point years later when I realized my financial resources at this point are basically what Elon's were when he went all-in on SpaceX and Tesla. I think in many corners he does not get the respect that he should about being a wealthy person that could just retire and [instead] he went all-in where he could have gone bust. There's plenty of people, you look at sad athletes or entertainers that had all the money in the world and blew it, and he could have been the business case example of that.
I have a great deal of admiration that he was willing to throw himself so completely into that, because in contrast with myself I was doing Armadillo Aerospace with this tightly bounded [limit], it was John's crazy money at the time that had a finite limit on it, it was never going to impact me or my family if it completely failed. I was still hedging my bets working at ID software at the time when he had been really all-in there and I have a huge amount of respect for that.
The other thing I get irritated with is people that say, “Oh Elon's just a business guy, he just was
gifted the money and he's just investing in all of this,” when he was really deeply involved in a lot of the decisions. Not all of them were perfect but he cared very much about engine material selection and propellant selection. For years he'd be telling me get off that hydrogen peroxide stuff, liquid oxygen is the only proper oxidizer for this.
The times that I've gone through the factories with him we're talking very detailed things about how this weld is made, how this sub assembly goes together, what are startup shutdown behaviors of the different things. He is really in there at a very detailed level and I think that he is the best modern example now of someone that can effectively micromanage some decisions on things on both Tesla and SpaceX to some degree where he cares enough about it.
I worry a lot that he’s stretched too thin with the Boring company and Neuralink and Twitter
and all the other possible things there. I know I’ve got limits on how much I can pay attention to and I have to box off different amounts of time.
I look back at my aerospace side of things, and it's like I did not go all-in on that. I did not commit myself at a level that it would have taken to be successful there. It's kind of a weird thing having a discussion while he's the richest man in the world right now. He operates on a level that is still very much in my wheelhouse on a technical side of things.
>> I understand the frustration... but my impression is that space exploration is one of the fields where very thorough, very systematic planning with very conservative change cycles is the most promising approach to get something that works at the first attempt - even if it takes longer and costs more than planned. Compare the JWST to the most recent "Starship" launch for illustration.
> This would sound plausible, except that Musk has succeeded by doing the opposite. I think this is why so many people are in love with Musk: he’s proven that valuing good ideas, moving fast, and not having bureaucracy can work, sort of, in a weird little bubble of his own creation. Yeah, the first Starship exploded, but most people predict Starship will eventually work, and when it does it will be a much more impressive feat of engineering than JWST or anything else created the “proper” way.
Weirdly, you seem to have misunderstood your own point of disagreement with Fluffy Buffalo. Musk hasn't succeeded at the only goal Fluffy Buffalo mentions - succeeding on the first try. You go on to claim that Musk is a counterexample to Fluffy Buffalo's ideas while conceding that he isn't. (?!)
The actual point here is that succeeding on the first try is not a valuable goal, not that planning isn't helpful in achieving it.
What isn't mentioned is how bureaucratic the organisations are that Musk's two most succesful companies compete with. Lack of good competition helps a lot. Explains his lack of success at Boring comopany, Twitter and that brain implant company.
Then what he did with Tesla is obviously impressive, but is it earth shattering? The company lost money for well over a decade. With negative margins of < -10% for most years while:
-Selling higher margin luxury cars.
-Underpaying and overworking employees.
-Delivering cars of subpar build quality.
-Providing poor after market service.
-Benefitting from billions in subsidies.
-Not even making the battery cells (the main tech) themselves
Other car companies looked at this and said to themselves, well we got unions to content with, and measures of quality control that are much higher, why on earth would we try to compete here and lose even more money?
Now that battery cells have come down enough in price and making EVs has become profitable, competition is flooding into the market.
Seems like Musks main skill here was keeping Tesla financially afloat and keeping the hype up, allowing him to sell a lot of stock and debt at high prices (exceptionally low interest rates helped a lot here as well). Being succesful at SpaceX was probably a major help to this.
Musk hasn't just beaten his competitors, he's beaten everyone else in Silicon Valley and around the world who could have tried putting themselves in a position to beat those same competitors. IE if it was easy, why didn't anyone else do it?
(cf Andrew Beal, Better Place, and all the other startups that tried to disrupt cars and rockets and failed)
Not the guy you replied to, and I haven't read all the comments and shit, but I've always seen Musk as a guy who has one particular superpower, which is an ability to raise practically infinite capital.
Even for twitter, which never really had a model for being profitable and he loaded up with debt to the point that it will definitely lose money, and which Musk has admitted he overpaid for and very clearly tried to back out of buying, he got a bunch of rich people to put in money.
Like, imagine you agree to buy an apartment building for way above market price, then very loudly try to wiggle out of it for months, then try to get other people to go in for a share, telling them how good a deal it is. Would anyone invest? Well if you were Elon Musk they would! Dude could sell people a 1% share of a bag containing $100, for $5.
If you and Musk both presented a plan for some new business model - producing electric planes or something - you'd have a really detailed proposal, a shitload of tech specs, a SWAT team of PhDs to nerdily explain all the nuances, and Musk would show up like "we're gonna cut costs 98% by hiring the Cat in the Hat! Here's a picture I drew in crayon" and he'd still raise like a trillion dollars to your 0.
And then they'd hemorrhage money for 15 years, crash 47 test planes, repeatedly break their own promises, etc, but at the end of the day there'd be a working electric plane.
The really dramatic/self-important description of the world of finance is that it's how society decides how to allocate resources to different endeavors.
So ... is it useful to allocate a shitload of resources to "whatever Elon Musk wants to do"?
Argument in favor: yes, look at this stuff he's done.
Argument against (assuming the stuff he's done is good): anyone who gets that much capital could do a lot of stuff. If you, Scott, got as much capital as Musk did, you'd probably totally transform how psychiatry works in the US - you'd get everything on your most fanciful wish list of "how I wish psychiatrists and adjacent people operated" and then some.
Argument in favor (assuming the above is true) it's still useful to have someone with the "raise infinite capital" cheat code on. The regular business world is so mired down in bureaucratic nonsense and conformism that randomly elevating one guy with slightly above replacement level ideas, to pursue those ideas, is still better than whatever that capital would otherwise be allocated to.
Perserverance. Since founded Tesla lost money with sizable negative profit margins for 16 years. And for 15 since Musk invested in it. And this is while running Tesla almost as a cult. Which probably lowered costs quite a bit. Which is not a luxury competitors had. (which is why I think also running SpaceX at the same time had a major synergistic effect here, helping Tesla survive).
Typically start ups start making a profit after 6-7 years and are shut down if they can't make it. For example Amazon lost money for 7 years. Google for 3-4 years, Facebook for 5 years.
And I don't think Tesla has beaten all competition. A lot of good EV's with long ranges are coming out now, that often have better build quality than Tesla. Now that it has become profitable to actually sell EV's. I have some friends who were in the market for an EV and they did not even seriously consider a Tesla (and not primarily because of Musk's politics).
good comment, and the first i've seen that touches on the "cult" aspect of Tesla. Tesla's retail investors are notoriously rabid and worship the ground Elon walks on. They provided a backstop to Tesla's market cap even when the company was generating losses quarter after quarter. They would 100% buy any promise Elon pushed and support it unconditionally, despite reality differing from the goal. Lots of examples here, but the most damning is Tesla's full self driving, which is going on like 8 years of 3 months maybe, 6 months definitely.
By far the largest reach of the Tesla influences is X, which I would argue controlling the narrtive of Tesla has existentially import. (it's possible Tesla has become large enough from a market cap perspective that only a "key person" risk event would cause a collapse)
BTW I am not arguing here that making Tesla a success was trivial. Just that it should be seen within context. I am pushing back against the "he did something that nobody thought was possible" narrative. The competition did think it was possible to make electric cars, just that it would not be profitable. So they waited until it did. As is happening now.
Although arguably Tesla did cause making EV's to become profitable sooner by probably a couple of years due to economies of scale in battery cell manufacturing.
Musk's unofficial marketing arm (with X being the platform with the largest reach/influence) is by far the most impactful driver of Tesla's ability to stay afload (via high market cap) for so many unprofitable years. How many company customer bases report best-in-class owner satisfaction with below average quaity and worst-in-class reliability? Musk's fanboys reached a critical mass and were able to keep Tesla's stock price elevated through the existentially precarious periods, ensuring Tesla could always raise more capital.
Also the birth of hedge fund manager / EV evangelist CNBC promoters (think Cathy Wood or Chamath) finding it easier to get an audience preaching faith in a company's future rather than any kind of logical or coherent argument based on typical company success metrics. Tesla is the posterchild for this sort of marketing
My view is the vast majority of discussion from the original book review to the following related discussion misses for the forest for the trees. Musk is almost entirely a story of access to retail capital providing a longer runway than basically any other founders. Tesla success means he can tap capital markets whenever he wants/needs based on his reputation of "never bet against musk" etc
The ledger of successes/failures or smart/dumb things Musk has done we can argue about, but the main story is Tesla. Without it nothing else is really possible.
