I'm particularly interested in reactions from readers familiar with using basic hierarchical Bayes methods for practical problems. I'd expect that there would be a few such readers of a blog whose slogan is "P(A|B) = [P(A)*P(B|A)]/P(B), all the rest is commentary."
I am looking to start my own medical practice. Talking to people about it has been strange, because I feel like I don't really understand what a lay-person is looking for when they choose a doctor. I am a PCP with a lot of extra skills and I constantly hear that people want "a doctor like that" (for example, I am also a yoga teacher). I'm just looking for some clues about how people found their current primary and what things would convince a person to change (and possibly pay more out of pocket)?
In the US, my primary consideration is whether they accept my insurance. Second is how easy they are to schedule and work with.
One doctor I used for a year (before I moved) had absolutely amazing service. He had a website like "Doctorblackmountainradio.com" which advertised specifically many of the insurance carriers he accepted and what it covered with him (insofar as annual checkups, gym rebates, preemptive testing, etc.), so I didn't have to go through an incomprehensible system to understand if I was going to have to pay or not, or how much I would have to. He also had super transparent scheduling, where next day, or even same-day scheduling was an option.
The transparency was great, but better was that he actually made me understand how the whole thing works. Exactly why I should schedule yearly checkups and how I can do so without paying. What he was actually doing and what other options I had after receiving a checkup, and why I would choose any of them. I ended up paying out of pocket for some additional health check things that I was mildly interested in ahead of time, simply because he discussed with me for a bit. I assume he realized I was a young man in his 20s (and thus likely loosely aware of things like testing testosterone and other biomarkers), and explained some actual stuff about what metrics mattered, what were likely to be important to me, and what was almost certainly a waste of money. I don't imagine that's easily scalable, and he was a relatively young doctor so I imagine he hasn't yet been beaten down by the monotony of entitled/insane/normal patients.
I highly recommend (if you're in the US) to sign up for Zocdoc, and try to get a few of your existing patients (or just friends or family) make at least one appointment through there, and leave you a glowing review, as that's how I've found my primary care physician each time I've moved to a new area.
I go to my insurance company's website and use the search tool to find out who within a geographically reasonable area accepts my insurance. Then with a handful of choices, maybe I look at a few online reviews to make sure I'm not accidentally signing up with someone terrible, but probably I pick the closest one.
Of course it may be different for other people, but my view of a primary care physician is pretty much like my view of an auto mechanic. I just want someone who can do the job and get me in and out efficiently. If you're looking for advice, I'd go with things like "it's very easy to make appointments!" and "They actually tell me how much money I am likely to be paying instead of grinning while we hit the insurance company roulette." Maybe, "The waiting room is very nice and doesn't blast me with Fox news."
Ah, thank you! Seems like it should be comfortable, stuff should just work, and it shouldn't involve surprise expenses.
How would you expect to see that last part signaled? My current academic medicine practice has a policy where if you bring up *anything* during your routine annual that isn't part of a routine annual (a refill, my toe hurts, etc.) then we hit you with a bundled visit and a copay. The policy is printed out and stuck on the inside of every room (print is small). I rarely get complaints, as the copay is usually less than $40, but it has always felt scummy.
I don't plan on carrying this policy over to my new practice, but other than the explanation I wrote above, how can I say "I won't do that"?
I mostly disagree with Marcus, but your post does open with "In June 2022, I bet a commenter $100 that *AI would master image compositionality by June 2025*." (emphasis mine)
Strictly speaking, it hasn't mastered it, has it? It's just much better than it used to be, but even now it will probably fail if I give it 10 relationships to track (which isn't too crazy if I have a specific image in mind I want it to approximate).
It depends on the friends. For our (my husband and my) closest friends, we see them 1-2 times a week. We play board games, eat together and chat - about work, mutual friends, family, religion and politics, whatever's on our minds. We don't text/email often outside of that, except to arrange things and/or for something particularly noteworthy. The case is similar for the friends I see only every few months (dependent on how much they like board games!).
We also have a group of friends we watch anime and movies with, in which case it's usually most of a day, hosted by one of us. We watch the films and eat and chat, usually about geeky things. That takes place once every couple of months or so.
In both cases, I enjoy the time with people and look forward to the arranged dates, but it is also nice to come home (or to wave goodbye) and relax afterwards.
I recently moved to a different country because of work, so my situations probably doesn't generalize.
> How often do you guys meet?
About once every 2 weeks, either I go back home, or friends come here, usually for a weekend. Also I sometimes call friends to talk about whats up (about once a week). Note: These numbers are for all friends together; I meet each individual friend about once every 1-2months.
> What do you guys do when you meet up?
When they come here, we do vacation-stuff (hiking, boating, museums etc). When I go there, we usually cook and watch anime/movies or we go to concerts/events.
> What are you talking about?
when we are together we talk about hobbies (e.g. I do 3d printing and show them some recent projects and failures) and work most of the time. Sometimes a friend talks about something no ones else cares about and then we carefully tell them to shut up.
> What are you texting about?
weekend-plans and memes
> And how does all of that make you feel?
good but exhausted. I have to force myself to socialize, because when I isolate myself to for too long, I become bitter and angry.
I've tried understanding the "adding more lanes to a road doesn't improve traffic" argument that many urban planning types and fans of public transportation like to advocate, but I just don't get it.
As I understand the argument: If you are a frustrated commuter with an hour long drive, you may want an additional lane added to your route to handle more traffic. But this won't actually improve your commute, because the added ability to handle more vehicles will cause people who take other roads to divert to the widened road.
I get how this wouldn't improve my individual commute. But from a utilitarian perspective, isn't the commute of the people switching to the widened road improved? Surely *someone's* commute got better, or else no one would switch routes.
This isn't the most common argument, but one issue is that in many places the bottleneck isn't an expandable main road. E.g. if you're driving along an Intercity highway but the city you're driving to has too many congested small roads to handle the traffic, you'll (at best) get a bit further down the highway before getting stuck waiting to enter the city and (at worst) have to wait even longer now because the extra highway lane convinced more people to drive to it but the actual bottleneck (city streets) can't actually handle more traffic. And city streets are often impractical to expand.
(This is the issue with the road I've been stuck on most, between Jerusalem and Tel Aviv. Adding a lane to it wouldn't really help, especially since the topography probably doesn't allow that in the tight sections anyway).
More generally, space need for cars scales superlinerly (somewhere between linear and quadratic) in the number of vehicles, as adding highway capacity requires multiple road expansions within the target cities, more parking, and so on. It's not (usually) strictly true that adding a lane makes traffic actively worse, but it does in general help much less than you're expected (since a congested highway is a symptom of an overburdened system and you may not be able to add capacity to the rest of it).
Near me there is a junction between two highways. You can take an offramp from one highway that merges into the other highway's left lane.
At some point somebody noticed that the ramp was wide enough to fit another lane. And it was divided. Now you had two lanes on the off/onramp that simultaneously merged into the same one lane of the target highway. The target highway was not expanded; you just had two ramp lanes merging into each other immediately as they merged into the target lane.
This caused truly incredible traffic jams.
The problem was ultimately solved by eliminating the extra lane from the ramp.
It was a pretty stunning example of lane expansion not always being a good idea, but as you note, I don't think this is the kind of thing people are generally talking about.
As i understand it, yes, individual participants can benefit in the short run, but the theory says that total demand goes up - people move further into the suburbs, stop using public transport etc. until eventually, the new equilibrium is the old equilibrium except with an extra lane.
The usual term is "induced demand". After equilibrium is achieved, it need not be the case that anyone's commute got better, but it must still be the case that someone's *life* got better.
That person might be someone who moved to the region after the lanes were expanded, making the question of whether you should vote for local lane expansion potentially hazy.
So apparently in New York State, where I live, a huge amoint of elementary-level education, including assessments, is done using Chromebooks. I have children entering elementary school soon and, bluntly, I am convinced that this is a *terrible* idea and want them to be using physical books and paper for everything short of writing papers at the high school level. (Where they should still be using physical textbooks).
Does anyone have experience with or awareness of public school districts in New York or neighboring states that *don’t* use laptops at the elementary level?
"Suppose you took 10,000 optimally selected people and dumped them into a region that had adequate forests, fields, and mountains for mining iron and coal, all in a 100 mile radius
They start with only 1 season of food supplies
How fast could they bootstrap to tech circa 1900?"
