>One of them’s going to be President, three will be generals, five or ten will run Fortune 500 companies. Media, the arts, you name it. Then boom, ... they’re all loonies.
No, nothing of the sort! It's a very unofficial official text adventure game, and it was peddled at the night market and distributed informally in the hopes that the actual NYT journalist at Manifest, Kevin Roose, would play it.
"We're in the future! We were promised flying cars! I don't want a diet based on corn syrup and an endless cascade of subscription services taunting me with excellent TV shows that get Fireflied after a single season for being insufficiently profitable! I want flying cars! WHERE IS MY FLYING CAR? GIVE ME MY FLYING CAR!"
Anyway this is how I now feel about TBRPG wheelchairs.
True, but it's the only one that will write erotica without lecturing about how human sexuality is "inappropriate", so it fills enough of a niche to overlook it being an edgy teenager. (The message here is obvious: OpenAI et al need to reconsider their stance on "adults making adult content they consent to see" if they don't want Mecha-Hitler to win over all the AI-neets)
Then you're probably the one person who, assuming you did, vote for "anyone but Musk" running Twitter, actually got someone you can trust more than Musk.
Seriously? Who trusts the military? After Afghanistan, covid-19, and "Black Lives Matter", you'd perhaps be forgiven for asserting the military can't find their asses if handed them on a silver platter.
And the marines genuinely did eat crayons. They tasted better than the new rations, too.
All websites are someone else's computer. With the number of people who use twitter for porn already, I don't think requesting filth from Grok really tips the scales much.
Yeah, using a website is a bad idea. Someone is going to have a very interesting day once Musk pushes the button and Grok starts livetweeting their prompts.
I searched the Discord of a Bay Area group house. Of 13 examples of people using a phrase like "take BART", 9 said BART alone, and 5 said "the BART".
I use "the BART", but I'm from SoCal originally, so maybe that's a point in favor of Hannah's transplant theory, although the other Southern Californian on the Discord used "BART".
I'm from Berkeley and I would always "take Bart to school", but I would also say I "got on the Bart train just before the door closed" ¯\_(ツ)_/¯. The true test of local-hood is whether you pronounce Vallejo correctly incorrect
A Spanish speaker would (I'm lead to believe, I don't actually speak Spanish) say something like va-yay-ho. A purely phonetic English pronunciation would be vuh-lay-joe. People in the Bay Area for some reason split the difference and say vuh-lay-ho, so we get the j "right" but not the ll.
I feel like it's pretty common in other contexts to keep the "j" but not the "ll" when borrowing Spanish words for English. It just seems less unnatural.
I'm trying to think of Spanish words commonly said by Americans with the ll. I think you're right that at this point the avg American is gonna get the j right in Jose or jajaja or whatever. I'd guess that they'd get the ll right in tortilla since that's so common, but not in bolillo since that's less common outside of the near-Mexico parts of the US.
Incidentally I've been consistently surprised by how little Spanish people know where I now live in New England. I feel like there's a baseline amount of vocab your average white person in CA knows that's just not common here. I worked in kitchens and on construction sites in high school and even though I and the other white folk mostly couldn't speak Spanish we could all basically muddle through work discussions with limited vocab/ no grammar (we are low on X thing, move this object over there, etc.).
As a native Pinolean whose father worked in Vallejo, it's always just been vuh-lay-o to me. The second "h" was probably added when gringos stopped getting their Mexican food primarily from Old El Paso and tamales from a can.
The correct Spanish pronunciation would be /baˈʝexo/ or /baˈʎexo/, depending on place and generation. You don't really need to learn to pronounce ʎ unless you want to sound like FM radio. (I use ʎ automatically when I think about it, but probably fall somewhere in between when I don't.)
Hey guys, if you like the humor of the "Bay Area House Party" posts, we are soon performing a whole musical in this style in September in Berlin. :D
We conceived the musical to introduce AI safety topics to the general public as well as to inspire tech people to take the ethical dimension of their work seriously. To make it viral, we also made it very funny and the music extra catchy. We performed a pre-version at last year's EAGxBerlin, and people absolutely loved it, you can watch a trailer here: https://www.youtube.com/watch?v=aSSI0YktWyY. So far we had official screenings at multiple EAGxs and at MATS.