Success/Smart
* UPenn and let's assume he was accepted to Stanford's graduate Physics program
* Zip2 - for sure benefitted from macro timing, but still he built something from nothing and got the payout from Compaq
* x.com - same comment as Zip2
* early investor in Tesla - sorta seems like a no-brainer since the founders already had a quick and sporty prototype, but EVs were seen as a losing bet at the time so this was definitely a source of Elon alpha
* SpaceX / Starlink and learning enough about rockets to be additive from an engineering perspective. Neither of these are novel ideas
* ability to consistently receive subsidies for all of his business ventures
Failures/Dumb
* Paypal/X.com period where he nearly killed the company by refusing to implement any compliance measure (chargebacks alone were totaling $2,300/hr as a result) and pushing an incentive program that the company could afford. Elon was ousted from CEO role and replaced with Thiel as a result
* SolarCity was quickly approaching bankruptcy in 2016 so Musk offered to merge SolarCity and Tesla, which was pretty clearly a musk bailout at the expense of Tesla shareholders. Very beneficial to the narrative of Musk to not have SC bankrupt
* Twitter - clownshow as has been well documented
* publicly committing securities fraud by faking a private buyout (taking Tesla private at $420)
* consistently missing driving automation timelines
* Neuralink / Boring Company - time sinks that are possible because of unlimited access to capital. Nothing these companies are doing is novel or profitable
* calling a cave rescue lead a pedo
* saying chess is a "too simple" after his feelings got hurt from a disagreement with Gary Kasparov over Ukraine
* using Tesla resources to build a home in Austin (ongoing criminal investigation)
Don't forget that he *also* broke security laws when acquiring his initial stake in Twitter in early 2022. (He filed the disclosure forms well after the required deadline and when he did finally disclose, he claimed on the forms that he had no intention of acquiring Twitter, immediately before he proceeded to do just that.)
`Lost money for a long time' would also seem to describe Amazon, or Facebook. Maybe even Google, although the dawn of Google is before I was an adult and paying attention.
Amazon already had positive operating margins in 2002. While being founded in July 1994. So 7 years of losses, give or take. Google was profitable almost right from the start.
Facebook was unprofitable for 5 years after founding.
Tesla was founded in 2003, Musk invested in it in 2004. They became profitable in 2019. So 15 years of losses give or take.
> AI experts, is this a big deal? Can other AI teams not access Twitter data through the public web? Is it a substantial amount of text compared to other corpuses? Is the structure (280-character blurbs written by morons) a limiting factor? Or is this a genuine treasure?
Not really an expert, but... at least before the API changes (and possibly still afterwards), Reddit comments were free and less-restricted than tweets. Their upvotes occasionally even correlate(d) with truthfulness/usefulness (see: why we google "best [thing] to buy reddit" instead of "best [thing] to buy").
I'm going with "not treasure".
Also potentially relevant: Twitter is no longer really accessible through the public web since Musk made it impossible to read tweets without an account, and I think at the same time also limited the amount a single account is allowed to access per day?
Weighing in on Twitter:
1) I've had to block FAR more people to keep my feed as consistent as it used to be, somewhere on the order of fifty a month when the number was previously less than ten. Part of it's the algorithm seemingly getting shittier. Another part of it is the top of every thread filling up with blue checks, who mysteriously tend to post the most asinine corporate boot-licking takes. A Twitter subscription service is not a BAD idea - in fact, this was in the works before Elon took over, to allow people to edit their posts, as far as I know. But one that allows people to boost their engagement just creates a bad incentive structure where the people dumb enough to piss away eight dollars on a free platform are promoted to the top. The best blue checks are people who bought it for the quality of life features, or the cynical businessmen looking to get an edge - both of which are in the minority. It's like Linkedin-lite, and Linkedin is the worst social media platform.
2) I get at least one cryptocurrency spam message a day under Elon's Twitter, that I didn't get before. At least half of the new accounts that follow me are bots with GPT generated bios and AI generated profile pictures. If his goal was to fix Twitter's bot problem, he's failed. He's partially not to blame for this - AI has expanded its capabilities far more than we expected in the last year - but at the same time, I don't see the same frequency of spam on any other social media platform.
3) There have been more major Twitter outages. This might just be anecdotal. But at the very least, the one where he pretended to have restricted the number of posts people can see per day to cover up a very loud and obvious engineering failure was entirely his fault. If you're trying to build up Twitter as something committed to free speech and transparency, it's not a good look to accidentally cripple it for close to 24 hours and then just lie and say you did it on purpose. If there are more incidents like this, more people will leave the platform, and because he lied about WHY it happened I haven't seen any evidence that he's taken steps to keep it from happening again other than reverting the changes.
There's more to it. Anecdotally, I feel like the environment has become more hostile. But if you want to boil down the mismanagement at Twitter and my current gripes with it to actual, quantifiable issues, that would be it. Twitter Blue cannot be tied to the promotion of tweets. If they want an organic platform and not a soulless collection of corporate robots like Linkedin, they need to decouple engagement from the amount of money you pay. They need now more than ever to have hundreds of people moderating the platform BY HAND, not with AI. I would not mind him laying off 90% of his engineers if it meant he brought in half as many moderators, but as it stands the platform is too vulnerable to spam. And if they don't want to lose users every time someone deploys shitty code and bricks the site for a day, they need to be actually transparent about what actually broke, instead of coming up with a stupid lie.
Max Levchin is CEO of affirm so don't think he is mostly retired anymore.
I remember reading somewhere that publicly diagnosing other (famous) people who aren't your patients is a violation of professional ethics. Is that not the case?
It's called the Goldwater rule (and primarily applies to political figures, though it's stated for any public figure). Presumably this heavily qualified "has some symptoms of X but not others, but I don't know what he doesn't share in public" is sufficiently careful for it.
Okay, and I suppose Musk opened the door on this a bit by discussing his mental health in public. Theyre not really diagnosing the person, they're diagnosing their online persona. It still doesn't sit well with me for some reason.
>I’m not sure whether this means that everyone else is an idiot with a pointless bureaucracy fetish, or that only a few very special people like Musk can make the non-bureaucratic version work.
You could ask Stockton Rush that question.
In other words, heavy survivorship bias going on here. We remember successes, not failures unless they're especially notable failures. Bureacracy is the institutional way of remembering failures.
That's very similar to my take. It's also how I think about government. To me, the best government is one lead by a singularly amazing leader who has significant power and control. That's also the worst government, depending on the person in control. And either one has a really hard time passing control to someone else without losing what makes it special or just totally falling apart.
"I’m not sure whether this means that everyone else is an idiot with a pointless bureaucracy fetish, or that only a few very special people like Musk can make the non-bureaucratic version work."
"Pointless" seems like going too far. Maybe everyone else is a lazy bureaucrat with a perfectly self-interested bureaucracy fetish!
Anecdotally I often parse bipolar but not bipolar as ADHD or something like it: tendency to excessive mood swings, extended "stuck on" or stuck off"", but in response to stuff exaggeratedly and spiralling, not actually random.
I don't care so much whether this is enough to be a medical condition or not, more whether "things you might expect to be under an adult's control but don't seem to be are harder to keep under control for the person, or whether the person doesn't care, or somewhere between". He seems to have SOME passionate intense work, and SOME giant pissy fits about things. But he doesn't seem to acknowledge that having angry meltdowns at people is a problem.
Re ability I expected him to be smart but not the best engineer at somewhere like SpaceX. I think that's probably still true. More like, all the OTHER things that make a driven leader good like breadth of knowledge, common sense, drive, vision, where is he on a spectrum between "good at things he knows the right amount about" and "petty and aggrandising at anything where he's even slightly out of his depth"
> This doesn’t even have to contradict Musk’s ADL claim! If Parag or whoever employed a thousand censors to keep the ADL happy, and the ADL becoming unhappy cuts Twitter profits by 60%, then there’s a strong business case for those censors!
I am less inclined to think of the ADL as causal than most others here, it seems like. I've worked with advertisers and they all said they were trained to be very careful of the brand presence, to the point where it looks panicy and puritanical to outsiders. I think if you took away the ADL specifically but left everything else the same, the world we see would not change. Here's why:
I used to work on the web site for a company that sold something bland and generically useful -- let's say refridgerators. We were beating the other refridgerator companies in our marketplaces, partially because our refridgerators were well made but mostly because our ad people were good at placing ads, targeting ads, and winning ad auctions without losing money. I did a lot of the SEO, and wrote a lot of the landing pages our ads pointed to, so I wound up talking to them a lot.
They were pretty chill to talk to, and clearly great their jobs, but in their work they were incredibly conservative. Any time our brand appeared next to anything even faintly controversial they went on the war path. Hate speech, sure, but also nudity, drugs, discussion of abortion; "anything that might make a suburban mom uncomfortable," as one of them summarized it once. If we placed an ad in a newspaper or something and it ran alongside a piece about a gang shooting, one of the ad folks would get on the phone and yell at the newspaper until they apologised, took it down, and refunded us.