Answers in the comments ranged from 2 years to 10k years, which both somehow seem plausible to me. It seems like thousands of technically capable people should be able to get a lot done in a few years, but then again there must have been some major bottle necks to the process that it took thousands of years irl.
A lot of answers got sidetracked by whether they'd just die from hunger or exposure, so say they start with seeds to plant and the climate is mild enough that the community is guaranteed to survive if they dedicate at least 90% of their man hours to farming and shelter in the beginning and become more efficient as their tech advances from there.
The coal/iron/steel is the easy part, honestly. Those are "big bulk things" that you don't have to get "precisely right."
Try cheesemaking, or distilling, or any number of the "practical chemistry" bits of technology you'd need to get to have "technology circa 1900" Not to mention, but you'd need glass for that. Also, 1900 is early enough for galoshes, so you'd need rubber.
And we haven't even touched clothing (textile mills) or "the great big bombs" that are steam engines. Waterwheels and Windmills seem easy enough to put together, though, so "practical automation" is doable.
1. You select 10000 people who are ideologically aligned, mentally stable, and knowledgable on the problems at hand (agriculture, manufacturing, etc. maybe skip medicine since this is a speedrun). This is a tough optimisation problem, so you have to sacrifice some knowledge on e.g. science in favour of people who are a bit less antisocial.
2. Prefer people who are better at establishing strong institutions and educating others, so that the next generation can learn how to do the labour.
3. 10000 people won't be enough to reach 1900 without kids, there will be a bottleneck in mining, science, or manufacturing somewhere.
So I think probably they could get it done in 120 years, or about 6 generations of kids.
If they were optimally selected, it wouldn’t take especially long before you could have basic ironworking producing tools and weapons. Maybe less than a year. From there it’s a pretty straight shot upward to steam power, but I think the precise ironworking and how difficult steel production is would involve a lot of trial and error. I’d be really surprised if anyone had both enough iron, and enough precision to produce an engine within a decade, but if they’re optimally selected… who knows?
The main problem would be food and access to resources. While it doesn’t matter if it takes a year or two to start making more complex iron tools, it does matter if you can or can’t get enough food production set up for 10,000 people (starting with no tools). I’d say the answer is likely no. Farming is very intensive work, 10,000 people is too much to be in one place, so you’ll have different groups at different levels of food production capability.
When you’re starving you’ll necessarily steal food to survive, and when you’re almost starving you’ll fight to make sure no one steals your food. Then you have social organization, centralization of food storage, tribes, and that itself if it becomes intense enough, might invalidate any technological progress. While I’m convinced optimally selected modern people could survive, I doubt their children could receive enough knowledge orally to do almost anything that their parents weren’t already capable of. If it’s just a vague idea of “You can mine coal, the black rock, out of the ground and burn it” and “if you produce iron in this specific way it’s stronger” that wouldn’t be enough to restart civilization, and might decay over more generations to exactly as much knowledge is useful to live in such an environment.
Just to clarify the question: are the "optimally selected people" able to overcome mutual conflicts, and resist burning most of the resources in various zero-sum games, such as "who is the boss" and "who can have sex with whom"?
Right, I think the “10,000 optimally selected people” would have to be, something like deeply faithful practicing Roman Catholics, with the right genetic distribution so only 1% had 99%ile iq and everyone was willing to be super obedient to the religious authority structure.
I... don't really think you need smart people for this. You need "butt-basic idiots" -- the folks who have learned how to do things the hard way, because it's fun. Smart people are good at solving "problems you've never seen before." But i don't think we're at that.
Even motivated midwits would work -- remember, with 10,000 people, you can have an "expert" on every link of the chain. Exposure shouldn't be an issue, you can always grab 10 Amish (or "Ron Swanson-esque" survivalists) who can order the rest of the folks about.
1900... Steam and locomotives. Mills, and upgraded farming.
Yeah, I think you may be right. You want reasonably intelligent persons selected much more for their combination of existing knowledge plus work ethic and pro-social norms. Though you may want one person tasked with getting this stuff written down lest the knowledge die off. I don’t think you could do this in a single human lifespan, you’d need cycles to bootstrap in the technology. It’s not just about knowing how but having the necessary tools, which have to go through iterations. Can’t go straight from raw iron to a lathe capable of micron-level precision. If your super knowledge persons were in their 20’s…. maybe you could pull it off in 40 years.
Of course you'd need cycles, but you don't need micron-level precision to get to 1900's technology. I think the iron/steel/coal troika is doable within ... maybe 5 years? Most of the steps there don't involve "precision" so much as "and now we make pig iron." (Pulling this from "that polish guy's science fiction," so I could be wrong ... but I don't think I am).
But... then you have "water purification" (aka BEER) Either you have the yeasts, or you don't (and, I believe, unlike bread, you need specific strains).
Distilling (also a pre-1900 thing I think) would require glassware.
And we have Yogurt! And Cheese! (If you don't think these are technology, boy, howdy). And sausage.
Basic looms/spinning are pretty easy to make. Not sure how they scale into "industrial looms and felting and..."
FluffyBuffalo here so people aren’t misled. Some amyloid theories are probably wrong, but some are almost certainly not. There is good evidence that amyloids are in fact involved in Alzheimer’s in some way, and that anti-amyloid therapies help moderately.
“I think it's important to be careful about what "subscribing to the amyloid hypothesis" really means. The evidence is very strong that amyloids play a significant role in the disease - genetic variants like APOE4 that influence amyloid accumulation are strong risk factors for AD; you see characteristic biomarker levels (in particular Ab42/Ab40 ratio in cerebro-spinal fluid) in AD patients that you don't get in other neurodegenerative diseases; you find the characteristic plaques in deceased patients, etc. That part is sound as far as I can tell, and pointing at a misguided study or two doesn't change that.
The question what exactly the amyloids DO, and whether the buildup uf the plaques is the whole story or just a part, is not so clear, and my impression is that researchers are open to the possibility that there's more going on. The new hot topic seems to be CAA: Cerebral Amyloid Angiopathy - it turns out that amyloids also accumulate in blood vessels, weakening the walls and leading to micro-bleeding. (Apparently, this got more attention when it was found that brain bleeding is a not-so-rare side effect of lecanemab.)”
Stupid question time: I've been trying to understand how exactly Heritability is defined, but the explanations I've found don't make sense.
Wikipedia tells me that, in simple cases, it can be computed as Variance(Genes)/Variance(Phenotype) where "Variance" means the expected squared distance from the expected value.
But this doesn't depend on any relationship between the two at all! By this explanation "Heritability" is just a measure of whether a set of genes is similarly spread out as the phenotype, not whether those genes explain anything.
Also, variance is measured in units of "distance in the underlying distribution" squared, so you can only take a ratio like this is the distances are comeasurable in some way, and I don't see how "polygenic score" and e.g. "has schizophrenia" can be comeasurable.
Stats in science are often bad, but I have trouble believing they're *that* bad. Can someone try and explain how this actually work
> Also, variance is measured in units of "distance in the underlying distribution" squared, so you can only take a ratio like this is the distances are comeasurable in some way
Variance infamously suffers from the problem that the same distribution at different scales has different variance. For example, If you measure everyone's height in inches, and if you measure everyone's height in centimeters, the height-in-centimeters distribution has more variance even though you measured the same set of people and they were all the same height both times.
The Alzheimer mouse review on Friday reminded me of Thrifty Printing Inc, whose business model of online photo-processing for cornershops didn't take off, apparently, so naturally the company pivoted, to drug development for such diseases as Alzheimer's.
But Anavex Life Sciences, that's the new name, doesn't seem much respected by the stockmarket. Was the pivot too ambitious? Or perhaps Anavex is seen as unserious by the professionals after hiring a runway model with no business experience as Director of Business Development and Investor Relations (Nell Rebowe). Or it's the hairstyle of the CEO (Dr. Christopher Missling). Or his presentation skills (that last one actually has some merit).
A recent Economist article, titled "The Alzheimer's drug pipeline is healthier than you might think", did not even bother to mention that Anavex has a pill under evaluation by the European Medicines Agency. Yet what will happen to the stock if it is approved in six months or so?
On the other hand, withdrawal or rejection would most likely send the stock crashing (not "certainly" because another drug's phase 2 trial for schizophrenia will report soon). To me, Anavex's pill appears disease-modifying (though not a cure) and safe (unlike the anti-amyloid drugs), but certain language in the company's latest 10-Q filing does strike me as ominous. Also, shareholders are up against high short interest, including Martin Shkreli, who has called Anavex "another retail trap" (referring to Cassava Sciences, popular with retail investors, whose Alzheimer's drug failed its phase 3 trial; that company's history involves fraud allegations). And whose strategy of shorting Alzheimer biotechs just scored another victory when INmune Bio's phase 2 trial failed to convince.