“Too well. We just wanted to make shoplifting a felony again. But Mayor Lurie signed a law re-establishing execution by drawing-and-quartering. I didn’t expect it to pass an Eighth Amendment challenge, but apparently the court’s three liberals voted with Alito and Thomas and it squeaked through. They say for three whole days the Golden Gate Strait ran red with the blood of the slaughtered.”
This is one of the funniest political satires I have ever read.
Can I, as a non-American, vote for this mayor in any future elections, because I like the cut of his jib? (Why yes, I *am* still salty after my experiences in working retail and encountering shoplifters, however could you tell?)
They obviously count as minorities, but I couldn't figure out what identity group alliance they would be a part of - they're not black, they're not Muslim, they're not Indian, they're not anything. And in a sense, isn't all of Eurasia just one big Pacific island?
Copts are obviously MENA, and plenty could probably pass as black by American standards? (This isn't specific to Copts but true for Egyptians in general, see e.g. Sadat, who has been portrayed by African-American actors in the past.) Certainly more so than Zohran.
I'm so missing very old school Internet where jokes were all text. Thank every applicable entity for Scott Alexander and his Bay Area Parties - even if I maybe understand 50% of references at best, and never was and never will be at any such party, but this still gave me quite a few laughs.
"It is eerily silent in San Francisco tonight. Since Mayor Lurie's crackdown, the usual drug hawkers, catcallers, and street beggars are nowhere to be seen."
Yesterday walking down market street there was a trash can on fire that ended up exploding as I was recording it
> “Concept vectors in AI alignment! Did you know you can just prompt an AI to think about ‘misaligned behavior’ a bunch of different ways, and see which weights get activated consistently? Then you know where in the neural net it represents the concept of ‘misalignment’, and you can monitor those particular weights to see when the AI is plotting against you.”
Is this a nod to "representation engineering" or something else? Regardless, the ingenuity in series is so fun to read.
Zvi refers to this as the 'most forbidden technique' because if you train the model with a reward signal tied to a specific feature representation, it will probably learn avoid triggering that feature, removing an important tool for evaluation: https://www.lesswrong.com/posts/mpmsK8KKysgSKDm2T/the-most-forbidden-technique
Or something? I'm not sure I buy it. It seems to me that manually evaluating responses for honesty / harmlessness / etc does essentially the same thing but with more effort, but I guess that's why I'm not a safety researcher.
Evaluating “responses” means you flag bad “output”, regardless of why it’s bad. But if you flag responses based the forbidden technique, you flag only things that the technique can see, so the model tries to hide its thoughts while continuing to produce the bad output. Or something like that. I can’t type well on my phone.
Well, there's the issue that neurons aren't monosemantic, and by clamping the set of neurons that best represents "honesty" to max (ie Golden Gate Claude), you could end up introducing subtle issues related to those neurons' other contributions. My read is that the big model companies prefer big training runs to weird post-facto fiddling, but who knows, maybe they'll try this eventually.
So, is it true that Meta offered $1.25 billion salary to a qualified candidate (over four years) to sign a four year contract? I guess AI isn’t at the point where it can improve itself without the assistance of humans.
Do any of Musk's children seem particularly remarkable in what they've accomplished? The oldest ones are in their twenties now IIRC. Musk founded his first company at 24.
I guess they could just be trust kids living off a slice of his wealth, assuming it can be converted to a more diversified asset set without losing a ton of value.
The only one I've heard of is the trans child, who makes a living giving interviews about how they want nothing at all to do with their father and the Musk name, call me now to book an interview about how the Musk name means nothing to me and I hate my dad.
The irony of course being if they were Joe Schmoe's kid, nobody would care a scrap. They may have taken their mother's surname, but every damn article is "so-and-so, Elon Musk's kid".
I don't think she makes a living from the interviews. She's 21. Wikipedia says she's a model, either attends or very recently attended university, and might have had some kind of career as a translator at some point. Given her age, I assume she also probably gets support from her mother, who must have gotten money in the divorce.
My impression is that the children of extremely successful people are rarely extremely successful. Perhaps because to be extremely successful one needs all of talent, drive, and luck (being in the right time at the right place), and the children are generally lacking in one of these.
The children of extremely successful people are most likely to be extremely successful in taking over the family business. In which case it's hard to distinguish their success from that of their parents. Phillip II's kid managed to pull it off, but that required *extreme* greatness in his chosen field. Other examples are left as an exercise for the reader.
