The empirical track record is that the top 3 AI research labs (Anthropic, DeepMind, and OpenAI) were all started by people worried that AI would be unsafe, who then went on to design and implement a bunch of unsafe AIs.
Closed Limelike Curves
Karma: 138
[Question] Improving Giving with Nudges
I suppose the long-term future doesn’t exist because AGI got to it first.
Actually, maybe I should clarify this. This is standard practice when you hire a decent statistician. We’ve known this since like… the 1940s, maybe?
But a lot of organizations and clinical trials don’t do this because they don’t consult with a statistician early enough. I’ve had people come to me and say “hey, here’s a pile of data, can you calculate a p-value?” too many times to count. Yes, I calculated a p-value, it’s like 0.06, and if you’d come to me at the start of the experiment we could’ve avoided the million-dollar boondoggle
that you just created.
You’re completely correct! However, it’s worth noting this is standard practice (when the treatment makes up most of the cost, which it usually doesn’t). Most statisticians will be able to tell you about this.
So I think I have two comments:
It’s actually pretty neat you figured this out by yourself, and shows you have a decent intuition for the subject.
However, if you’re a researcher at any kind of research institution, and you run or design RCTs, this suggests an organizational problem. You’re reinventing the wheel, and need to consult with a statistician. It’s very, very difficult to do good research without a statistician, no matter how clever you are. (If you’d like, I’m happy to help if you send me a DM.)
[Question] Research on Nudging People to Increase Donations to Effective Charities
There’s an EA cause area in this, but it’s not tooth-related. Our jaws are perfectly fine. The problem is bad incentives in the healthcare system; dentists get paid to take wisdom teeth out, not to leave them in, so about 90% of wisdom tooth extractions are unnecessary.
The mortality rate of anesthesia is not quite negligible; a couple hundred people have died because of anesthesia during wisdom tooth removal, and there’s also the risk of infection.
With regards to protein: studies consistently show getting enough protein to prevent sarcopenia or other diseases is not especially difficult, and in fact most people get substantially more than the recommended daily allowances. Also important is that consuming animal protein is positively correlated with all-cause mortality, while plant protein is negatively correlated. From what I can tell, though, this is probably because fiber is confounding the results.
Moving on from observational studies and to randomized controlled trials, it’s unclear whether protein even has a meaningful effect on muscle strength, let alone something like lifespan, with some studies finding little or no effect from supplementation. It’s really not clear protein would help with sarcopenia from the studies, but we’d honestly expect null results from theory alone. We know from surveys that the limiting factor in most people’s muscle strength is physical activity, not protein. Most people get more than enough protein, and excess protein consumption turns into fat, not muscle. For most people, protein supplementation just can’t help because they aren’t moving enough.
The primary dietary determinants of health seem instead are mostly:
Polyunsaturated fatty acids (found in fish and vegetable oils)
Fiber (typically supplemented with psyllium husk)
I’m busy, so I’ll stick to looking at the evidence base we’ve acquired just on fiber supplementation. Here’s a bunch of insanely overpowered observational studies with sample sizes stretching into the hundreds of thousands, all showing reductions in all-cause mortality. Meta-analyses of RCTs? We’ve got those in spades.
There is a reason every doctor’s dietary advice is exactly the same: “Replace refined with whole grains, avoid saturated fats and replace them with unsaturated fats.” The reason is because we have million-person studies showing these interventions reduce heart disease. Note that heart disease is by far the leading cause of death among the elderly; sarcopenia doesn’t really kill, although it does make life substantially worse. And importantly, both of these are substantially more common in plants than in animal foods, which offers a pretty good explanation for why vegetarians seem to live longer, even after controlling for confounders. And also for why vegetarian diets seem to improve health markers in RCTs.
Thanks!
“AI is powerful and uncontrollable and could kill all of humanity, like seriously” is not a complicated message.