Anavex's pill is meant to stimulate autophagy and is not related to the amyloid hypothesis, as far as I know (I don't know much, but was still tempted to complain on Friday when someone called Leqembi a "proof of principle" for the hypothesis --- if people try for decades to develop drugs on the basis of a paradigm, wouldn't you expect them, even if they are mistaken, to come up with something eventually that shows an accidental effect, via some other mechanism?).
"A project in Co Mayo is generating renewable electricity through the flying of kites, which its operator has described as a potential "game changer" in the wind energy sector.
...The site, which is the first designated airborne wind energy test site in the world is being operated by Kitepower, a zero emissions energy solutions spin-off from Delft University in the Netherlands.
Kitepower's system employs a yo-yo effect, where a kite, measuring 60sq/m is flown at altitudes of up to 425m attached to a rope that is wound around a drum - which itself is connected to a ground-based generator.
The kites can generate 2.5 to 4 tonnes of force on the tether.
The pull from this force then rotates the ground-based drum at a high speed.
This rotation then generates electricity that can be stored in a battery system for deployment wherever and whenever it is needed.
The kites are flown using the knowledge and skills of kitesurfing professionals, combined with a highly specialised computerised GPS-guided steering system.
They fly upwards repeatedly in a figure of eight pattern for periods of 45 seconds.
The flight pattern is important because it forces the kites to behave like sails on a boat, maximising the pull of the wind to increase speed so electricity can be generated.
After 45 seconds, the kites are levelled up so that the pull from the wind is momentarily minimised.
This enables the tether to be wound back in, using only a fraction of the electricity generated when it was being spun out.
The result is a net gain in renewable power at the simple cost of flying a kite.
Then the cycle is repeated, again and again, potentially for hours on end."
(2) Well, looks like AI is *already* taking er jerbs (at least if you're a graduate in finance):
"Also this week, the latest 'Employment Monitor' from recruitment firm Morgan McKinley Ireland found notable reductions in graduate hiring by major firms in the accountancy and finance sectors because of the adoption of AI.
And on Thursday, AIB announced a major AI rollout for staff in conjunction with Microsoft Ireland, sparking concerns from trade unions.
Morgan McKinley Ireland's Employment Monitor for the second quarter of the year was published on Thursday.
The recruitment firm said that the standout development of the quarter was the significant impact of AI and automation, particularly within the accountancy and finance sectors.
"The notable reduction in graduate hiring by major firms, driven by AI capabilities, highlights potential challenges ahead," the report found.
"Companies are increasingly leveraging AI capabilities to automate routine tasks such as accounts payable, accounts receivable, credit control, and payroll."
...[Allied Irish Bank] announced a new artificial intelligence rollout for staff in conjunction with Microsoft Ireland on Thursday.
The bank said the new tools will reduce time spent on repetitive tasks, freeing up employees for higher-value work.
The plan will involve the widespread deployment of Microsoft 365 Copilot, embedding AI into everyday tools like Outlook, Word, Excel, Teams, and PowerPoint.
...Last month, the Chief Executive of AIB Colin Hunt took part in a panel discussion at a Bloomberg event in Dublin.
Asked what impact AI will have on staffing numbers at the bank over the next five years, Mr Hunt said it may lead to a small reduction in net headcount.
"I do think that there are certain manual processes that we do now that will be done by AI in the future, and probably net headcount will be broadly stable with a slight downward bias maybe," Mr Hunt said."
Who knew that the real knife-ears were the ones we created along the way! (See "The Rings of Power": "Elf ships on our shore; Elf workers taking your trades. Workers who don't sleep, don't tire, don't age. I say, the Queen's either blind or an Elf lover, just like her father.")
Is this not "fourth power of wind speed"? Because that's the issue, if you want "renewable energy" to stop generating more greenhouse gas (via the necessity of "keeping the grid stable" through spikes and dips -- the natural gas required for this is significantly more than "if we didn't use wind power at all.")
That story about graduates not finding jobs has been repeated across British papers for the last few months. If it is a trend then AI is not boding well.
It's mostly worrying about what all these graduates were being hired to do. I can see why graphic design would be toast, but the only other things AI can seeming do are "google this and write about it," "take this information and reformat it" and "spitball arbitrarily about this topic."
Sorry to be laughing about the tooth claims, because vision problems are no joke, but yeah. There's always something.
It's very unlikely the fancy tooth treatment is to blame, but it's not impossible, and it's only when you have real users out in the wild, as it were, that these one-in-a-million things crop up. Now we know the problems Big Pharma and the FDA face!
Reading about the twin studies... twins are similar not only because of genes but also because of a shared initial history in the womb. And not all twins are the same, it depends at which development stage they split in two different embryos (see "mirror twins")
If you like road trips and you still believe in the American Dream, join us in April 2026 for The Great Route 66 Centennial Convergence. We're driving from Chicago to LA over the course of a few weeks. There will be never-before-seen mysteries, side quests, and prizes. It costs nothing. Find more details on Facebook, Instagram, and Youtube.
Interviewees and video co-hosts wanted. We won't just be talking about hot rods and midcentury architecture. We'll be discussing the American Dream, capitalism, and human progress. It might even be a paid gig if you're charismatic, pleasing to look at, or slightly famous.
I'm a fresh grad about to start a Data Science-esque role for a big UK insurer. Never having worked in corporate environment, I'd like to ask the ACX community for sage advice on questions such as:
How can I learn the most?
How can I negotiate my salary throughout?
How do I handle 'office politics'?
or any questions someone in my position should think about.
If your office politics involve poisoning each other, just quit. If poisoning each other has been banned as being "not within the spirit of office politics" -- just quit.
This took me a long time to learn via experience, as it's not necessarily intuitive to rational people.
Aside from complying with things like critical regulatory requirements which might get you arrested for violating, you really only have one real job at work:
Make your boss [1] happy.
That is often very difficult for rational people leaving the merit-based school system and entering the social hierarchy of the workplace to fully grasp. You may have felt that school was your "job" and that your teachers were your "bosses," but this wasn't actually the case; as a student, you were a *customer buying a product,* which meant you had an absolute, enforceable *right* to receive clear expectations and an objective measure of your performance *as part of the product you were purchasing.*
In the case of your employer, they are the customer purchasing your time and attention... in the way they want you to provide those things Your paycheck and career aren't going to be based on you scrupulously, objectively performing all the duties of your job description or doing what's best for the company as a whole. Practically speaking, there's no way to really track that, and people who are determined not to see proof of your good work won't look at the evidence anyway. The people making decisions about your role at the company are going to be making them based on how they *FEEL* about you, which means your first priority needs to be making your boss *FEEL* good about you.
Keep.
Your Boss.
Happy.
So if your boss is very attached to a dumb process that you just KNOW could be vastly improved, your *real* job - the one which may get you promoted or saved from layoffs - is to make them *FEEL* good by surrendering to their dumb process. If your boss is overly-emotional and sensitive to criticism/rejection and perceive suggestions/corrections/warnings as personal attacks, your *real* job at all times is to make them *FEEL* good by avoiding triggering their negative emotions, even if that means allowing them to damage the company.
And remember that your grandboss and above haven't fixed or fired your crappy boss because there's something about your boss which makes *them* happy enough to want to keep your boss around. Don't assume *they* want to fix problems for the good of the company, either.
Surrender to the stupidity. That's what your customer wants, and that's your job now.
[1] "Boss" doesn't necessarily refer to your direct supervisor (although it probably will), but rather the persons/people above you who have the power to promote or fire you, including your grandboss, etc.
You should also keep in mind what your trajectory would ideally be during the next few years. Ideally you'll go from teachable youngster who is easy to work with and only needs to be told anything once, to junior colleague who can be trusted with small responsibilities, to actual peer who knows the ropes and can be trusted with substantial (and even unknown) challenges.
But right now you are _just_ the teachable youngster, and you need to embrace that role. Be conspicously eager to learn and respectful of advice from your older colleages. Do anything you are told to do, or even that you are suggested you might do, unless it is literally illegal. And if you haven't been given anything specific to do and everyone seems really busy, at the very least don't get in the way.