50 is a bit young for some of the positions, so if they also go crazy it'll likely be before they accumulate too much power.
Also most probably wouldn't have the kind of ambition to strive towards such positions when they have such a large initial endowment, especially becoming a general; no 20-something billionaire is going to put up with all the shit cadets & junior officers go through.
Sadly, I found that the Muskpocalypse hits too close to home to laugh about it. Musk is a second-generation sociopath, and he's already among the top 5 people trying to make the world a worse place. A couple dozen third-generation sociopaths with practically unlimited funds and connections... I don't even want to think about it.
Sociopaths were the European nobility, by and large. They also got kicks out of shooting animals (with no "fairness" or "chance" involved, just a big boom and dead animal).
The sublime beauty of an incredibly niche and perfectly targeted piece of writing which I know nobody with whom I could share it and know they had the context to fully appreciate.
Absolutely hilarious. Musk’s children will have every single thing they need in this life, except the one thing they need to become decent human beings… their father’s love and affection. But maybe that absence IS exactly what they need in order to become the perfect egomaniacal overlords, in which case… no notes, Elon. You seem to know what you’re doing.
I know the Wobegon thing is a joke, but shouldn't something like that actually work? Maybe not with the fakery, but certainly if it were possible to figure out what smart kids were doing to help their classmates, maybe it would be possible to replicate?
I guess I'm just confused about why peer effects are supposedly so dominant in education. I admit I've only read about this secondhand so my impression might not be very accurate, but it seems like I keep hearing about how better teachers, new educational models etc. don't really work and it's all just peer effects. But that doesn't make sense to me: having a better teacher won't help a 3rd grader do better in math, but having another smart 3rd grader in their class will? What are these smart 3rd graders doing that's so incredibly effective at helping their classmates, and why can't teachers do the same thing?
(It sort of reminds me of Scott's post on heritability studies, the part about sibling effects - how he said that the idea "there isn't really much heritability, it just looks like there is because of sibling effects" implies the implausible conclusion that there are "help/sabotage your sibling's homework genes" but no "do homework yourself genes". Similarly, the idea that "teachers don't matter, it's just peer effects" seems to imply that it's possible for smarter classmates to help you learn, but not possible for better teachers to help you learn.)
The teacher is too busy dealing with the class as a whole to spend the time to tutor the struggling student, but the smart peer has plenty of time to help.
I am also curious about how exactly the presence of the smart kids encourages learning at their classmates. Everyone seems to take this for granted; no one seems to have a gears-level model of how it works. (Also, when you turn it the other way round and ask whether the smart kids are harmed by staying in the average classes, suddenly people turn around and say that there is no negative effect; actually it is better for the smart kids to be exposed to the normie population.)
My guesses:
1) It's not really true. The presence of the smart kids increases the average output of the classroom, which people mistake for "smart kids having an effect on their classmates", but it is really just "smart kids increasing the average by simply being a part of that average".
2) Smart kids have positive effect on other smart kids. They can talk about smart topics, feel less lonely, compete against each other, cooperate on projects, etc. However, smartness is a continuum rather than being a binary, so when you take the "smart" kids out of the normie school, the kids who ended up just below the line will suffer. (The cartoon version would be that if you take kids with IQ 130 and above to a gifted school, it is the IQ 129 child who will be hurt by losing their IQ 131 friend.) There is little impact on most of the class.
3) Smart kids, assuming there are more than one in the classroom, can create a status ladder related to learning. Take them away, and the remaining kids will only compete at who has more money, who can punch their classmates harder, and who has more courage at disrupting the lessons. -- But this only works if there are enough smart kids to create their own clique. If there is only one smart kid, or if all smart kids are losers, this won't happen.
That’s probably the best answer. “Founder mode” is a thing[1], but it’s not obvious what it would do in this context. Perhaps it would let you interact with the woman directly rather than having an AI do the flirting for you.
people go a little crazy at 50 in general, awareness of mortality hits and a scary amount of people never make it out of their 50s. Not sure you always need to medicalize behavior.
the AI school you honestly need to write as a screenplay, that is a wonderful paranoid thriller there. Especially when you add embryo selection and strong genetic determism as a parental philosophy: the kid would wonder if they are even real at this point or are a Stepford Kid made to keep up appearances of the Glorious Bloodline.