It never made sense to me, but when I'd try to ask about it they'd mostly brush me off; this was how they were trained, it clearly worked, so I had to trust them. Finally, one night at the end of a company party I was drinking with the guy in charge of ads for north america, and I decided to rise the issue directly. "Why _don't_ we run ads against porn?" I asked. "After all, people who jerk off need refridgerators too!" "Yes," he patiently replied, "but then our brand is 'the refridgerator for jerk-offs'. We won't get enough sales to be worth that." Then he said a lot of stuff about brand awareness, but it wasn't phrased as a searing burn so I don't remember it precisely.
As far as I know the ADL never called us, but if they had our ad folks would have reacted right away. Not because they were scared of the ADL, but because they were scared of the same things the ADL is, plus a bunch more things besides. The ADL has probably organized some boycotts in their time, or tried to, but my guess is that most of their work is more like raising awareness, and the _modal_ ADL interaction is them calling a brand to say "Hey here's a picture of your ad running next to an ASCII swastika" and the brand replying "Thank you so much, we also care about that for self-interested profit reasons and will take action immediately," with no coercion even implied.
Now, separate from all that: are the advertisers actually right? Honestly I still don't believe it, and I could imagine a smart engineer seeing all this from the outside and assuming it was a lie meant to cover up something more sinister. But, at least in my corner of the industry, it wasn't.
> I thought there was a week or two when everyone threatened to switch to Mastodon, then found they didn’t like Mastodon and went back? So where did everyone go? Was it Mastodon after all? Facebook Threads? Blue Sky? Or did they all start learning to paint and spending time with their friends and families?
Personally I started spending more time bouldering and bicycling, and substituted Manifold Markets as my source of "what's going viral on Twitter right now" news.
In fact in some cases Manifold led to a better sense of what's actually going on in the world, such as in the room-temperature superconductor markets where people who read Twitter all day gave it way more credibility than was warranted, leading to a big profit opportunity.
I find it so interesting how resistant people seem to be to the idea that Elon Musk is very smart and capable and that has contributed positively to his success. It seems people are looking for any possible thing they can point to to say "see he's not that smart after all, he just got lucky". Maybe it's a response to the Musk-worshipping that was going on around 2018-2021, but it's a weird manifestation of a 'God of the Gaps' style argument.
Yes! Well-said! This is thing that bothered me about a number of comments too. "Look at his imperfect bathing average! Clearly a failure! Mentally ill too!" Etc.
It is because stuff like “pedo guy” or hyperloop seems so hard to reconcile with the smart Musk hypothesis.
I just don't see why the conclusion drawn from those incidents is "Musk must not be so smart despite all the evidence pointing to him being smart" rather than "It turns out very smart people are capable of doing things that seem very dumb". We have plenty of examples of the latter case being true. In fact in the case of Hyperloop, there must be plenty of very smart engineers/scientists who all did very well on their SATs and got into great grad school programs who devoted great efforts to working on Hyperloop
Yes. I should have accented “seems”.
and that he publically commited securities fraud by faking a private buyout of Tesla
My immediate reaction is "Why are we debating Musk's intelligence?" Yeah, you have a higher IQ than Musk. Who cares; have you built a successful rocket company? A little more subtly, everybody knows you can build a Starship-class rocket, and more or less how to do it; we did it fifty years ago and the Soviets came close. The tricky part is actually doing it.
The point that comes out that is unusual is that he has the ability to absorb a lot of detailed knowledge and work with it and using that to both impress and manage the herd of extremely good and highly motivated people he's recruited to do the work.
Weird to see Musk being compared to everyone including Fëanor, but not to John Galt (or any other Randian hero).
I did make a joke (not on that post) about all the non-defective Elon Musk clones going off into a Galt's Gulch retreat. It seemed like a convenient explanation for his energy, ubiquity, and erratic behavior. :-)
Every time I see a conversation about Mars colonization I ctrl-F for "perchlorate" and "soil" to see if they're serious. They rarely are.
I am well aware of the existence and implications of perchlorates in Martian soil, I have even put a fair degree of thought into how to put them to good use. But I don't find it necessary to mention them every time I talk about any aspect of Mars colonization, and I don't think that's a particularly good metric to determine the seriousness of a discussion.
If your model is that perchlorates make Mars colonization impractical without some new miracle, but the Space Nerds don't know about the perchlorates because they're just ignorant fanboys, then no. We know about the perchlorates, we don't think they make Mars colonization impractical, we've got reasons for that, but we've also got reasons to talk about a *whole lot* of things that aren't soil perchlorates.
Next time, try to be less insulting.
If you're going to make a hermetically sealed colony that has no interaction with the outside atmosphere or soil, then Antarctica is back on the table.
> AI experts, is this a big deal? Can other AI teams not access Twitter data through the public web? Is it a substantial amount of text compared to other corpuses? Is the structure (280-character blurbs written by morons) a limiting factor? Or is this a genuine treasure?
More data is basically always better in the LLM game, but having high quality clean curated factually grounded data is even better. See for example Textbooks Are All You Need: https://arxiv.org/abs/2306.11644
Twitter is basically the opposite of that.
The "python script.saga" seems to be two tweets. Im not really getting anything from.it.
With regards to the Vitamin C problem of building tunnels on earth- Rhubarb needs very little light to grow extraordinarily quickly. One cup of rhubarb is 16% DV Vitamin C. I think that the scurvy problem with Chicxulub 2 would be nil.
> For example, I’m surprised to hear this! I thought there was a week or two when everyone threatened to switch to Mastodon, then found they didn’t like Mastodon and went back? So where did everyone go? Was it Mastodon after all? Facebook Threads? Blue Sky? Or did they all start learning to paint and spending time with their friends and families?
If you want to see how bad Twitter can be now thanks to the prioritization of bluechecks, see this thread: https://twitter.com/cremieuxrecueil/status/1702132016831971365
Look how far down the replies you have to go before you see something besides explicitly anti-vaxxer gibberish. Blue check-dominance is atrociously bad for conversations that take place at IQs above room temperature and involve/allow many respondents.
Right now there's not a good alternative AFAICT, so mostly people are just spending less time there and moving to non-identical websites (e.g. Reddit, Facebook, Instagram).
You have an awful lot of mind-killing phrases there for someone pretending to be interested in conversation.
I am interested in conversation. I am not interested in conversing with people who think that getting vaccinated for COVID-19 is bad, on that subject, because that position is extremely, mindbogglingly stupid and based in 0 evidence. I don't care how many "mind killing phrases" are in that sentence, I also don't want the top fifty responses to "belief in spherical Earth correlates with IQ" or "belief that the 2016 and 2020 US Presidential elections were legitimate correlate with IQ" to be "that just proves that IQ doesn't make you smart." I cannot have a useful conversation with such people, because I would first have to spend 18 months slowly dredging their brains of whatever conspiratorial sludge has currently polluted them so that we could have an evidence-based conversation.
You made yourself quite clear the first time. Redundancy is unnecessary. Neither is your idiosyncratic definition of "useful."
Posting that was also unnecessary, but that didn't stop you.
1. Is a 1,700% increased death signal from VAERS "based in 0 evidence"? If the official vaccine safety monitoring system of the CDC and FDA doesn't count, what else would possibly constitute 'real' evidence? https://www.floridahealth.gov/newsroom/2023/02/20230215-updated-health-alert.pr.html
2. Is a 4 percent increase in the percent of the labor force that has some form of disability (the highest in 15 years, and the largest 1-year increase in disability on record), in the exact period during and immediately following the vaccine rollout, "based in 0 evidence"? https://fred.stlouisfed.org/series/LNU01374597
3. In the Pfizer study that led to the emergency approval of the vaccine, the vaccine group had around 4x as many adverse events and 7% more deaths than the control group (actually, 21% more deaths, but Pfizer explicitly excluded several of the dead vaxxers from their results sectiuon, as they mention in the study). Notably, 20% of the vaccine group deaths were due to heart attack, vs 0% in the control group. Is this "based in 0 evidence"? What else would possibly constitute evidence? https://www.ncbi.nlm.nih.gov/pmc/articles/PMC8461570/
I know enough to know that VAERS has fundamentally junk data gathering methods and can only be used as a direction of possible investigation, not as evidence in itself, so I'm definitely not interested in digging through your other two points to find if they are equally cherrypicked.
70% of the global population has gotten vaccinated and all you can do is point to this mush. If it was actually killing and maiming people in any appreciable amount doctors - in the generic sense, not like, one specific doctor - would have noticed by now. I know you don't trust doctors, but that's the thing: I don't want to argue up to "actually, the medical establishment that has overseen decades of rising life expectancy is not wildly incompetent enough to miss that something administered to seven out of every ten humans is actually really bad." That's not a good use of my time. I won't learn anything. Maybe I could persuade you but I don't care about you.
> VAERS has fundamentally junk data gathering methods and can only be used as a direction of possible investigation
You've just argued that *the primary safety monitoring system* for the vaccine has "fundamentally junk data gathering methods." And you think this bodes well for the safety of the vaccine?
> I'm definitely not interested in digging through your other two points to find if they are equally cherrypicked.