Office politics doesn't need to be tricky, especially as a junior; it's just the effect of companies being staffed with fallible flesh-and-blood human beings. Focus on making sure that people know you, and they like you, and allow that to be more important than being right all the time, and you'll probably do fine.
Similer to one of Wasteland Firebird's points, long term you'll do better if you hop sideways and upward between employers as opposed to spending years and years with the same one. This is because with each job change you should get a step up in position and salary, as well as a wider experience of the industry (and you don't necessarily have to stay with insurance companies - data science skills are fairly transferrable). But of course you shouldn't overdo the job hopping frequency, or potential employers will wonder why you can't stay in any one place for long.
Also, people tend to be better at office politics in inverse proportion to their technical ability: A weak performer needs to be crafty at office politics to offset their technical shortcomings, and conversely a tech guru, recognised as such, can often afford to disdain the politics.
When starting a new job, you should be suspicious of a colleague who seems too solicitous of your welfare when it isn't their assigned duty. It's similar to prison, although I've never had any experience of that, in that shortly after you arrive some apparently helpful fellow convict will sidle up to you and offer to provide anything you need and show you the ropes and so on. But ultimately it's for their benefit not yours!
When a company downsizes, the bean counters take zero account of the competence or otherwise of those being laid off.
> When starting a new job, you should be suspicious of a colleague who seems too solicitous of your welfare when it isn't their assigned duty.
I often do that
> But ultimately it's for their benefit not yours!
My "benefit" is that it makes the new guy feel welcome. Also talking to people is more fun than actual work. Also I guess it also makes my own boss and grand-boss happy when the new people come to me with their problems and not to them (boss and grand-boss).
Learning: Most of what I've learned, I learned inevitably on the job, or from switching jobs a lot. I haven't done much work in my spare time to keep up with the industry. It seems kind of silly to do that because you never really know what's going to be expected of you in your next job anyway.
Salary: The best way to get raises is just to take whatever you get, at first, and always be looking around for something that pays more. You don't even necessarily have to take the job that pays more if you don't want, you can get an offer and bring it back to show it to your current employer, and ask for a raise then. If you feel weird asking for a raise, don't ask for one. Just tell them you're leaving, explain why, and let them get the idea to make you a counteroffer. Eventually, you'll get to where you earn and/or have "enough" money. Figure out what that means for yourself. Then, the good part comes. Keep changing jobs, but this time, only change into jobs that make you happier! Sooner or later you'll get into a job where you simply can't realistically expect to find anything better. That's where I am now. I am hoping this will be the last job I ever have to take seriously.
Office politics: Even as a nerdy person who used to have no social skills, and who is probably, like many of us here, what might probably be diagnosed as a "person on the autism spectrum," I have to say, I actually really enjoy office politics. "Soft power" is a big thing. A lot of times you don't have any authority to actually make people do things, but you can get people to do things anyway. Ideally, you can figure out how to express the thing you want to do in terms people will appreciate, and then they'll join you in your quest. That requires learning real empathy, which is a difficult thing in itself. But when that doesn't work, sometimes there are little clever tricks you can pull that let you get your way. For example, there's a project I've been wanting to do, but no one is allotting my team any time to do it. So instead, I've looked around for other similar projects that were allotted time. I've treated those projects like they are higher priority than they actually are, and seen them through to completion. After a couple years, my dream project is 80% complete, and we haven't even technically worked on it at all yet.
Another fun trick I've used multiple times: You need to do something. You want to do it in a simple way. People with power over the project insist that the project must be done in a very complicated way that will take far too much time. Quietly complete the project on time, in the simple way. Get it as far as you can take it, so all they have to do is say, "Fine, good enough, ship it." The choice you've now forced on them is to do it your way, or to deliver it late.
Me and Lord Hobo Boom Sauce have been ruminating tonight about having an ACX site where we put up photos of ourselves as kids. Seems there would be very little risk of doxxing ourselves, so long as it's pix of us as *young* kids. And it would nudge us all a bit away from the godawful online illusion that the entity we are talking to via texts consist of a funky little fake name and a cluster of, like, 5 opinions. Anyone like this idea. or have I just gone down 11 notches in everybody's respect just for suggesting it?
I'm generally very dubious about putting up any kind of identifiable information because someone out there will pick up on it and try to identify real world you. I know that sounds paranoid, and I don't really have much to lose if someone figures out "oh Deiseach is really so-and-so" (apart from my jealously-guarded privacy) but for people who *do* have something to lose, I would be way more careful.
People *have* gone after Scott and those associated with him, see Sneerclub and the infamous Cade Metz story. It's not beyond the bounds of possibility that someone with a grudge about rationalists/Scott/SSC/ACX/TheMotte/LessWrong/you name it would latch on to anything like photos and try to work out who you are in real life and then send nasty little emails to your employer about "were you aware that Eremolalos is involved with right-wing fascists and racists?" (the HBD/IQ stuff is catnip to people with axes to grind).
I'm not saying "don't do it", just "be very very sure about the level of security".
I'm mildly intrigued, but I don't think I have any pictures of myself from a young age. Presumably my parents had some, but I think my mothers' scant few pictures of me as a little one went onto the "toss" pile when I was sorting her assets after her passing. I don't mind sharing more recent pictures, though, from when I lived somewhere else entirely (an attitude that can probably be surmised from my avatar).
An essay on the transformation of academic tutors into Turing cops, marking into an imitation game, and Al generated homework into the trigger for the technological singularity:
Maybe we'll see a return to the old-fashioned system of verbal examinations, whereby the student links via Zoom to an AI interviewer, which then fires questions at the student, who is required to make immediate extempore replies without referring to notes.
Hey Vincent, I read it, and have a non-ironic suggestion:
-It is probably possible to train an AI to be an excellent judge whether there is AI contribution to an essay, and how much, and whether it's in the main idea, the overall structure, or individual sentences or paragraphs. AI's -- which overall fucking suck, IMHO -- are pattern-identification geniuses. They are better than professionals at identifying melanoma, various retinal diseases, etc, from scans. You might need to hire somebodydo a bit of extra training to improve an LLM's ability to recognize these things, but I'm pretty sure it would be possible.
-OK, so then tell the students that if they submit something that scores more than a certain low percent on AI content, you will not read the piece and it will be graded by AI. I actually think that would discourage people quite a bit from turning in AI-contaminated work. I had a professor who would grade late essays, but did not put any comments on them. I really wanted to get comments, so never turned in things late to that prof, even though it would not have harmed my grade, and I was in general actually fairly bad about turning in papers late.
There are existing AI detectors that check for this, but the current models often give false negatives or wildly divergent responses. These could be improved of course, so I do agree with your suggestion. I guess the loophole then becomes that students could ask their AIs to write in ways that avoid the telltale signs. So we would still get an AI arms race of sorts.
> It is probably possible to train an AI to be an excellent judge whether there is AI contribution to an essay,
I just tried that by asking chatGPT about cats and then in a different instance asking it the probability of the cat text being AI. It was fairly certain it was.
This is admittedly a repetitious point, but for anybody who missed my response to Scott's analysis of Covid origins, here it is.
https://michaelweissman.substack.com/p/open-letter-to-scott-alexander
I'm particularly interested in reactions from readers familiar with using basic hierarchical Bayes methods for practical problems. I'd expect that there would be a few such readers of a blog whose slogan is "P(A|B) = [P(A)*P(B|A)]/P(B), all the rest is commentary."
I am looking to start my own medical practice. Talking to people about it has been strange, because I feel like I don't really understand what a lay-person is looking for when they choose a doctor. I am a PCP with a lot of extra skills and I constantly hear that people want "a doctor like that" (for example, I am also a yoga teacher). I'm just looking for some clues about how people found their current primary and what things would convince a person to change (and possibly pay more out of pocket)?
In the US, my primary consideration is whether they accept my insurance. Second is how easy they are to schedule and work with.
One doctor I used for a year (before I moved) had absolutely amazing service. He had a website like "Doctorblackmountainradio.com" which advertised specifically many of the insurance carriers he accepted and what it covered with him (insofar as annual checkups, gym rebates, preemptive testing, etc.), so I didn't have to go through an incomprehensible system to understand if I was going to have to pay or not, or how much I would have to. He also had super transparent scheduling, where next day, or even same-day scheduling was an option.