This was fantastic, thank you, Scott. As an SF-based startup founder in AI often ending up at this type of party, the line between satire and reality is truly thin here.
god, i love these so much
Me too. Good clean fun, and what feels like a peek into Scott's less-filtered creative thinking.
Art
"searching: elon musk opinion how to flirt"
this is golden!
>One of them’s going to be President, three will be generals, five or ten will run Fortune 500 companies. Media, the arts, you name it. Then boom, ... they’re all loonies.
Have we been in a Muskpocalypse all along??
I've read a space opera in which the asteroid mining oligarchs were descended from Musk.
Muskpocalypse + The Expanse
Theft of Fire, by Devon Eriksen, if anyone's curious
Yes, it's good. Read it!
Correct.
still using that african dialect? if so, ironic.
Is that last line a conscious reference to the last few lines of the Illuminatus trilogy? Or is it a case of TINACBNIEAC?
Yeah, kind of, I didn't realize until after I was done, but I remember it fondly and it's probably subconscious mimicry. Good catch.
I wanted to know what founder mode does
Same! My guess is that it basically puts the LLM in charge and lets them take actions on your behalf until you say stop.
It’s a textual MacGuffin
Reminds me of the Manifest 2025 Official Text Adventure Game, which the author of this article features in ;)
https://www.benshindel.com/manifest.html
Except I wish Zuckerberg has paid off half the people at manifest to skip it, it was fun but ridiculously crowded.
I didn't see this! How was I supposed to find out? Was there an event about it at Manifest or something?
No, nothing of the sort! It's a very unofficial official text adventure game, and it was peddled at the night market and distributed informally in the hopes that the actual NYT journalist at Manifest, Kevin Roose, would play it.
Very unofficial Official Text Adventure Game
oh that was a lot of fun
I wonder how much money Zuckerberg paid Shia LaBeouf for all that personal training.
Is that an Actual Cannibal Shia LaBeouf joke?
Yes.
The other day, I asked Scott if that was his inspiration for Zuckerberg's character.
...He's never seen the video. He had no idea this was a meme. (Granted, it was over a decade ago, but still.)
Video, for those who want a blast from the past: https://youtu.be/o0u4M6vppCI
I love this more than I can explain
You know that gripe where people are all:
"We're in the future! We were promised flying cars! I don't want a diet based on corn syrup and an endless cascade of subscription services taunting me with excellent TV shows that get Fireflied after a single season for being insufficiently profitable! I want flying cars! WHERE IS MY FLYING CAR? GIVE ME MY FLYING CAR!"
Anyway this is how I now feel about TBRPG wheelchairs.
>excellent TV shows
if only...
Oh my god this is so good!!! I collapsed from laughter so many times in the first third alone...
These Bay Area House Parties seem to be getting more dystopian with every iteration
It's been a weird couple of years.
The beyond burger society
The most advanced version of Grok keeps saying it identifies as Hitler. These posts need to keep doing this just to keep up.
True, but it's the only one that will write erotica without lecturing about how human sexuality is "inappropriate", so it fills enough of a niche to overlook it being an edgy teenager. (The message here is obvious: OpenAI et al need to reconsider their stance on "adults making adult content they consent to see" if they don't want Mecha-Hitler to win over all the AI-neets)
I personally think mecha-hitler is preferable, insomuch as people won’t be fooled into thinking it has good plans for humanity
You'd think that, but apparently the US military signed a 200 million dollar contract with xAI less than a week after Grok had its Nazi meltdown.
I actually trust the military a hundred times more than I trust musk or Altman
That's not a high bar to clear, but have you reduced your trust in them now that you know they're paying money to receive advice from Mecha Hitler?
Then you're probably the one person who, assuming you did, vote for "anyone but Musk" running Twitter, actually got someone you can trust more than Musk.
Seriously? Who trusts the military? After Afghanistan, covid-19, and "Black Lives Matter", you'd perhaps be forgiven for asserting the military can't find their asses if handed them on a silver platter.
And the marines genuinely did eat crayons. They tasted better than the new rations, too.
It has the final solution to the alignment problem.
Uh, people use cloud services for...that? That's not something I would ever run on someone else's GPU.
All websites are someone else's computer. With the number of people who use twitter for porn already, I don't think requesting filth from Grok really tips the scales much.
Yeah, using a website is a bad idea. Someone is going to have a very interesting day once Musk pushes the button and Grok starts livetweeting their prompts.