Not surprised, pretty typical behavior for your type. Calls the official safety monitoring system's data "cherrypicked," and then proceeds to cover their ears and yell "LALALALALALA I CAN'T HEAR YOU!!!" when presented with evidence that goes against their beliefs. And then somehow still thinks that their views are based in "evidence".
> 70% of the global population has gotten vaccinated and all you can do is point to this mush. If it was actually killing and maiming people in any appreciable amount doctors - in the generic sense, not like, one specific doctor - would have noticed by now.
There were more deaths in 2021 after the rollout of the vaccine than in 2020 without the vaccine. In the US alone, millions are dead. Insurance companies reported a vastly higher death rate (up to 2x higher than baseline) *specifically for young people* in 2021 immediately following the vaccine rollout, *not* among older people, as would be predicted if the 2021 deaths were primarily due to COVID itself.
> is not wildly incompetent enough to miss that something administered to seven out of every ten humans is actually really bad
No, incompetence alone doesn't explain it. For most, it's a mix of the general hyperconformity and fear of standing out from the crowd that all humans have, along with fear of losing your medical license or your career being ruined if you disagree with the mainstream narrative due to backslash from people like you whenever evidence contrary to your viewpoint is presented. And then of course there are the genuinely evil and malicious at the top who just wanted people to die (or didn't care how many died as long as they sold more shots), but they are not the majority.
Note that we have a strong precedent for people failing to speak up while atrocities are being committed: https://www.scientificamerican.com/article/the-nazi-and-the-psychiatrist/
Since you are exactly the kind of person that would have just gone along with those atrocities - "Oh, I'm sure ze authorities are just helping ze Jews, Helga, zey are good people. They can't be doing anything bad, surely one of them would have spoken up about zis if it were true!" - I don't care about you either. Be sure to schedule your 7th booster. And don't worry about that weird cramp in your leg or that irregular heart fluttering you get, I'm sure it's nothing. I'm just here to inform some of the more "Trust the Science!!!" types that these "No Evidence" claims are propagandistic drivel.
> It's worth going to Mars *because it's fucking going to Mars.*
Always relevant: https://www.smbc-comics.com/comic/2010-12-09
Does making Twitter a paid-only service (source: https://www.bbc.com/news/technology-66850821) destroy the "everything app" goal for X?
No, because if XTwitter is the "everything app", then everybody who doesn't want to be a nobody has to pay for it. And if it's on the order of $10/month, people who don't want to be nobodies will eventually suck it up, stop whining about how it used to be free, and send Elon their $10/month. Which might be enough to make him a trillionaire.
Assuming he can accomplish the "everything app" part, which is extremely dubious.
>This would sound plausible, except that Musk has succeeded by doing the opposite. I think this is why so many people are in love with Musk: he’s proven that valuing good ideas, moving fast, and not having bureaucracy can work, sort of, in a weird little bubble of his own creation.
There's a much more important structural issue to understand when discussing why SpaceX moves so fast while aerospace incumbents move soooooo slow:
Incumbents built the bulk of their engineering processes around the "waterfall" development model, where requirements definition, architecture, design, implementation, testing, deployment, and maintenance are all distinct phases, gated by exhaustive documentation and formal review processes before proceeding on to the next step. They did this because it was state-of-the-art engineering methodology back in the 60's and 70's, and it was how NASA and the DoD wanted it done. The waterfall model affects all of the incumbents' business processes. It's fair to say that Boeing, NorGrumm, and LockMart are essentially big machines designed to execute waterfall engineering, in a symbiotic relationship with the DoD and NASA, which are big machines for accepting waterfall engineering.
SpaceX, on the other hand, is an iterative company. How much of that is simply because it was founded when iterative and agile engineering processes were state of the art is unclear. With an iterative engineering process, you work really hard at pairing down requirements to the absolute minimum needed to make progress, you keep your architecture fairly fluid, and design / implementation / test / deployment are unified, short, continuous cycles.
SpaceX has been a huge success with NASA and the DoD (now mostly, but not completely, Space Force) because both organizations have realized that the waterfall model simply won't scale going forward, and they made reforms to purchasing that kinda looked like a "waterfall lite" model, which allowed companies like SpaceX to implement their own processes, with a smaller number of checkpoints to make it compatible with government acquisition regulations. This is why the commercial services programs that got spun up in the late oh-oh's were so important.
The results have been spectacular, and no better example exists than the comparison between the SpaceX F9/Dragon 2 crew system and the Boeing Starliner. D2's been pumping out 2-3 crewed missions for the last four years, while Starliner is going back into the shop for its third set of major problems and still hasn't launched a crew.
It's tempting to conclude that Boeing is simply incompetent with stuff like this, and that's possibly part of the problem. But the more important issue is that Boeing simply isn't set up to move like SpaceX is, and reforms will cost billions of dollars and take the entire organization offline while they revamp engineering management.
Iterative development is just plain better. SpaceX uses it because it was best practice when they spun up. Boeing can't use it because their engineering organization can't adapt to it.
This is interesting; do you have a link to more information about what these engineering methods are?
"Agile" is a common iterative software model. Iterative software is the reason that web software is now constantly updated, and that locally executed apps have largely transitioned to an auto-update model, where new features appear semi-magically.
Iterative hardware is more difficult. That SpaceX has succeeded in implementing an iterative methodology is an achievement not to be discounted.
PS: I realized I didn't really answer your question. Wikipedia has perfectly good descriptions. See "waterfall model", "iterative and incremental development", and "agile software development". Agile's a bit of a cult, albeit an extremely functional cult. I'm not sure you can apply the word "agile" to a hardware project, but you can certainly design hardware iteratively.
The original values from the 'agile manifesto' are (copied from Wikipedia):
Individuals and interactions over processes and tools
Working software over comprehensive documentation
Customer collaboration over contract negotiation
Responding to change over following a plan
Basically, work in small loops, often ~2 weeks, to get the smallest possible unit of work you can done and delivered rather than making small progress on a big plan and not delivering anything until the end. Delivery might mean delivered to the customer or just that it is merged into the main build (for software).
It's worth noting that there are a large number of very dogmatic branded implementations of agile such as Kanban, Scrum, Extreme Programming. Personally I don't believe strict adherence to any of them matters. The main value is that product management can find small bits of work, hand them off to engineers, and then not need to interfere for a couple weeks. At the end they can see how things look and easily adapt for the next cycle (or 'sprint').
I've worked for government contractors that use waterfall, startups that use agile, and large companies that say they use agile but are closer to waterfall. From a developer standpoint, agile means I spend more time accomplishing and less time making plans that will fall apart anyways.
Agile/iterative doesn't completely relieve you of the need to provide a reasonable architecture, but it does force you to architect stuff that's easy to rip out and modify when it (inevitably) turns out that what you thought were going to be the big issues aren't, and things you didn't think of at all turn out to be serious problems.
But there's no substitute for having a good guess as to what your product will grow up to be, and how your customers will use it. Products based on a bad guess fail. Products based on a good guess still morph beyond all recognition, but the kernel of the good guess can guide development through a lot of stuff that would otherwise kill the product.
I still haven't decided whether Musk's guess for Starship was good or bad. The architecture has been consistently driven to support high-scale missions to Mars and back, which I still find wildly improbable. But the things that requirement has forced on the project--ultra-heavy lift, on-orbit refueling, full reusability, extremely high launch cadence, and lots of earth-orbit-rendezvous--seem to coincidentally have built a system that's pretty good for cislunar and lunar surface applications, excellent for launching Starlinks (which hadn't even been imagined when Starship's architecture was first promulgated), and likely to enable a lot of other earth-orbit applications that nobody had imagined.
Manifestos aside, I'd say that the difference between Agile and Waterfall is organizational philosophy.
Under Waterfall, before performing any nitty-gritty work, your entire organization must produce a detailed designed document outlining every single component of your final product, and every single step on the way to getting there. The customer must sign off on this document before a single line of code is written (or a single bolt is tightened). Amending this document is possible, but very difficult (by design), requiring sign-offs by everyone involved. The advantage of the Waterfall model is predictability and accountability: the client knows exactly what he's getting, when he's getting it, and how it's going to perform.
Under Agile, teams iterate on short-term deliverables every couple weeks. The end product is understood in terms of vague generalities (i.e. "we want a word processor" or "we want to launch a satellite into orbit"), but these are attractors rather than firm commitments. The short-term deliverables *are* firm commitments; even though they could be changed, doing so without a good reason is discouraged. The advantage of Agile is rapid iteration. If you are developing some truly innovative product, then discovering mid-stream that your client asked for something physically impossible, nonsensical, or inconvenient for the users is not a problem: you can pivot toward something more reasonable over the next two weeks.
The disadvantage of Waterfall is lack of innovation. Innovation requires exploration, and exploration cannot be baked into the design document and chiseled into stone tablets once and for all.
The disadvantage of Agile is lack of accountability and lack of focus. Once you start exploring, it's possible to explore your way right off a cliff, and deliver some kind of a self-replicating food processor instead of a rocket.
There are ways to circumvent the disadvantages of both approaches, but ultimately you'd want to use Agile for innovation, and Waterfall for reliability.
>Iterative development is just plain better.