The transparency was great, but better was that he actually made me understand how the whole thing works. Exactly why I should schedule yearly checkups and how I can do so without paying. What he was actually doing and what other options I had after receiving a checkup, and why I would choose any of them. I ended up paying out of pocket for some additional health check things that I was mildly interested in ahead of time, simply because he discussed with me for a bit. I assume he realized I was a young man in his 20s (and thus likely loosely aware of things like testing testosterone and other biomarkers), and explained some actual stuff about what metrics mattered, what were likely to be important to me, and what was almost certainly a waste of money. I don't imagine that's easily scalable, and he was a relatively young doctor so I imagine he hasn't yet been beaten down by the monotony of entitled/insane/normal patients.
I highly recommend (if you're in the US) to sign up for Zocdoc, and try to get a few of your existing patients (or just friends or family) make at least one appointment through there, and leave you a glowing review, as that's how I've found my primary care physician each time I've moved to a new area.
I go to my insurance company's website and use the search tool to find out who within a geographically reasonable area accepts my insurance. Then with a handful of choices, maybe I look at a few online reviews to make sure I'm not accidentally signing up with someone terrible, but probably I pick the closest one.
Of course it may be different for other people, but my view of a primary care physician is pretty much like my view of an auto mechanic. I just want someone who can do the job and get me in and out efficiently. If you're looking for advice, I'd go with things like "it's very easy to make appointments!" and "They actually tell me how much money I am likely to be paying instead of grinning while we hit the insurance company roulette." Maybe, "The waiting room is very nice and doesn't blast me with Fox news."
Business customer service stuff.
Ah, thank you! Seems like it should be comfortable, stuff should just work, and it shouldn't involve surprise expenses.
How would you expect to see that last part signaled? My current academic medicine practice has a policy where if you bring up *anything* during your routine annual that isn't part of a routine annual (a refill, my toe hurts, etc.) then we hit you with a bundled visit and a copay. The policy is printed out and stuck on the inside of every room (print is small). I rarely get complaints, as the copay is usually less than $40, but it has always felt scummy.
I don't plan on carrying this policy over to my new practice, but other than the explanation I wrote above, how can I say "I won't do that"?
I mostly disagree with Marcus, but your post does open with "In June 2022, I bet a commenter $100 that *AI would master image compositionality by June 2025*." (emphasis mine)
Strictly speaking, it hasn't mastered it, has it? It's just much better than it used to be, but even now it will probably fail if I give it 10 relationships to track (which isn't too crazy if I have a specific image in mind I want it to approximate).
I'm struggling with my friendships, so I'd love to hear some anecdata. What are your friendships like? By that I mean things like:
How often do you guys meet?
What do you guys do when you meet up?
What are you talking about?
What are you texting about?
And how does all of that make you feel?
(NB: Obviously I'm not asking for details of the talks - "talking about a mutual friend's struggles" rather than "about Marissa's failing marriage".)
It depends on the friends. For our (my husband and my) closest friends, we see them 1-2 times a week. We play board games, eat together and chat - about work, mutual friends, family, religion and politics, whatever's on our minds. We don't text/email often outside of that, except to arrange things and/or for something particularly noteworthy. The case is similar for the friends I see only every few months (dependent on how much they like board games!).
We also have a group of friends we watch anime and movies with, in which case it's usually most of a day, hosted by one of us. We watch the films and eat and chat, usually about geeky things. That takes place once every couple of months or so.
In both cases, I enjoy the time with people and look forward to the arranged dates, but it is also nice to come home (or to wave goodbye) and relax afterwards.
I recently moved to a different country because of work, so my situations probably doesn't generalize.
> How often do you guys meet?
About once every 2 weeks, either I go back home, or friends come here, usually for a weekend. Also I sometimes call friends to talk about whats up (about once a week). Note: These numbers are for all friends together; I meet each individual friend about once every 1-2months.
> What do you guys do when you meet up?
When they come here, we do vacation-stuff (hiking, boating, museums etc). When I go there, we usually cook and watch anime/movies or we go to concerts/events.
> What are you talking about?
when we are together we talk about hobbies (e.g. I do 3d printing and show them some recent projects and failures) and work most of the time. Sometimes a friend talks about something no ones else cares about and then we carefully tell them to shut up.
> What are you texting about?
weekend-plans and memes
> And how does all of that make you feel?
good but exhausted. I have to force myself to socialize, because when I isolate myself to for too long, I become bitter and angry.
I've tried understanding the "adding more lanes to a road doesn't improve traffic" argument that many urban planning types and fans of public transportation like to advocate, but I just don't get it.
As I understand the argument: If you are a frustrated commuter with an hour long drive, you may want an additional lane added to your route to handle more traffic. But this won't actually improve your commute, because the added ability to handle more vehicles will cause people who take other roads to divert to the widened road.
I get how this wouldn't improve my individual commute. But from a utilitarian perspective, isn't the commute of the people switching to the widened road improved? Surely *someone's* commute got better, or else no one would switch routes.
What am I not understanding?
This isn't the most common argument, but one issue is that in many places the bottleneck isn't an expandable main road. E.g. if you're driving along an Intercity highway but the city you're driving to has too many congested small roads to handle the traffic, you'll (at best) get a bit further down the highway before getting stuck waiting to enter the city and (at worst) have to wait even longer now because the extra highway lane convinced more people to drive to it but the actual bottleneck (city streets) can't actually handle more traffic. And city streets are often impractical to expand.
(This is the issue with the road I've been stuck on most, between Jerusalem and Tel Aviv. Adding a lane to it wouldn't really help, especially since the topography probably doesn't allow that in the tight sections anyway).
More generally, space need for cars scales superlinerly (somewhere between linear and quadratic) in the number of vehicles, as adding highway capacity requires multiple road expansions within the target cities, more parking, and so on. It's not (usually) strictly true that adding a lane makes traffic actively worse, but it does in general help much less than you're expected (since a congested highway is a symptom of an overburdened system and you may not be able to add capacity to the rest of it).
Near me there is a junction between two highways. You can take an offramp from one highway that merges into the other highway's left lane.
At some point somebody noticed that the ramp was wide enough to fit another lane. And it was divided. Now you had two lanes on the off/onramp that simultaneously merged into the same one lane of the target highway. The target highway was not expanded; you just had two ramp lanes merging into each other immediately as they merged into the target lane.
This caused truly incredible traffic jams.
The problem was ultimately solved by eliminating the extra lane from the ramp.
It was a pretty stunning example of lane expansion not always being a good idea, but as you note, I don't think this is the kind of thing people are generally talking about.
Edit: actually, it's probably not equivalent, just related
Check out Steve Mould's video on a related topic (additional road, not additional lane).
https://youtu.be/Cg73j3QYRJc?si=-D8VBcbZGNmjZOXK
As i understand it, yes, individual participants can benefit in the short run, but the theory says that total demand goes up - people move further into the suburbs, stop using public transport etc. until eventually, the new equilibrium is the old equilibrium except with an extra lane.
The usual term is "induced demand". After equilibrium is achieved, it need not be the case that anyone's commute got better, but it must still be the case that someone's *life* got better.
That person might be someone who moved to the region after the lanes were expanded, making the question of whether you should vote for local lane expansion potentially hazy.
So apparently in New York State, where I live, a huge amoint of elementary-level education, including assessments, is done using Chromebooks. I have children entering elementary school soon and, bluntly, I am convinced that this is a *terrible* idea and want them to be using physical books and paper for everything short of writing papers at the high school level. (Where they should still be using physical textbooks).
Does anyone have experience with or awareness of public school districts in New York or neighboring states that *don’t* use laptops at the elementary level?
Try the shtetls? If nothing else, there you have a religious argument for "why paper textbooks are good" -- the kids can read on Shabbos.
I saw this question on twitter and thought it was interesting:
https://x.com/GrantSlatton/status/1944089586084311198
"Suppose you took 10,000 optimally selected people and dumped them into a region that had adequate forests, fields, and mountains for mining iron and coal, all in a 100 mile radius
They start with only 1 season of food supplies
How fast could they bootstrap to tech circa 1900?"
Answers in the comments ranged from 2 years to 10k years, which both somehow seem plausible to me. It seems like thousands of technically capable people should be able to get a lot done in a few years, but then again there must have been some major bottle necks to the process that it took thousands of years irl.
A lot of answers got sidetracked by whether they'd just die from hunger or exposure, so say they start with seeds to plant and the climate is mild enough that the community is guaranteed to survive if they dedicate at least 90% of their man hours to farming and shelter in the beginning and become more efficient as their tech advances from there.