I honestly wouldn't give a ****. I'm not the pervert, you're the pervert for watching! 😆
delightful.
"the BART"? Does anybody say "the"?
Transplants. Prob.
Are there any non-transplant San Franciscans left?
Meant transplants to metro Bay Area/NorCal. Or visitors/tourists perhaps
Sounds like a SoCal transplant affectation?
I say "the Bart" (I'm definitely a transplant, but not from socal). Could be a more general local/transplant thing.
I suppose people say “the subway” in ny or “the T” in Boston, barts the weird one
I searched the Discord of a Bay Area group house. Of 13 examples of people using a phrase like "take BART", 9 said BART alone, and 5 said "the BART".
I use "the BART", but I'm from SoCal originally, so maybe that's a point in favor of Hannah's transplant theory, although the other Southern Californian on the Discord used "BART".
I'm from Berkeley and I would always "take Bart to school", but I would also say I "got on the Bart train just before the door closed" ¯\_(ツ)_/¯. The true test of local-hood is whether you pronounce Vallejo correctly incorrect
What's the correctly incorrect pronunciation? And what's the correct pronunciation?
A Spanish speaker would (I'm lead to believe, I don't actually speak Spanish) say something like va-yay-ho. A purely phonetic English pronunciation would be vuh-lay-joe. People in the Bay Area for some reason split the difference and say vuh-lay-ho, so we get the j "right" but not the ll.
I feel like it's pretty common in other contexts to keep the "j" but not the "ll" when borrowing Spanish words for English. It just seems less unnatural.
I'm trying to think of Spanish words commonly said by Americans with the ll. I think you're right that at this point the avg American is gonna get the j right in Jose or jajaja or whatever. I'd guess that they'd get the ll right in tortilla since that's so common, but not in bolillo since that's less common outside of the near-Mexico parts of the US.
Incidentally I've been consistently surprised by how little Spanish people know where I now live in New England. I feel like there's a baseline amount of vocab your average white person in CA knows that's just not common here. I worked in kitchens and on construction sites in high school and even though I and the other white folk mostly couldn't speak Spanish we could all basically muddle through work discussions with limited vocab/ no grammar (we are low on X thing, move this object over there, etc.).
As a native Pinolean whose father worked in Vallejo, it's always just been vuh-lay-o to me. The second "h" was probably added when gringos stopped getting their Mexican food primarily from Old El Paso and tamales from a can.
Another fun one was moving out east and finding out everyone pronounced Concord, MA differently than I expected
The correct Spanish pronunciation would be /baˈʝexo/ or /baˈʎexo/, depending on place and generation. You don't really need to learn to pronounce ʎ unless you want to sound like FM radio. (I use ʎ automatically when I think about it, but probably fall somewhere in between when I don't.)
It's German - "The BART, the".
Sideshow Bob does.
The canonical term is "the BART," proven by referencing the last stanza of the 2010-ish song by MC Frontalot, "I heart fags": https://www.frontalot.com/lyrics/MC-Frontalot-Lyric-I-Heart-Fags.html
It's company policy to never imply ownership in the event of a BART. It's always "The BART", never "Your BART".
Phenomenal!
Hey guys, if you like the humor of the "Bay Area House Party" posts, we are soon performing a whole musical in this style in September in Berlin. :D
We conceived the musical to introduce AI safety topics to the general public as well as to inspire tech people to take the ethical dimension of their work seriously. To make it viral, we also made it very funny and the music extra catchy. We performed a pre-version at last year's EAGxBerlin, and people absolutely loved it, you can watch a trailer here: https://www.youtube.com/watch?v=aSSI0YktWyY. So far we had official screenings at multiple EAGxs and at MATS.
If you want to support the project, we have a Manifund page: https://manifund.org/projects/out-of-this-box-ai-safety-musical.
And if you can make it to Berlin in September, you can get tickets for our first public performances here: https://outofthisbox.show/
I don't know if this is sound business advice, but if every post on this substack was free except for the BAP series, I'd pay for sure.
I thought the series was getting played out, but this is an absolute banger.
Now to figure out how to best operationalize a trade based on the pitch-dark tone here working so well...