Iterative development greatly increases the chances that you will fail catastrophically because half your people didn't fully understand how you changed the design midstream. It has its advantages, and is better in some contexts, but it is not "just plain better".
Software and hardware have different answers to that issue. In software, the development and configuration control tools are so good that "refactoring" isn't nearly as big a deal as it once was. Pervasive use of object-based design has helped a lot as well.
Hardware is a different story. I'd say that substantial re-architecting of hardware systems is well-nigh to impossible, but fiddling with the pieces-parts inside a good architecture is pretty easy.
For all their internal complexity, the high-level stuff in a rocket architecture isn't that complex. You have to make very, very careful decisions up front, and then you get to live with those almost forever. But there's a lot of room for incremental improvements in subsystems.
There's also lots of room to wreck your rocket by making a small incremental "improvement" that isn't, because the person making the change doesn't understand all the interactions.
If it absolutely, positively has to work, and if you're flying people or billion-dollar payloads it kind of really has to work, then once you start actually *building* the rocket you go into a solid "any proposed improvements will wait on the Mark II rocket" mode. And you don't start building the Mark II until you get everybody in a room to agree on exactly what you're going to build. SpaceX learned to do this on the Falcon 9, about the time they started getting contracts to launch national security payloads on Falcon 9, and it's not a coincidence that's the one actually reliable thing they've built.
Not as reliable as ULA's rockets, but close enough for government work.
I think NASA required a stable design 5 flights before Demo-2, the first crewed test. That would have been mission #79.
The last loss-of-payload accident, the Amos-6 pad explosion, would have been mission #29. So SpaceX ran 51 missions with an unfrozen design, during which time they handled 13 NASA missions and 3 DoD missions. So at least the billion-dollar payload issue wasn't really a problem.
My understanding is that SpaceX is still iterating on Block 5, albeit much more carefully and incrementally, and with NASA and DoD sign-offs on changes.
As for reliability, that's complicated. If you count just Falcon vs. Atlas V, SpaceX is more reliable--even with the 2 failures and 1 partial failure. That's a result of the sample size being so much larger, with the correspondingly narrower confidence interval. If you count all Delta and Atlas launches under ULA management vs. all SpaceX operational launches, then ULA is slightly more reliable.
The Air Force was requiring design stability and Aerospace-grade mission assurance for GPS launches, which came several years before commercial crew. And ULA's 142 operational launch attempts isn't *that* much smaller than the Falcon 9's 258. Certainly not enough to compensate for two failures vs. zero.
I don't know what to tell you. NROL-76 launched on an F9 FT (NRO birds are expensive, but this one only went to VLEO, which may mean that it was a cheaper asset), and USAF launched DISCOVR on a v1.1 and an X-37B (not a cheap asset) on a Block 4. You may be misconstruing the meaning of "design stability," or maybe GPS had funny requirements. There wasn't an F9-based GPS launch until Block 5 was deployed. And I have no idea what "aerospace-grade mission assurance" is.
Where'd you get 142 ULA launches from? I counted 111 Deltas and Atlases from December 1, 2006, which is when ULA was incorporated. Before that, it's Boeing for Delta and LockMart for Atlas.
If you go much further back in the Delta II/III/IV history, there are some failures. There's also a partial Atlas V failure of an NRO mission in 2007 (payloads went into low but viable orbits), which I missed the first time through.
Object-oriented programming is more of an academic concept and not heavily practiced in industry in 2023.
Huh? OOP is no longer the hot new thing, and functional programming is expanding out of academia, but I still use object-oriented languages and APIs every day. Which part of the industry do you work in?
No one really does "functional" either. You just... figure out what you want to achieve, and then write the code that makes the computer achieve that thing. That's it. No contemplation of the true Platonic essence of an object and how it should relate to other objects in the hierarchy, no quibbling over a function that has a side effect or could be refactored as a quadruply-curried higher-order function - you just write the thing you want to do.
I think you are using a uselessly narrow definition of "practiced in industry". Sure, OOP as practiced in industry usually falls short of the strict forms discussed in academic papers. But OOP in its less strict, git'er'done form is *everywhere* - and, for the purposes of this discussion, that's enough to have had a big effect on the refactorability of code.
There are some use cases like desktop GUIs or simulations where OO programming is well suited for the task and continues to be used for good reason.
Desktop GUIs are all Chrome now, and I'm sure no simulation software uses OOP.
Um. What? Java (released mid 90s) is built around OO (admittedly a pretty simplistic and limited version of it). So are most mainstream languages today (e.g. Python and C++). C is probably the biggest exception. People have been using OO in industry since at least the late 80s with Smalltalk.
I'm old enough that pervasive use of OOP was a big mid-life paradigm shift, and I spent a fair amount of time trying to wrap my head around it. But today, the reason you don't see a lot of discussion of OOP is because it's pervasive. It's what all the cool kids learned in school, and they don't even think about it.
That's not to say that there aren't plenty of very poorly designed class libraries and interfaces. But that's no different than how coding has always been: 80% of the stuff gets done by 20% of the coders, who have to struggle mightily to unsnarl the messes made by their less talented colleagues.
Say what now ? If you've ever used the "." operator in Python, or iterated over a list, then you've been using OOP. I suppose OOP is not of much use in firmware programming, but there's more to the world than firmware...
Theory aside, I think it's fairly clear that software quality has degraded over the last decade or two; if, as you say, this coincides with a shift to agile development, this is at least some evidence against agile in software (though not overwhelming evidence, there are other potential explanations).
I think it's fairer to say that, for applications where software quality isn't that important, quality has declined. For mission-critical applications, quality--and functionality--have improved considerably.
I've been retired for a while, so my knowledge of the tools isn't super-duper these days. My impression is that integrated development/test environments are so much better than they were, say, 15 years ago, that you can get arbitrarily dense test coverage over your entire code base. So it's really a question of how much you want to invest in building test suites, and how hard you want to ride your coders on unit testing.
But yeah, for stupid websites that nobody cares about, quality has plummeted. It'll be interesting to see what happens as generative AI grows into that space. I have no clue what'll happen.
There is a wasteland of failure for small rocket startups that were high risk iterative models.
It will be much more difficult for SpaceX to stay the way it is once success has been achieved. Making dramatic iterative changes to a rocket with 230 consecutive successful launches and a low relative cost doesn't have much upside. The long story short is there is no guarantee that they just won't become another Boeing because that is the natural trajectory for things like this. Fast scrappy successful startups turn into fat lazy engorged monoliths. Google. Apple.
I wish them success, following Starship closely.
SpaceX has been quite public about the F9 design being largely frozen. That's a perfectly reasonable thing to do with your cash cow. It's especially reasonable when the cash cow generates a ton of paperwork to NASA and USSF any time you breathe on it.
The same is not the case for Starship, where SpaceX is dreaming up weird stuff all the time and trying it out. A lot of it never makes it off the pad, but the pad is a major--possibly even the dominant--component of the launch system.
I don't disagree that SpaceX will eventually get slow, bloated, and stupid. But it hasn't happened so far. I wish they had a viable competitor to stave off the sclerosis.
It's a lot easier to iterate faster on software, especially when almost perfect abstraction barriers are the norm. But I agree that Boeing seems to be iterating much more slowly than is ideal. One big potential reason is that when you have interactions between different organizations you often have contracts specifying what needs to be delivered which are very hard to revise. There's a barrier like that between SpaceX and NASA, but otherwise SpaceX gets to keep revising within those bounds. Boeing is the prime contractor among a web of outsourcing and so this sort of rigidity reaches much further down with them. You also have politics developing withing organizations, per the Iron Law of Bureaucracy, which can make different departments interface like separate legal organizations.
The difference is primarily that Boeing is owned by stockholders who mostly don’t care about the long run and sure don’t like talk about “bet the company”…hence the 737Max.
I would argue that Boeing doesn't iterate **at all**. Yes, Boeing will produce elaborations of previous designs, but they all go through the full waterfall process before anybody starts bending metal or writing code.
A fair--and important--point, however, that Boeing's outsourcing is a huge encumbrance to moving fast and breaking things. All of Musk's hardware businesses start out with a minimum of outsourcing, then get more and more vertical over time. He wants control over the entire supply chain, so he can fiddle with any portion of it at will.
As a person who has experience in building stuff with multidisciplinary teams (SW, HW, RF and MECH), with hard problems in each discipline, I want to push back hard on Agile having anything to do with SpaceX being better or faster.
When people talk about agile usually mean a mix of two things:
- The rituals: Standups, retrospectives, and so on. Not focusing on this as it's just ways of trying to make people talk with each other and pressuring each other to perform.
- Iterative design: The great discovery of SW is that iterations are for free. You can build something minimal and then continue adding features forever. When a SW company builds an app, it's going to start extracting revenue from the minimal product (or your VC), and is going to continue iterating on it for a long time, instead of having a fixed delivery date with fixed requirements, you just prioritize which feature goes next based on costumer feedback.
The problem of the rest of the disciplines is that iterations are not for free. They consume money and time. Waterfall project management is just the assumption that the money and time are not infinite, and if at some point you want to integrate the stuff that each discipline has made, you need to have your ducks in a row in a certain point in time.