Anything longer than 8 morphemes is probably not going to survive Twitter or CNN getting their hands on it. I like the original version (“Literally everyone will die”) better.
In what sense? The problem of potential helium scarcity has been (effectively) solved in the last few years by just looking for more helium deposits.
In a sense, we are “running out” (because there’s only a finite supply of Helium), we’re just running out very, very slowly.
Could be marketed as a medication for hypersomnia, narcolepsy, chronic fatigue, and ADHD.
With regard to differing forms of , the article you linked notes:
Despite these minor differences in processing, each of these forms is probably equally effective when equal doses are given.
Which is understating things a bit: All the minor variants will be identical after dissolving in the body—anyhydrous creatine will become hydrated, and particle size is 0 after the creatine has dissolved. The only (minor) difference is that micronized creatine will dissolve slightly faster if you mix it with a drink, because of the smaller particle size. This can be mildly convenient, but I’ve never been bothered by the monohydrate being too slow to dissolve (it’s almost instant for both forms).
With respect to the impacts on cognitive performance, Gwern completed an in-depth look here. Most important is the section on publication bias; the authors of these papers failed to replicate their results several times, but these failures were never published. Gwern concludes (and I agree) that creatine almost certainly doesn’t affect intelligence in nonvegetarians, and has at best
The study you quote at the end about vegetarianism has a sample size in the low dozens, and I wouldn’t put much stock in it. The large meta-analyses of this all find better life expectancy and health markers in vegetarians/vegans. The one important exception is B12 (for which the study you cite finds a deficiency in vegans). Luckily, th
is can easily be fixed by supplements.
If you’re wondering whether the above studies leave some room for including fish in your diet: My guess is yes, pescetarianism is likely to be slightly healthier, because of the aforementioned effects of polyunsaturated fats (including omega-3 fatty acids). This is why doctors also recommend the Mediterranean diet a lot, although it’s not clear how much of the effect comes from fish and how much from vegetables. Omega-3 supplements definitely help.
The quality of research is already there: See my comment here.
I’m not aware of any newer studies on creatine after the original one that spawned this belief; I will point out the original studies were likely aggressively p-hacked, showed impossible effect sizes, and make very little sense because brain creatine levels are not substantially different between vegetarians and vegans.
Still, it’s not completely out of the question there’s a real effect. We still want to avoid the fallacy of the one-sided bet. It’s definitely possible that vegetarianism causes a drop in . But given vegetarianism seems to lower blood pressure, LDL cholesterol, and obesity rates, I’d be willing to put money on the exact opposite. A healthier kid is a smarter kid, and vegetarianism seems to have positive effects on health in general.
If you’re still skeptical, over-the-counter creatine is extremely cheap (and has benefits for physical performance anyways). Trying to get rid of LDL cholesterol, on the other hand, is pretty hard.
>The egoistic motivation for donating is highly scope insensitive – giving away $500 feels roughly as good as giving away $50,000. I haven’t found any academic evidence on this, but it’s been robustly true in my experience.
There’s academic evidence and it disagrees; the amount of donations matters. (It’s nonlinear, but so is the effect of spending on anything.) I suppose you could probably increase it by giving smaller amounts more frequently—having a tiny notification on your computer that goes “Congrats on donating $10!” or whatever.
I’d rephrase that, maybe. Whether we’re asking people to make a change doesn’t matter—that’s assigning an unearned privilege to old ideas. The reason we have to justify this claim is because it’s a claim. If you’re saying “I know vegetarianism is healthier,” you should be able to explain how you know that.
…is what you tell yourself before you get exposed for committing massive fraud, costing far more billions than you ended up with.
If SBF did commit fraud, it looks like he did it to keep Alameda from going bankrupt. If that’s the case, he ended up destroying billions of dollars of potential donations from FTX. “Take a risk that might let you earn a billion dollars illegally” and “Make 0 dollars” are not your only options here! You could have taken not-illegal risks that might have won big instead. Those tend to have higher EV.