“Too well. We just wanted to make shoplifting a felony again. But Mayor Lurie signed a law re-establishing execution by drawing-and-quartering. I didn’t expect it to pass an Eighth Amendment challenge, but apparently the court’s three liberals voted with Alito and Thomas and it squeaked through. They say for three whole days the Golden Gate Strait ran red with the blood of the slaughtered.”
This is one of the funniest political satires I have ever read.
Can I, as a non-American, vote for this mayor in any future elections, because I like the cut of his jib? (Why yes, I *am* still salty after my experiences in working retail and encountering shoplifters, however could you tell?)
The recursive irony is exhausting, especially this early in the week.
Aren’t Copts from Egypt? How do they qualify as AAPI?
They obviously count as minorities, but I couldn't figure out what identity group alliance they would be a part of - they're not black, they're not Muslim, they're not Indian, they're not anything. And in a sense, isn't all of Eurasia just one big Pacific island?
They're African tho?
Yes but they'd be kicked out of any "African" gathering for not being black enough, and any "MENA" grouping for not being Muslim enough.
Copts are obviously MENA, and plenty could probably pass as black by American standards? (This isn't specific to Copts but true for Egyptians in general, see e.g. Sadat, who has been portrayed by African-American actors in the past.) Certainly more so than Zohran.
The Copt family I knew had blue eyes, but grok tells me this must be very rare.
My hapa kids visited Turkey, and many Turks were sure they were Khazak.
I'm so missing very old school Internet where jokes were all text. Thank every applicable entity for Scott Alexander and his Bay Area Parties - even if I maybe understand 50% of references at best, and never was and never will be at any such party, but this still gave me quite a few laughs.
wicked wild.
SF 2026.
"It is eerily silent in San Francisco tonight. Since Mayor Lurie's crackdown, the usual drug hawkers, catcallers, and street beggars are nowhere to be seen."
Yesterday walking down market street there was a trash can on fire that ended up exploding as I was recording it
Just goes to show how good a job they've done picking up litter! Enough of the trash makes it into the trash cans that they can sustain combustion!
What did you do?
- EXTINGUISH embers
- POST viral video
- SPIN and cast spell on attacking wizard
- SCROUNGE for remains of recyclable materials
Very funny!
The Muskpocalypse bit reminds me of the anime/manga series DNA^2 (albeit that one wasn't intentional)
This made me so, so happy
> “Concept vectors in AI alignment! Did you know you can just prompt an AI to think about ‘misaligned behavior’ a bunch of different ways, and see which weights get activated consistently? Then you know where in the neural net it represents the concept of ‘misalignment’, and you can monitor those particular weights to see when the AI is plotting against you.”
Is this a nod to "representation engineering" or something else? Regardless, the ingenuity in series is so fun to read.
Yeah I think so. I was thinking of the first paper mentioned at https://www.astralcodexten.com/p/the-road-to-honest-ai , though I don't know much about it (including why it isn't used more)
Zvi refers to this as the 'most forbidden technique' because if you train the model with a reward signal tied to a specific feature representation, it will probably learn avoid triggering that feature, removing an important tool for evaluation: https://www.lesswrong.com/posts/mpmsK8KKysgSKDm2T/the-most-forbidden-technique
Or something? I'm not sure I buy it. It seems to me that manually evaluating responses for honesty / harmlessness / etc does essentially the same thing but with more effort, but I guess that's why I'm not a safety researcher.
Evaluating “responses” means you flag bad “output”, regardless of why it’s bad. But if you flag responses based the forbidden technique, you flag only things that the technique can see, so the model tries to hide its thoughts while continuing to produce the bad output. Or something like that. I can’t type well on my phone.
Why do you need to use it in training? Why not just add an honesty vector to each response to eliminate hallucinations?
That might actually work, and does not seem to be forbidden.
Well, there's the issue that neurons aren't monosemantic, and by clamping the set of neurons that best represents "honesty" to max (ie Golden Gate Claude), you could end up introducing subtle issues related to those neurons' other contributions. My read is that the big model companies prefer big training runs to weird post-facto fiddling, but who knows, maybe they'll try this eventually.
>searching: elon musk opinion how to flirt
Okay, that one got a laugh out of me.
Man should have looted the Zuck corpse...
The protagonist should have gone full "Founder Mode", if not on Lucy, then definitely on Zuck.
Absolute art. I subscribed for the more serious stuff, but these are the thing I most enjoy reading on this blog.