SW people are used to working just with other SW people, and have absolutely no idea how hard is to make a guy with a phD in theoretical magnetometrics communicate with a mechanical guy who is used to hammer away his problems: they could as well be speaking different languages. You just can't assume that if you let those people iterate in their own, the end result will be something manufacturable that does what it's supposed to do, fits where is supposed to be placed, and it's compliant with the relevant standards.
So the solution is to set official reviews in place, plan for a number of iterations, set processes in place that define the interfaces on the disciplines... And you will be looking at something that is quite similar to what Boing and NASA do. The problem is that these dinosaurs set their processes in a time when CAD was in it's infancy, drawings where actual paper drawings, reviews could take months and they have not incentives to change them.
Now, there is quite a lot of low hanging fruit in the CAD and simulation space. I don't know about the field of rocketry, but in my neck of the woods, most of the enterprise tools were built in the 90s, are an absolute pain in the ass to use, integrate with each other horribly and have terrible version control.
I can imagine that a company with a monomaniatical focus on reducing iteration time could make a review that takes months in those dinosaur companies in hours. In the company I work for, the first times we where doing cross RF-HW-MECH simulations it would take us days to make sure that what every discipline had in it's CAD was the same as the rest, now we are down to minutes by automating the integration between the tools, but it was not easy and goes against what a lot of senior people are used to.
Mind you, this has nothing to do with Agile vs Waterfall. I have a gantt chart always open where I can see what me and my team needs to deliver and when, there are several iterations planned, some of them including cross-discipline dependencies, and if the alternative is a bunch of stickers on a wall, you'll pry it from my dead hands.
Personally, I would say that the biggest thing that Elon has done for making his company special is hiring very smart and driven people, not let them be encumbered by legacy processes, and hammer them continuously in a very specific direction. Smart people, if you don't let them diverge, will end up figuring out the best way of doing things.
EDIT: Rephrasing.
Interesting pushback. I wonder how many will read your description of improved waterfall and think "mmm yes, agile".
Dunno... In my mind if you have delivery dates, fixed external requirements that don't move through the project, and a gantt chart, it's just not agile. Of course, Agile proponents have this Not-True-Scotsman thing going on where everything that works is Agile, and if it doesn't you are just not using it correctly...
it IS agile, but only because agile is a big spectrum.
In the context of this discussion: Every thing that is not classic waterfall model has a non-zero amount of agile in it.
You have classic waterfall on one end with fixed deadlines and a single iteration. And you have stuff like extreme programming on the other end. Scrum is somewhere closer to XP (i.e. it does have fixed dates for delivery, but it's open what will be delivered). And spiral model or V modell or "improved waterfall" are closer to the classic waterfall modell. (I am sure you know all this, but I am explaining it to 3rd parties, who might stumble upon this comment chain)
It's iterative, but not necessarily agile.
To a first-order approximation, SpaceX **doesn't** have delivery dates. Instead, they have a very smart but fundamentally unreasonable leader whose personal multiplier (the actual-delivery-date-to-initial-delivery-estimate ratio) is north of 2 (not uncommon amongst engineers). If Elon had a manager, they'd be able to trim his estimates back before they hit the SpaceX Twitter account. But he doesn't.
So everybody has to scramble before Elon loses patience and fires them. I'm sure that makes the culture fairly painful, but also rewarding, because stuff gets done that otherwise wouldn't--or couldn't.
Musk joked a while back that the unofficial company motto was, "Making the impossible merely late."
I hope I was clear that "iterative" and "agile" aren't the same thing. All agile is iterative, but not all iterative is agile.
There's no question that the iteration cycle for hardware is longer than that for software. However, in SpaceX's case, there are three features that enable their iteration:
1) The first, somewhat reductively, is that they don't use waterfall very much, if at all. They don't sweat the requirements. Indeed Musk has a standard "all requirements are dumb" mantra. He even goes so far as to say that you should especially mistrust requirements from a smart person that you otherwise trust. They do the absolute minimum to get something off the pad.
2) They're incredibly software-rich. They build their own tools. They even build stuff like CFD suites in-house. I don't know what they use for CAD, but if there are tweaks that improve their process and productivity, they'll invest in them.
3) They run incredibly hardware-rich. This is largely a function of Musk being cash-rich, with lots of like-minded investors, so they're not afraid to blow stuff up. Being private helps as well, since they don't have to write down the stuff they blow up in public accounting.
Hiring well obviously helps. But SpaceX is big, and big companies can easily attenuate the efforts of their smartest employees. Process is always the backbone of every successful project. They have some secret sauce here.
The Isaacson bio makes reference to Elon’s interest in the mid 20th century Southern CA defense industry, and the idea to use a tent to rapidly expand manufacturing space as they did once at Tesla.
That story comes from Lockheed Skunkworks division. What seems to distinguish Skunkworks, when you look closely at their original operating principles, is an early version of iterative vs waterfall manufacturing.
Because of the tent and the Isaacson reference, I strongly suspect this was a primary influence for Elon, along with iterative being the de facto method in software.
For hardware, I also suspect “iterative” design is more about rapid prototyping into a stable production design, versus true continuous delivery. Still a very big advantage over waterfall.
Another thing I think gets overlooked is that SpaceX is at least as much a software company as a hardware one.
Back in the Olden Times, lots of aerodynamic and structural regimes simply weren't available to the engineer, because they weren't dynamically stable. You had to make your hardware dumb, because the compute power and software sophistication to stabilize dynamically unstable systems was impossible. That started to change in the late 70's, but the hardware was still pathetic.
When SpaceX spun up, they could essentially throw arbitrary amounts of compute power at problems for almost no mass penalty. Ultra-fast networks and sub-millisecond minor cycles let you design hardware a lot more flexibly--and cover up a lot of design flaws that would otherwise send you back to the (literal) drawing board--than was previously possible.
Again, it's much easier for a company founded in the oh-oh's to capitalize on this change than it is for a company that stopped doing methodology innovation before most of its current employees were born.
Great points. Also, out my depth here, but I wonder what they have in place to create a CICD-like toolset for their overall design and engineering pipeline.
I'm curious about the state of the art here. Does anyone know what proprietary tools they have in that realm?
I don't know the answer to that. I'd kinda guess that continuous delivery would be a little dicey for flight software, but they also produce a huge number of tools in-house, and I'd bet that they use a CICD model for those. There's also of course a huge amount of launch processing and ground segment software, which probably needs fairly tight control, but not as tight as what flies on the vehicles.
I'm too lazy to dig it up, but there was a really interesting presentation by the guy who was in charge of SpaceX's in-house computational fluid dynamics system. They've done some really innovative things to tailor the size of the CFD grid to what they have available for compute resources. Rather than a fixed grid of points on the vehicle, they have an algorithm that makes the gird sparser or denser on various parts, based on how the CFD solution appears to be shaping up. And it's all completely custom, proprietary code.
On Musk's mental health:
Something that hasn't gotten a lot of discussion is his ongoing cervical radiculopathy problems. He had a total disc replacement a few years back, which, by his account, failed and had to be redone, presumably as a more standard fusion. (The Blofeld-like scarves and neckerchiefs he's often seen wearing are, I believe, an attempt to cover up the scars.)
I've had a similar problem, and I can attest that the pain is excruciating and chronic, and the only thing that really helps are opioids. My problem occurred just before cervical TDRs became a widespread technique in the US, and I hung on, using ice packs and Norco, until I was able to get a TDR. My surgery was successful, and I was fortunately able to wean myself off the opioids without too much trouble, but I wouldn't be at all surprised if Musk had to rely on them. Even if you're careful, getting the balance right between enduring the pain and not being gorked out on the meds isn't easy.
Musk has always been a bit erratic, but things got markedly worse about four years ago. That was, I believe, close to the time when his surgeries were occurring. Correlation is obviously not causation, but the timing kinda matches up.
I'd speculate that mental <--> physical issues tend to feed one another.
"It is from this obsessive identification with the body, I believe, that many physical and emotional problems arise. In meditation, as we learn at deeper and deeper levels that the basis of our personality is not physical but spiritual, such problems fall away. Often we do not even have to confront them. We simply go deeper, move away from the tenements of consciousness where they arise.
…
You can leave psychosomatic problems hanging in the closet and find another home in a much safer neighborhood; they will never be the wiser. This approach is very different from the conventional wisdom of the modern world. It cannot work without meditation. For it is much more than a change in lifestyle; it is a transformation of thought-style.”
--Eknath Easwaran, Conquest of Mind, 1988
Elon Musk was noticeably in a grumpier than usual mood during the 3 part Starbase tour he did with Tim Dodd (The Everyday Astronaut). I'm having trouble finding it now, but somewhere Tim flashed up a tweet/message from Elon where he apologized and said it was due to back pain.