The coal/iron/steel is the easy part, honestly. Those are "big bulk things" that you don't have to get "precisely right."
Try cheesemaking, or distilling, or any number of the "practical chemistry" bits of technology you'd need to get to have "technology circa 1900" Not to mention, but you'd need glass for that. Also, 1900 is early enough for galoshes, so you'd need rubber.
And we haven't even touched clothing (textile mills) or "the great big bombs" that are steam engines. Waterwheels and Windmills seem easy enough to put together, though, so "practical automation" is doable.
1. You select 10000 people who are ideologically aligned, mentally stable, and knowledgable on the problems at hand (agriculture, manufacturing, etc. maybe skip medicine since this is a speedrun). This is a tough optimisation problem, so you have to sacrifice some knowledge on e.g. science in favour of people who are a bit less antisocial.
2. Prefer people who are better at establishing strong institutions and educating others, so that the next generation can learn how to do the labour.
3. 10000 people won't be enough to reach 1900 without kids, there will be a bottleneck in mining, science, or manufacturing somewhere.
So I think probably they could get it done in 120 years, or about 6 generations of kids.
Someone started building a real-life tech tree, which may yield some relevant data to the question:
https://www.historicaltechtree.com/
If they were optimally selected, it wouldn’t take especially long before you could have basic ironworking producing tools and weapons. Maybe less than a year. From there it’s a pretty straight shot upward to steam power, but I think the precise ironworking and how difficult steel production is would involve a lot of trial and error. I’d be really surprised if anyone had both enough iron, and enough precision to produce an engine within a decade, but if they’re optimally selected… who knows?
The main problem would be food and access to resources. While it doesn’t matter if it takes a year or two to start making more complex iron tools, it does matter if you can or can’t get enough food production set up for 10,000 people (starting with no tools). I’d say the answer is likely no. Farming is very intensive work, 10,000 people is too much to be in one place, so you’ll have different groups at different levels of food production capability.
When you’re starving you’ll necessarily steal food to survive, and when you’re almost starving you’ll fight to make sure no one steals your food. Then you have social organization, centralization of food storage, tribes, and that itself if it becomes intense enough, might invalidate any technological progress. While I’m convinced optimally selected modern people could survive, I doubt their children could receive enough knowledge orally to do almost anything that their parents weren’t already capable of. If it’s just a vague idea of “You can mine coal, the black rock, out of the ground and burn it” and “if you produce iron in this specific way it’s stronger” that wouldn’t be enough to restart civilization, and might decay over more generations to exactly as much knowledge is useful to live in such an environment.
Just to clarify the question: are the "optimally selected people" able to overcome mutual conflicts, and resist burning most of the resources in various zero-sum games, such as "who is the boss" and "who can have sex with whom"?
It's not originally my scenario, but I think it's more interesting to assume they're very dedicated to the mission and able to cooperate.
Right, I think the “10,000 optimally selected people” would have to be, something like deeply faithful practicing Roman Catholics, with the right genetic distribution so only 1% had 99%ile iq and everyone was willing to be super obedient to the religious authority structure.
I... don't really think you need smart people for this. You need "butt-basic idiots" -- the folks who have learned how to do things the hard way, because it's fun. Smart people are good at solving "problems you've never seen before." But i don't think we're at that.
Even motivated midwits would work -- remember, with 10,000 people, you can have an "expert" on every link of the chain. Exposure shouldn't be an issue, you can always grab 10 Amish (or "Ron Swanson-esque" survivalists) who can order the rest of the folks about.
1900... Steam and locomotives. Mills, and upgraded farming.
Yeah, I think you may be right. You want reasonably intelligent persons selected much more for their combination of existing knowledge plus work ethic and pro-social norms. Though you may want one person tasked with getting this stuff written down lest the knowledge die off. I don’t think you could do this in a single human lifespan, you’d need cycles to bootstrap in the technology. It’s not just about knowing how but having the necessary tools, which have to go through iterations. Can’t go straight from raw iron to a lathe capable of micron-level precision. If your super knowledge persons were in their 20’s…. maybe you could pull it off in 40 years.
Of course you'd need cycles, but you don't need micron-level precision to get to 1900's technology. I think the iron/steel/coal troika is doable within ... maybe 5 years? Most of the steps there don't involve "precision" so much as "and now we make pig iron." (Pulling this from "that polish guy's science fiction," so I could be wrong ... but I don't think I am).
But... then you have "water purification" (aka BEER) Either you have the yeasts, or you don't (and, I believe, unlike bread, you need specific strains).
Distilling (also a pre-1900 thing I think) would require glassware.
And we have Yogurt! And Cheese! (If you don't think these are technology, boy, howdy). And sausage.
Basic looms/spinning are pretty easy to make. Not sure how they scale into "industrial looms and felting and..."
Reposting this comment from
FluffyBuffalo here so people aren’t misled. Some amyloid theories are probably wrong, but some are almost certainly not. There is good evidence that amyloids are in fact involved in Alzheimer’s in some way, and that anti-amyloid therapies help moderately.
“I think it's important to be careful about what "subscribing to the amyloid hypothesis" really means. The evidence is very strong that amyloids play a significant role in the disease - genetic variants like APOE4 that influence amyloid accumulation are strong risk factors for AD; you see characteristic biomarker levels (in particular Ab42/Ab40 ratio in cerebro-spinal fluid) in AD patients that you don't get in other neurodegenerative diseases; you find the characteristic plaques in deceased patients, etc. That part is sound as far as I can tell, and pointing at a misguided study or two doesn't change that.
The question what exactly the amyloids DO, and whether the buildup uf the plaques is the whole story or just a part, is not so clear, and my impression is that researchers are open to the possibility that there's more going on. The new hot topic seems to be CAA: Cerebral Amyloid Angiopathy - it turns out that amyloids also accumulate in blood vessels, weakening the walls and leading to micro-bleeding. (Apparently, this got more attention when it was found that brain bleeding is a not-so-rare side effect of lecanemab.)”
Devansh on the value of reading Hannah Arendt, totalitarianism scholar.
https://artificialintelligencemadesimple.substack.com/p/why-you-should-read-hannah-arendt-5bc
Stupid question time: I've been trying to understand how exactly Heritability is defined, but the explanations I've found don't make sense.
Wikipedia tells me that, in simple cases, it can be computed as Variance(Genes)/Variance(Phenotype) where "Variance" means the expected squared distance from the expected value.
But this doesn't depend on any relationship between the two at all! By this explanation "Heritability" is just a measure of whether a set of genes is similarly spread out as the phenotype, not whether those genes explain anything.
Also, variance is measured in units of "distance in the underlying distribution" squared, so you can only take a ratio like this is the distances are comeasurable in some way, and I don't see how "polygenic score" and e.g. "has schizophrenia" can be comeasurable.
Stats in science are often bad, but I have trouble believing they're *that* bad. Can someone try and explain how this actually work
> Also, variance is measured in units of "distance in the underlying distribution" squared, so you can only take a ratio like this is the distances are comeasurable in some way
Variance infamously suffers from the problem that the same distribution at different scales has different variance. For example, If you measure everyone's height in inches, and if you measure everyone's height in centimeters, the height-in-centimeters distribution has more variance even though you measured the same set of people and they were all the same height both times.
That's just a rounding issue isn't it? If you measure height in rounded kilometers then everyone is 0km tall, no variance at all.
I found this explanation good:
https://dynomight.net/heritability/
The Alzheimer mouse review on Friday reminded me of Thrifty Printing Inc, whose business model of online photo-processing for cornershops didn't take off, apparently, so naturally the company pivoted, to drug development for such diseases as Alzheimer's.
But Anavex Life Sciences, that's the new name, doesn't seem much respected by the stockmarket. Was the pivot too ambitious? Or perhaps Anavex is seen as unserious by the professionals after hiring a runway model with no business experience as Director of Business Development and Investor Relations (Nell Rebowe). Or it's the hairstyle of the CEO (Dr. Christopher Missling). Or his presentation skills (that last one actually has some merit).
A recent Economist article, titled "The Alzheimer's drug pipeline is healthier than you might think", did not even bother to mention that Anavex has a pill under evaluation by the European Medicines Agency. Yet what will happen to the stock if it is approved in six months or so?