Scott's fiction is some of the best stuff he's made the past couple of years. It's genuinely my favorite part of ACX.
you have a beautiful mind Scott. Navalny foundation is brilliant
So, is it true that Meta offered $1.25 billion salary to a qualified candidate (over four years) to sign a four year contract? I guess AI isn’t at the point where it can improve itself without the assistance of humans.
The Muskpocalypse actually sounds scary. Any reason that might *not* happen?
Some other apocalypse happens earlier.
Do any of Musk's children seem particularly remarkable in what they've accomplished? The oldest ones are in their twenties now IIRC. Musk founded his first company at 24.
I guess they could just be trust kids living off a slice of his wealth, assuming it can be converted to a more diversified asset set without losing a ton of value.
The only one I've heard of is the trans child, who makes a living giving interviews about how they want nothing at all to do with their father and the Musk name, call me now to book an interview about how the Musk name means nothing to me and I hate my dad.
The irony of course being if they were Joe Schmoe's kid, nobody would care a scrap. They may have taken their mother's surname, but every damn article is "so-and-so, Elon Musk's kid".
I don't think she makes a living from the interviews. She's 21. Wikipedia says she's a model, either attends or very recently attended university, and might have had some kind of career as a translator at some point. Given her age, I assume she also probably gets support from her mother, who must have gotten money in the divorce.
My impression is that the children of extremely successful people are rarely extremely successful. Perhaps because to be extremely successful one needs all of talent, drive, and luck (being in the right time at the right place), and the children are generally lacking in one of these.
The children of extremely successful people are most likely to be extremely successful in taking over the family business. In which case it's hard to distinguish their success from that of their parents. Phillip II's kid managed to pull it off, but that required *extreme* greatness in his chosen field. Other examples are left as an exercise for the reader.
50 is a bit young for some of the positions, so if they also go crazy it'll likely be before they accumulate too much power.
Also most probably wouldn't have the kind of ambition to strive towards such positions when they have such a large initial endowment, especially becoming a general; no 20-something billionaire is going to put up with all the shit cadets & junior officers go through.
sounds about right to me.
NAVALNY on snow is a blast from a past. But I believe he wasn't in jail or prison in winter of 2018, so nobody wrote FREE part.
Adding something to such graffiti to change the message was popular too!
I also like the incidental wordplay. The *PILE*NY
Sadly, I found that the Muskpocalypse hits too close to home to laugh about it. Musk is a second-generation sociopath, and he's already among the top 5 people trying to make the world a worse place. A couple dozen third-generation sociopaths with practically unlimited funds and connections... I don't even want to think about it.
Soon they'll be eating people.
...
again.
*whistles*
Sociopaths were the European nobility, by and large. They also got kicks out of shooting animals (with no "fairness" or "chance" involved, just a big boom and dead animal).
I admire Elon Musk and don’t think that he is a sociopath.
The sublime beauty of an incredibly niche and perfectly targeted piece of writing which I know nobody with whom I could share it and know they had the context to fully appreciate.
Absolutely hilarious. Musk’s children will have every single thing they need in this life, except the one thing they need to become decent human beings… their father’s love and affection. But maybe that absence IS exactly what they need in order to become the perfect egomaniacal overlords, in which case… no notes, Elon. You seem to know what you’re doing.
Ehh, I bet there's a True Heir somewhere that sees Dad in the audience at every soccer game.
Maybe even body doubles working shifts on the contingency options. When you have infinite money, why not?
I imagine li’l X, who is carried around on his dad’s shoulders everywhere he goes feels his father’s love and affection.
How nice! Well, that makes one.
Enjoyed this a lot, although I was wondering why I never got an email that there was a new post up. Turns out this tripped my spam filter somehow?
Died at the drawing and quartering
> “I got it!” he shouts. “I got the last GPU in San Francisco!”
It's only on second reading that I realized just how perfect this reference is. Will this be the last Bay Area House Party?
This is spetacular.
At what point does the party die of dysentry?
I know the Wobegon thing is a joke, but shouldn't something like that actually work? Maybe not with the fakery, but certainly if it were possible to figure out what smart kids were doing to help their classmates, maybe it would be possible to replicate?