He also tweeted about it how he smashed his "c5-c6 disk" during Sumo wrestling
https://x.com/elonmusk/status/1505789670776610817?s=20
Yes, and that's why he was putting off the Zuckerberg MMA match (and probably would have lost). It's an interesting anecdote because you have to wonder if it's related and thus an example of the sort of behavior that contributes to all the mood-disorder comorbidities: you don't see other comparable men like Richard Branson or Larry Ellison or Jeff Bezos getting injured doing things like... [checks notes]... "trying to throw an actual sumo wrestler who weighs twice or thrice what they do by sumo wrestling". (Men their age are usually a *little* more aware of the fact that at that age, injuries, especially joint injuries, no longer go away.) But it is just one of countless anecdotes about Musk's risktaking.
Yeah, agreed! I actually was not aware of the sumo wrestling thing, just came across it while searching Twitter this morning.
Would be interesting to try to track down when he started taking sleeping pills also. Back pain and mental health issues both make it harder to sleep. He's tweeted a couple times how he takes sleeping pills occasionally, mentioning Ambien on at least one occasion. Prescription sleep medications can trigger depression in some people as well as memory and cognitive issues.
rebuttal to tg56's comment mistakenly comparing Tesla's balance sheet vs Ford's by citing long-term debt. tldr: the long term debt is a function of Ford having a financing arm and is more than offset by financial assets and equity cushion. Details in link below
https://seekingalpha.com/article/4417862-captive-financing-arms-avoiding-pitfalls-in-comparing-teslas-balance-sheet-to-those-of-auto-oems
I do not think that the Boring Company is going to build effective public transit.
Musk does not seem to want to build transit that involves sharing a vehicle with strangers. He's said this explicitly: "there's like a bunch of random strangers, one of who might be a serial killer."[1] It's also what he has designed. The Vegas Loop involves individual cars. Hyperloop was supposed to be composed of 'pods'.
This is a major problem. Putting multiple people in a vehicle allows you to get much higher densities.[2] By insisting on individual cars, you're accepting a 5x (relative to bus) - 50x (relative to large subways and commuter rail) reduction in the capacity of the transit line.
The choice to use cars is not because it's a small diameter tunnel. The London Tube has 12 foot tunnels. They use unusually small trains for a subway, but are still able to move a lot more people than cars could.
The Boring Company can probably reduce the cost of tunneling in the United States. The US builds transit at 10x the cost of Italy or Spain.[3] Plausibly Musk can even improve on global best practices. But because of his choice of what to put in the tunnels, he's starting with an order-of-magnitude handicap.
[1] https://www.wired.com/story/elon-musk-awkward-dislike-mass-transit/
[2] https://en.wikipedia.org/wiki/Passengers_per_hour_per_direction
[3] https://transitcosts.com/
> By insisting on individual cars, you're accepting a 5x (relative to bus) - 50x (relative to large subways and commuter rail) reduction in the capacity of the transit line.
Only compared to a transit line that consists of a constant end-to-end stream of buses/trains though. This almost never happens; a typical transit line has a train and then a gap of several kilometres and then another train behind it. A constant stream of private cars has a much higher passenger density than a train every N minutes, especially if it doesn't need to stop at every station.
It's also a helluva lot more pleasant, and I'm not a believer in making people suffer for the sake of efficiency anyway.
A constant stream of passenger cars along one lane of a highway has similar passenger density as a train every hour. At this frequency, it's not worth building a tunnel for a train either.
As an example, there are two ways to travel from Oakland to San Francisco: the Bay Bridge and the Transbay Tube. The bridge has 10 lanes for cars going towards San Francisco, while the tube has one "lane" for trains going towards San Francisco, which run every 2.5 minutes (maybe 4 km apart). Peak traffic occurs during morning rush hour, when the bridge carries 14,000 people per hour and the tube carries 27,000 people per hour. [1]
I like pleasant private transportation too. I just think that it's worth recognizing that this is what the Boring Company is working on. They're not going to revolutionize mass public transit by putting electric cars in a tunnel.
[1] As of 2018. See page 8 of https://www.bart.gov/sites/default/files/docs/BART%20HSCMCP%20Final.pdf
Fine, now I've sat down and done the maths. A jam-packed subway train (1500 people) every three minutes has a capacity of 30,000 people in an hour. You'd need 8.3 people going past every second to manage that with a single lane of traffic, which is implausible even under generous assumptions. So you're right on that one.
But of course if tunnels were cheap, you wouldn't have a single chokepoint like that. People don't want to go from West Oakland to Embarcadero, they want to go from Alameda to the Marina, or Berkeley to SFO, or Richmond to Richmond. You'd have a whole bunch of tunnels crossing the Bay in a whole bunch of different places so that people can get from where they are to where they want to be, efficiently.
That would only make sense if the cars themselves took up no space and had no need to keep any distance from each other in case one of them suddenly stops; otherwise, you'd have every single one of them crashing into the car in front of them in a massive chain reaction.
All the same limitations that apply to trains also apply to cars (it seems strange to me that anyone would not see this), except that people can sit or stand directly next to each other in far higher density in trains or buses than they can while sitting and driving in cars.
Public transportation is also much safer than cars. Musk being afraid of the potential serial killer next to him seems comical, considering the prevalence of serial killers. He should be much more afraid of his own capabilities to consistently drive safely and the capabilities of every driver on the road. You don't need a crazed murderer to die on the road; you only need a moment of inattention or a lapse in judgment.
Driving in a car is also at most moderately more comfortable than taking a train (or not at all if you actually have to drive yourself), unless you have a deep disdain or disgust for other people, which might be the case for a member of the ivory tower bourgeoisie. However, that's not something to be proud of; on the contrary.
I also feel that much of this perspective comes from an American standpoint. Maybe you (and Musk) would not dislike public transportation as much if you had actually experienced good public transportation. Where I live, nearly no one I know owns a car; there's no point. Public transportation is most of the time more comfortable, vastly cheaper, and safer. And often, it's even faster because finding a parking spot in the city sucks.
Re: going to Mars, since Elon has said the most entertaining outcome is the most likely, then I am putting my money on a massive asteroid wiping out the first SpaceX colony there about 15 years in.
The only one of those kids' names that would be normal to anyone with an understanding of American pop culture would be the one that was changed (to another normal name). 9% is, uh, abysmal as a batting average.
Also, that painting is something else. I hope that comment regarding calling Elon Musk "daddy" is some effort at trolling.
Should have written that a ton of furries have migrated from twitter to bluesky: i thibk about 10% of blueskye is furries, and one of the biggest accounts on blueskye, braeburned, is a furry porn artist that has 10% of the site following him
Many left leaning or progressive minded people have made alternate accounts on other websites but havnt left twitter yet, because they are uncertain if its actually gonna collapse
RE: sclmlw's comment on the Boring Company and other comments about size/cost
The Boring Company's car tunnels are about 14ft outer diameter (it says 12ft inner diameter on their website, assuming 1ft liner thickness, I think 14ft is fair). A typical subway tunnel might be 15ft-21ft outer diameter. Not much difference in diameter, but the cost approximately goes up with the cross sectional area, because of the volume excavated. So comparing 14ft to 20ft that's a 2x difference in volume of excavation. Also a larger diameter tunnel will require a thicker tunnel liner. It's reasonable to claim that the Boring Company could build tunnels maybe 1.5x to 2.5x cheaper than a typical subway tunnel.*
Score some points for the Boring Company - it is significantly cheaper. However it's not the 10x cost savings they're aiming for according to their website, that will need genuine technological innovation. Also I would counter that they can build the tunnel for 50% of the cost, but using cars, you're not even close to 50% of the passengers per hour (The Chaostician did a great example in another comment).
*(For an apples to apples comparison, this comparing 2 14ft tunnels to the more traditional "twin bore" subway system with 2 15-20ft tunnels, one in each direction of travel. There's also a modern "single bore" approach with a single 35ft-50ft diameter tunnel. This is more of an apples to oranges comparison because the tunnelling costs significantly more. But you save money by putting station platforms and misc parts of the station inside the tunnel itself. Plausibly this is the method west coast US people are hearing about in the news, because the Seattle Alaskan Way Viaduct and the BART to San Jose projects use single bore?)
RE: Safety
Safety wasn't a big part of my original comment because I don't have much experience there. I said they're going light on ventilation, exit walkways, and fire suppression systems (sprinklers). I was wrong about fire suppression. Looking at Youtube videos, they have a sprinkler system visible at the top of the existing Vegas Loop tunnel. Cutting out the exit walkways makes sense for a car tunnel - the bottom of the tunnel is already flat, instead of a subway tunnel with rails and other random crap like sump pumps and cables to trip over. My biggest gripe is the ventilation.
The ventilation of a tunnel has 3 purposes:
- Ventilate exhaust fumes (scratch this one for electric car tunnels)
- In case of fire, blow the fire/smoke in 1 direction, instead of letting it spread in 2 directions. Preferably use reversible fans to blow the fire/smoke to the direction with fewer passengers, or whichever direction the fire department desires.
- In case of fire, cycle fresh oxygen in the tunnel so passengers and firefighters don't suffocate (Fire consumes the oxygen. In principle you could NOT ventilate and try to smother the fire with lack of oxygen, but in practice I believe it's better to circulate air and send firefighters in there)
This is why in the bay area there's very large fans visible at the entrances to the Caldecott tunnel. IIRC they were retrofitted after a gasoline tanker caught fire in the tunnel.