On the other hand, withdrawal or rejection would most likely send the stock crashing (not "certainly" because another drug's phase 2 trial for schizophrenia will report soon). To me, Anavex's pill appears disease-modifying (though not a cure) and safe (unlike the anti-amyloid drugs), but certain language in the company's latest 10-Q filing does strike me as ominous. Also, shareholders are up against high short interest, including Martin Shkreli, who has called Anavex "another retail trap" (referring to Cassava Sciences, popular with retail investors, whose Alzheimer's drug failed its phase 3 trial; that company's history involves fraud allegations). And whose strategy of shorting Alzheimer biotechs just scored another victory when INmune Bio's phase 2 trial failed to convince.
Anavex's pill is meant to stimulate autophagy and is not related to the amyloid hypothesis, as far as I know (I don't know much, but was still tempted to complain on Friday when someone called Leqembi a "proof of principle" for the hypothesis --- if people try for decades to develop drugs on the basis of a paradigm, wouldn't you expect them, even if they are mistaken, to come up with something eventually that shows an accidental effect, via some other mechanism?).
Two things in recent Irish news coverage, and the fun one first:
(1) "Go fly a kite!" may be good advice if you want to generate energy
https://www.rte.ie/news/ireland/2025/0714/1523349-kite-flying-electricity/
"A project in Co Mayo is generating renewable electricity through the flying of kites, which its operator has described as a potential "game changer" in the wind energy sector.
...The site, which is the first designated airborne wind energy test site in the world is being operated by Kitepower, a zero emissions energy solutions spin-off from Delft University in the Netherlands.
Kitepower's system employs a yo-yo effect, where a kite, measuring 60sq/m is flown at altitudes of up to 425m attached to a rope that is wound around a drum - which itself is connected to a ground-based generator.
The kites can generate 2.5 to 4 tonnes of force on the tether.
The pull from this force then rotates the ground-based drum at a high speed.
This rotation then generates electricity that can be stored in a battery system for deployment wherever and whenever it is needed.
The kites are flown using the knowledge and skills of kitesurfing professionals, combined with a highly specialised computerised GPS-guided steering system.
They fly upwards repeatedly in a figure of eight pattern for periods of 45 seconds.
The flight pattern is important because it forces the kites to behave like sails on a boat, maximising the pull of the wind to increase speed so electricity can be generated.
After 45 seconds, the kites are levelled up so that the pull from the wind is momentarily minimised.
This enables the tether to be wound back in, using only a fraction of the electricity generated when it was being spun out.
The result is a net gain in renewable power at the simple cost of flying a kite.
Then the cycle is repeated, again and again, potentially for hours on end."
(2) Well, looks like AI is *already* taking er jerbs (at least if you're a graduate in finance):
https://www.rte.ie/news/analysis-and-comment/2025/0713/1523109-ai-job-losses-ireland/
"Also this week, the latest 'Employment Monitor' from recruitment firm Morgan McKinley Ireland found notable reductions in graduate hiring by major firms in the accountancy and finance sectors because of the adoption of AI.
And on Thursday, AIB announced a major AI rollout for staff in conjunction with Microsoft Ireland, sparking concerns from trade unions.
Morgan McKinley Ireland's Employment Monitor for the second quarter of the year was published on Thursday.
The recruitment firm said that the standout development of the quarter was the significant impact of AI and automation, particularly within the accountancy and finance sectors.
"The notable reduction in graduate hiring by major firms, driven by AI capabilities, highlights potential challenges ahead," the report found.
"Companies are increasingly leveraging AI capabilities to automate routine tasks such as accounts payable, accounts receivable, credit control, and payroll."
...[Allied Irish Bank] announced a new artificial intelligence rollout for staff in conjunction with Microsoft Ireland on Thursday.
The bank said the new tools will reduce time spent on repetitive tasks, freeing up employees for higher-value work.
The plan will involve the widespread deployment of Microsoft 365 Copilot, embedding AI into everyday tools like Outlook, Word, Excel, Teams, and PowerPoint.
...Last month, the Chief Executive of AIB Colin Hunt took part in a panel discussion at a Bloomberg event in Dublin.
Asked what impact AI will have on staffing numbers at the bank over the next five years, Mr Hunt said it may lead to a small reduction in net headcount.
"I do think that there are certain manual processes that we do now that will be done by AI in the future, and probably net headcount will be broadly stable with a slight downward bias maybe," Mr Hunt said."
Who knew that the real knife-ears were the ones we created along the way! (See "The Rings of Power": "Elf ships on our shore; Elf workers taking your trades. Workers who don't sleep, don't tire, don't age. I say, the Queen's either blind or an Elf lover, just like her father.")
Is this not "fourth power of wind speed"? Because that's the issue, if you want "renewable energy" to stop generating more greenhouse gas (via the necessity of "keeping the grid stable" through spikes and dips -- the natural gas required for this is significantly more than "if we didn't use wind power at all.")
That story about graduates not finding jobs has been repeated across British papers for the last few months. If it is a trend then AI is not boding well.
It's mostly worrying about what all these graduates were being hired to do. I can see why graphic design would be toast, but the only other things AI can seeming do are "google this and write about it," "take this information and reformat it" and "spitball arbitrarily about this topic."
What do you guys think about my sick invention?
https://x.com/SailAcross_/status/1858697261888192657
Pitfalls, criticism welcome.
With "invention", do you mean you built a working prototype, or modeled one in some detail?
I think some version of this will work, but I suspect it will be super weak and not very durable.
Sorry to be laughing about the tooth claims, because vision problems are no joke, but yeah. There's always something.
It's very unlikely the fancy tooth treatment is to blame, but it's not impossible, and it's only when you have real users out in the wild, as it were, that these one-in-a-million things crop up. Now we know the problems Big Pharma and the FDA face!
Reading about the twin studies... twins are similar not only because of genes but also because of a shared initial history in the womb. And not all twins are the same, it depends at which development stage they split in two different embryos (see "mirror twins")
If you like road trips and you still believe in the American Dream, join us in April 2026 for The Great Route 66 Centennial Convergence. We're driving from Chicago to LA over the course of a few weeks. There will be never-before-seen mysteries, side quests, and prizes. It costs nothing. Find more details on Facebook, Instagram, and Youtube.
Interviewees and video co-hosts wanted. We won't just be talking about hot rods and midcentury architecture. We'll be discussing the American Dream, capitalism, and human progress. It might even be a paid gig if you're charismatic, pleasing to look at, or slightly famous.
I'm a fresh grad about to start a Data Science-esque role for a big UK insurer. Never having worked in corporate environment, I'd like to ask the ACX community for sage advice on questions such as:
How can I learn the most?
How can I negotiate my salary throughout?
How do I handle 'office politics'?
or any questions someone in my position should think about.
Thanks in advance for your wisdom!
If your office politics involve poisoning each other, just quit. If poisoning each other has been banned as being "not within the spirit of office politics" -- just quit.
I'm confused. They should quit if there's poisoning, and also if the poisoning is banned?
Yes. Having enough poisonings that "it's been banned" implies that your workplace has gone beyond "toxic" into some strange new country.
Ah Makes sense, cheers
This took me a long time to learn via experience, as it's not necessarily intuitive to rational people.
Aside from complying with things like critical regulatory requirements which might get you arrested for violating, you really only have one real job at work:
Make your boss [1] happy.
That is often very difficult for rational people leaving the merit-based school system and entering the social hierarchy of the workplace to fully grasp. You may have felt that school was your "job" and that your teachers were your "bosses," but this wasn't actually the case; as a student, you were a *customer buying a product,* which meant you had an absolute, enforceable *right* to receive clear expectations and an objective measure of your performance *as part of the product you were purchasing.*
In the case of your employer, they are the customer purchasing your time and attention... in the way they want you to provide those things Your paycheck and career aren't going to be based on you scrupulously, objectively performing all the duties of your job description or doing what's best for the company as a whole. Practically speaking, there's no way to really track that, and people who are determined not to see proof of your good work won't look at the evidence anyway. The people making decisions about your role at the company are going to be making them based on how they *FEEL* about you, which means your first priority needs to be making your boss *FEEL* good about you.
Keep.
Your Boss.
Happy.
So if your boss is very attached to a dumb process that you just KNOW could be vastly improved, your *real* job - the one which may get you promoted or saved from layoffs - is to make them *FEEL* good by surrendering to their dumb process. If your boss is overly-emotional and sensitive to criticism/rejection and perceive suggestions/corrections/warnings as personal attacks, your *real* job at all times is to make them *FEEL* good by avoiding triggering their negative emotions, even if that means allowing them to damage the company.