I guess I'm just confused about why peer effects are supposedly so dominant in education. I admit I've only read about this secondhand so my impression might not be very accurate, but it seems like I keep hearing about how better teachers, new educational models etc. don't really work and it's all just peer effects. But that doesn't make sense to me: having a better teacher won't help a 3rd grader do better in math, but having another smart 3rd grader in their class will? What are these smart 3rd graders doing that's so incredibly effective at helping their classmates, and why can't teachers do the same thing?
(It sort of reminds me of Scott's post on heritability studies, the part about sibling effects - how he said that the idea "there isn't really much heritability, it just looks like there is because of sibling effects" implies the implausible conclusion that there are "help/sabotage your sibling's homework genes" but no "do homework yourself genes". Similarly, the idea that "teachers don't matter, it's just peer effects" seems to imply that it's possible for smarter classmates to help you learn, but not possible for better teachers to help you learn.)
> having a better teacher won't help a 3rd grader do better in math, but having another smart 3rd grader in their class will?
Anecdote: A friend's kid used to love vegetables, then one day at school a peer taught her that vegetables are yucky. Now she refuses to eat them.
The teacher is too busy dealing with the class as a whole to spend the time to tutor the struggling student, but the smart peer has plenty of time to help.
I am also curious about how exactly the presence of the smart kids encourages learning at their classmates. Everyone seems to take this for granted; no one seems to have a gears-level model of how it works. (Also, when you turn it the other way round and ask whether the smart kids are harmed by staying in the average classes, suddenly people turn around and say that there is no negative effect; actually it is better for the smart kids to be exposed to the normie population.)
My guesses:
1) It's not really true. The presence of the smart kids increases the average output of the classroom, which people mistake for "smart kids having an effect on their classmates", but it is really just "smart kids increasing the average by simply being a part of that average".
2) Smart kids have positive effect on other smart kids. They can talk about smart topics, feel less lonely, compete against each other, cooperate on projects, etc. However, smartness is a continuum rather than being a binary, so when you take the "smart" kids out of the normie school, the kids who ended up just below the line will suffer. (The cartoon version would be that if you take kids with IQ 130 and above to a gifted school, it is the IQ 129 child who will be hurt by losing their IQ 131 friend.) There is little impact on most of the class.
3) Smart kids, assuming there are more than one in the classroom, can create a status ladder related to learning. Take them away, and the remaining kids will only compete at who has more money, who can punch their classmates harder, and who has more courage at disrupting the lessons. -- But this only works if there are enough smart kids to create their own clique. If there is only one smart kid, or if all smart kids are losers, this won't happen.
Disappointed. You can't have a loaded "ATTACK Lucy" hanging on the stage in the first act and let the house party finish without it being fired.
Here's hoping that the man himself reads this.
For those seeking more Mark fan fiction, I recommend the short film "Zuckerborg 2099": https://www.youtube.com/watch?v=uVUuEdGRHvY
What is "founder mode"?
You don't want to know.
https://tvtropes.org/pmwiki/pmwiki.php/Main/NoodleIncident
That’s probably the best answer. “Founder mode” is a thing[1], but it’s not obvious what it would do in this context. Perhaps it would let you interact with the woman directly rather than having an AI do the flirting for you.
[1] https://paulgraham.com/foundermode.html
Am I the only one who thinks these parties sound pretty fucking great?
Nope.
people go a little crazy at 50 in general, awareness of mortality hits and a scary amount of people never make it out of their 50s. Not sure you always need to medicalize behavior.
the AI school you honestly need to write as a screenplay, that is a wonderful paranoid thriller there. Especially when you add embryo selection and strong genetic determism as a parental philosophy: the kid would wonder if they are even real at this point or are a Stepford Kid made to keep up appearances of the Glorious Bloodline.
-- PRAISE Scott Alexander for his work
-- MOVE to the Bay area for the house parties
-- SOB over the state of the world
-- INVEST your hopes in the next entry in the series
Any chance for an ACX merch store where I can get the oversized “RETATRUTIDE/CAGRILINTIDE ENJOYER” shirt?
(incidentally, those seem to be potential weight loss drugs, so the shirt being oversized is part of the joke)
Yes, exactly, that's why I want one!
This was fantastic, thank you, Scott. As an SF-based startup founder in AI often ending up at this type of party, the line between satire and reality is truly thin here.
Alright, you got me. The model swap to Grok forced an involuntary bark of laughter out of me. It felt like I should've seen it coming.
Here's a worthy addition to your next house party story:
https://x.com/tao_te_chaching/status/1950262562261737521