Electric car tunnels still need a plan to deal with fire. (Google "tesla catches fire"). It says they have a fire safety system and a ventilation system on their website, but I didn't see any ventilation on a quick scan of Vegas Loop Youtube videos. I don't want to badmouth safety on a project I'm not familiar with, maybe they have hidden ventilation, or presumably if they're skipping ventilation they have a detailed fire safety plan ready to go with the Vegas fire department. Maybe electric car fires are a slow burn compared to gasoline car fires, and there's more time to send firefighters down there?
Sorry, scratch that last sentence about EVs being a slow burning fire https://www.youtube.com/watch?v=K5vDWhMHTwE
The Seattle Alaskan Way Viaduct replacement tunnel is a double-decker car tunnel, with no stops. You can see a cross-section on Wikipedia, and it seems like a decent way to handle 4 lanes of traffic, including large trucks? There were some major delays, including hitting steel that the tunneling company knew was there, and this might have made the news.
https://en.wikipedia.org/wiki/State_Route_99_tunnel
The Seattle light rail tunnels are a twin bore pair of 18-21 foot tunnels that handle 1 light rail track each (although the 18 foot part can also handle city buses). The stations are dug out separately.
https://en.wikipedia.org/wiki/Downtown_Seattle_Transit_Tunnel
https://en.wikipedia.org/wiki/University_Link_tunnel
https://en.wikipedia.org/wiki/Northgate_Link_tunnel
I'm a small hedge fund manager, race car driver (I own a Model S Plaid as a daily driver), and ex-software engineer. I've been following Musk for professional reasons for about 5 years now. While he demand tremendous respect in public markets, I've never got the impression he's super smart. Some anecdotes that come to mind:
His tweet on Tesla polar moments (https://twitter.com/elonmusk/status/1042842449008197632) is nonsensical. Outside of racing, no polar moment is not extremely important, and there's no way a Model 3's is better than a lightweight, mid-engined car.
When he announced the Tesla Semi, I analyzed his claims using data from the DOE SuperTruck competition (analyzing range is fairly easy if you have Crr and CdA figures). I found them to be dependent on rapid drops in battery prices, which I think has been shown to be accurate.
The Hyperloop whitepaper. It's all hand-waving nonsense, glossing over the important questions that would determine if the idea is viable. Can the air bearings support its weight? What does it even weigh? What happens if the tube leaks, or someone shoots a hole in it?
FSD. I have a subscription to it for market research purposes. I don't think this is Musk being dumb as much as it is fraud. Tesla really needed that FSD cash flow early in its life, so he said whatever he needed to. Now the company is walking back some of the earlier claims.
I followed his Twitter acquisition from the moment he suggested it on Twitter (thanks for that Elon). He bought at a really dumb time, when tech companies were valued absurdly and interest rates were rising. I can't completely fault him for this, as it's not his job (as it is mine) to track asset bubbles. However he did not seem to understand the reason social media companies cater to the left is because of advertising revenue? How could anyone causally following the business not know this?
Making Mars a backup Earth? You don't just need a colony, you need millions, probably billions of people to produce the technology needed to survive on Mars. This is basic economics. Ergo Mars is much more of a biological and terraforming problem than a rocketry problem.
He wanted x.com (or PayPal; I can’t recall the timeline) to switch from a LAMP stack to Windows in the 90s? My eyes can only roll so much. Good thing he was fired.
I've just never seen him say anything revelatory, and I learn things from objectively dumb people all the time. I pay close attention to Elon because he moves markets, not because he teaches me anything.
Going over his big successes:
Landing rockets: Well I know nothing about rocketry, but this seems really awesome. I don't know how much of this was Elon, and how much Gwynne or other engineers. Anyone know who originally had the idea?
Electric cars: Anyone who knew anything about cars knew BEVs were better if the batteries got good enough. He should be given some points here.
Starlink: This is more of a derivative of cheap launch costs. Good on him for realizing it.
Twitter: Assuming it succeeds (my guess is it will), that is some points in his favor. However Twitter was the worst-run big tech company on the planet. This would be a business success, but not evidence of super-genius.
I see his successes more as evidence of how terrible our society is at utilizing intelligence. There are lots of people out there smarter than Musk, but they lack the motivation, charisma, cult following, and narcissism to command the sort of capital he does. Musk can try all sorts of dumb shit, fail a dozen ways, and somehow still command respect and status. Most people are too excessively afraid of failure to do the things he does.
I think he’s smart enough to realize when the mainstream heard has it wrong, narcissistic enough to believe only he can fix things, and charismatic enough to build a cult of personality around this.
I wonder if his success is due to the existence of so many rich, low-neuroticism tech bros? He can do and say all sorts of stupid shit that most people would (as the media does) focus on, yet his fans merely shrug off. They know his successes are far more important, while neurotic wordcels focus on his many flaws. If this is the case, he might not have succeeded at any other point in history.
If years trading public markets have taught me anything, it’s that people are quick to form cults around their favorite stocks and CEOs. A ticker (more a person in this case) is simultaneously a flag to rally around and a bet that might make them fabulously wealthy. Whatever else he is, Elon is the greatest stock pumper of all time.
I agree with basically everything you've said. Just a small remark:
> Ergo Mars is much more of a biological and terraforming problem than a rocketry problem.
The biological and terraforming problems can only be approached once the rocketry problem has been solved. Getting to Mars safely and efficiently is a necessary precondition for making it inhabitable.
Of course, a much smarter path would be to colonize Earth's orbit first. Unlike Mars, orbital habitats actually provide an economic benefit in the form of a location advantage not found anywhere on Earth: microgravity.
Eventually yes, I agree. However surely there's a lot of theoretical work to do before we start sending rockets over? Maybe we're at the point where we need run tests on the red planet itself?
e.g. if nuking the poles is the answer, we ought to do that when there aren't any humans on the planet. If there are people there, they might actually slow progress down.
I tend to think the sending rockets to mars obsession has more to do with his narcissism than anything else.
We're at the point where we need to run tests on the Red Planet itself. And we're doing those tests, e.g. https://en.wikipedia.org/wiki/Mars_Oxygen_ISRU_Experiment.
But for now, those tests are pitifully inadequate because of the expense of getting anything at all to Mars; MOXIE had to be substantially descoped to fit the mission.
>But out of all of that, my biggest take-away was that Teslas..... just aren't very good? Their structures up to the Model 3 are quite inefficient and don't have great rigidity. The dimensional variation is shocking (far beyond even SBU, IYKYK). The hang-on parts are generally relatively poorly performing on their own. They can't touch our structural or powertrain durability tests. Rate and handling is bad, ergonomics fails to meets package targets, NVH and sound quality are poor, and we pay JD Power far too much to find out just how bad the quality numbers are (hilariously bad). I don't think it's an exaggeration to say that most other OEMs can't make a Tesla, because our systems and processes prevent us from releasing something that half-baked.
This is basically what I'd expect from Tesla trying to build a whole auto manufacturing ecosystem from the ground up - obviously it's not going to be as good as the incumbent one. That's presumably why you almost never see a brand new car company launch and successfully compete. Tesla managed it because they had a killer feature that no one else had - a functioning electric car!
>So where did everyone go? Was it Mastodon after all? Facebook Threads? Blue Sky? Or did they all start learning to paint and spending time with their friends and families?
I, for one, started reading actual published novels again, which I've barely done for the last 5ish years.
7 children was not a lot back when (Catholic) Tolkien was writing.
Yes, you’re right. The bio repeatedly describes Musk as completely insensitive to emotions — quite the opposite of the borderline type.
There does seem to be something fundamentally disorganized about his personality, though. Particularly the highly unstable relationships (divorcing Riley twice etc). I just don’t buy the bipolar story because of the rapidity of the shifts.
I’d be very curious to hear a real psychiatrist weigh in.
Physician here. Here is the grand hypothesis coming up in the medical community at the moment: chronic pain is best understood as the endpoint of an evolution within the nervous system that is increasingly called "central sensitization syndrome." (CSS) This syndrome is caused by a chronic dissociative or avoidant stress response, usually having it's origin in early childhood adversity. It's common in people who have a "push through" or type A approach to life. And you run into that type of personality in day laborers as well as fortune 500 ceos. Frequently in people with chronic pain there are other manifestations of CSS, such as IBS, chronic pain, TMD, etc. Every new stress, be it a viral syndrome, a grief event, a difficult move of house, what-have-you, can cause the central sensitivity to further evolve. Elon is an absolutely classic chronic pain patient. The fact he had a laminectomy a few years ago with ongoing chronic pain is evidence that maybe the laminectomy was not the right treatment. In general, the right treatment for chronic pain is addressing the "push through" and avoidant stress response originating in early trauma.
" A complex system that works is invariably found to have evolved from a simple system that worked. A complex system designed from scratch never works and cannot be patched up to make it work." -- John Gall (1975)
I've written about Elon Musk and autism. Might interest people as additional context to this discussion.
https://oneautisticperson.substack.com/p/forty-four