And remember that your grandboss and above haven't fixed or fired your crappy boss because there's something about your boss which makes *them* happy enough to want to keep your boss around. Don't assume *they* want to fix problems for the good of the company, either.
Surrender to the stupidity. That's what your customer wants, and that's your job now.
[1] "Boss" doesn't necessarily refer to your direct supervisor (although it probably will), but rather the persons/people above you who have the power to promote or fire you, including your grandboss, etc.
All of that is true, and let me add to it.
You should also keep in mind what your trajectory would ideally be during the next few years. Ideally you'll go from teachable youngster who is easy to work with and only needs to be told anything once, to junior colleague who can be trusted with small responsibilities, to actual peer who knows the ropes and can be trusted with substantial (and even unknown) challenges.
But right now you are _just_ the teachable youngster, and you need to embrace that role. Be conspicously eager to learn and respectful of advice from your older colleages. Do anything you are told to do, or even that you are suggested you might do, unless it is literally illegal. And if you haven't been given anything specific to do and everyone seems really busy, at the very least don't get in the way.
Office politics doesn't need to be tricky, especially as a junior; it's just the effect of companies being staffed with fallible flesh-and-blood human beings. Focus on making sure that people know you, and they like you, and allow that to be more important than being right all the time, and you'll probably do fine.
A few things I've learned from experience:
Similer to one of Wasteland Firebird's points, long term you'll do better if you hop sideways and upward between employers as opposed to spending years and years with the same one. This is because with each job change you should get a step up in position and salary, as well as a wider experience of the industry (and you don't necessarily have to stay with insurance companies - data science skills are fairly transferrable). But of course you shouldn't overdo the job hopping frequency, or potential employers will wonder why you can't stay in any one place for long.
Also, people tend to be better at office politics in inverse proportion to their technical ability: A weak performer needs to be crafty at office politics to offset their technical shortcomings, and conversely a tech guru, recognised as such, can often afford to disdain the politics.
When starting a new job, you should be suspicious of a colleague who seems too solicitous of your welfare when it isn't their assigned duty. It's similar to prison, although I've never had any experience of that, in that shortly after you arrive some apparently helpful fellow convict will sidle up to you and offer to provide anything you need and show you the ropes and so on. But ultimately it's for their benefit not yours!
When a company downsizes, the bean counters take zero account of the competence or otherwise of those being laid off.
> When starting a new job, you should be suspicious of a colleague who seems too solicitous of your welfare when it isn't their assigned duty.
I often do that
> But ultimately it's for their benefit not yours!
My "benefit" is that it makes the new guy feel welcome. Also talking to people is more fun than actual work. Also I guess it also makes my own boss and grand-boss happy when the new people come to me with their problems and not to them (boss and grand-boss).
What exactly are the risks of such an overly helpful colleague?
Also, is the last paragraph complete? They don't look at the competence, so what do they look at?
Learning: Most of what I've learned, I learned inevitably on the job, or from switching jobs a lot. I haven't done much work in my spare time to keep up with the industry. It seems kind of silly to do that because you never really know what's going to be expected of you in your next job anyway.
Salary: The best way to get raises is just to take whatever you get, at first, and always be looking around for something that pays more. You don't even necessarily have to take the job that pays more if you don't want, you can get an offer and bring it back to show it to your current employer, and ask for a raise then. If you feel weird asking for a raise, don't ask for one. Just tell them you're leaving, explain why, and let them get the idea to make you a counteroffer. Eventually, you'll get to where you earn and/or have "enough" money. Figure out what that means for yourself. Then, the good part comes. Keep changing jobs, but this time, only change into jobs that make you happier! Sooner or later you'll get into a job where you simply can't realistically expect to find anything better. That's where I am now. I am hoping this will be the last job I ever have to take seriously.
Office politics: Even as a nerdy person who used to have no social skills, and who is probably, like many of us here, what might probably be diagnosed as a "person on the autism spectrum," I have to say, I actually really enjoy office politics. "Soft power" is a big thing. A lot of times you don't have any authority to actually make people do things, but you can get people to do things anyway. Ideally, you can figure out how to express the thing you want to do in terms people will appreciate, and then they'll join you in your quest. That requires learning real empathy, which is a difficult thing in itself. But when that doesn't work, sometimes there are little clever tricks you can pull that let you get your way. For example, there's a project I've been wanting to do, but no one is allotting my team any time to do it. So instead, I've looked around for other similar projects that were allotted time. I've treated those projects like they are higher priority than they actually are, and seen them through to completion. After a couple years, my dream project is 80% complete, and we haven't even technically worked on it at all yet.
Another fun trick I've used multiple times: You need to do something. You want to do it in a simple way. People with power over the project insist that the project must be done in a very complicated way that will take far too much time. Quietly complete the project on time, in the simple way. Get it as far as you can take it, so all they have to do is say, "Fine, good enough, ship it." The choice you've now forced on them is to do it your way, or to deliver it late.
Me and Lord Hobo Boom Sauce have been ruminating tonight about having an ACX site where we put up photos of ourselves as kids. Seems there would be very little risk of doxxing ourselves, so long as it's pix of us as *young* kids. And it would nudge us all a bit away from the godawful online illusion that the entity we are talking to via texts consist of a funky little fake name and a cluster of, like, 5 opinions. Anyone like this idea. or have I just gone down 11 notches in everybody's respect just for suggesting it?
Seems like that would summon pedophiles.
I'm generally very dubious about putting up any kind of identifiable information because someone out there will pick up on it and try to identify real world you. I know that sounds paranoid, and I don't really have much to lose if someone figures out "oh Deiseach is really so-and-so" (apart from my jealously-guarded privacy) but for people who *do* have something to lose, I would be way more careful.
People *have* gone after Scott and those associated with him, see Sneerclub and the infamous Cade Metz story. It's not beyond the bounds of possibility that someone with a grudge about rationalists/Scott/SSC/ACX/TheMotte/LessWrong/you name it would latch on to anything like photos and try to work out who you are in real life and then send nasty little emails to your employer about "were you aware that Eremolalos is involved with right-wing fascists and racists?" (the HBD/IQ stuff is catnip to people with axes to grind).
I'm not saying "don't do it", just "be very very sure about the level of security".
I'm mildly intrigued, but I don't think I have any pictures of myself from a young age. Presumably my parents had some, but I think my mothers' scant few pictures of me as a little one went onto the "toss" pile when I was sorting her assets after her passing. I don't mind sharing more recent pictures, though, from when I lived somewhere else entirely (an attitude that can probably be surmised from my avatar).
An essay on the transformation of academic tutors into Turing cops, marking into an imitation game, and Al generated homework into the trigger for the technological singularity:
https://open.substack.com/pub/vincentl3/p/a-modest-software-patch-for-preventing?r=b9rct&utm_medium=ios
Maybe we'll see a return to the old-fashioned system of verbal examinations, whereby the student links via Zoom to an AI interviewer, which then fires questions at the student, who is required to make immediate extempore replies without referring to notes.
Current fashionable buzzword in universities for this kind of thing is “secure assessment,” betraying the academy’s very own AI control problem.
Hey Vincent, I read it, and have a non-ironic suggestion:
-It is probably possible to train an AI to be an excellent judge whether there is AI contribution to an essay, and how much, and whether it's in the main idea, the overall structure, or individual sentences or paragraphs. AI's -- which overall fucking suck, IMHO -- are pattern-identification geniuses. They are better than professionals at identifying melanoma, various retinal diseases, etc, from scans. You might need to hire somebodydo a bit of extra training to improve an LLM's ability to recognize these things, but I'm pretty sure it would be possible.
-OK, so then tell the students that if they submit something that scores more than a certain low percent on AI content, you will not read the piece and it will be graded by AI. I actually think that would discourage people quite a bit from turning in AI-contaminated work. I had a professor who would grade late essays, but did not put any comments on them. I really wanted to get comments, so never turned in things late to that prof, even though it would not have harmed my grade, and I was in general actually fairly bad about turning in papers late.
There are existing AI detectors that check for this, but the current models often give false negatives or wildly divergent responses. These could be improved of course, so I do agree with your suggestion. I guess the loophole then becomes that students could ask their AIs to write in ways that avoid the telltale signs. So we would still get an AI arms race of sorts.
> It is probably possible to train an AI to be an excellent judge whether there is AI contribution to an essay,
I just tried that by asking chatGPT about cats and then in a different instance asking it the probability of the cat text being AI. It was fairly certain it was.