Anyone who's spent substantial time using one of the state of the art AIs today can attest that this statement is clueless and ridiculous. Makes me wonder about the fitness of Yann Lecun to lead a major research lab. It does appear to me 1) he's blind to existing data 2) he doesn't know how to collect data (namely spend 10-30 mins with a SoTA AI) 3) He's blinded by his theoretical biases.
AI is clearly in the right track regarding performance and the speed of progress. Is it human level in every task? Not really. It clearly has room to grow. Is it super human in many tasks? I think so. Just learn how to use it. Peace.
All computers are "superhuman in many tasks." But no computer has a cat's planning abilities or common sense, and it will be many decades before we get there.
Diet and exercise are universally failed interventions to stop major diseases (CVD, cancer...). People who have a vested interest in them promote them (e.g. exercise specialists, keto experts, vegan gurus, your favorite TV Dr, harvard nutrition dept whose livelihood depends on the effectiveness of "optimal diet" etc), but the results in intervention trials have been extremely disappointing.
1) The best diet studies where people were asked to follow Mediterranean diet only managed to lower CVD and/or mortality by 10s of % points. This is a very poor outcome. At best adds a few years to someone's lifespan. At best. The average Mediterranean dieter fares a lot worse.
https://pubmed.ncbi.nlm.nih.gov/26528631/https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4368053
Don't get me started about Keto, Vegan etc. Those have even less convincing evidence regarding hard outcomes (e.g. MACE, Cancer etc) though our opinion might change if they ever get tested in large scale.
2) Exercise is even worse. It has virtually zero effect on CVD outcomes, mortality outcomes in trials where participants were asked to change their behaviors. "Exercise did not reduce all-cause mortality and incident CVD in older adults or in people with chronic conditions, based on RCTs comprising ∼50,000 participants"
https://scholar.google.com/scholar?cluster=10512580439138189...
It's a myth that diet and exercise will save you from Cancer or CVD. There's virtually zero evidence for that, when you look at things from intervention point of view -- intervention evidence is the only evidence acceptable in science. The rest is pseudo-science or proto-science.
Btw, I believe weight loss is the best thing you can do to live longer (that is if you're overweight or obese). But diets still universally fail. E.g. bariatric surgery probably works, maybe drugs will be shown to work (Tirzepatide/Semaglutide I'm looking at you). A publication on how 90% of obese dieters fail after 1-2 yrs on the same diet: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5764193/
> Exercise is even worse. It has virtually zero effect on CVD outcomes, mortality outcomes in trials where participants were asked to change their behaviors.
This is simply not true and you jump through some curious hoops to make it look credible. The paragraph you cite basically says "if you start after 60 it's too late". Is then proceeds to add that a few weeks of exercise are not enough, and a few months of follow-up don't catch measurable differences.
But if you actually read the full study that you cite and check the papers it reviews, you'll see that multi-year interventions at middle age with multi-year follow-ups are extremely effective.
> During the 30-year follow-up, compared with control, the combined intervention group had a median delay in diabetes onset of 3·96 years (95% CI 1·25 to 6·67; p=0·0042), fewer cardiovascular disease events (hazard ratio 0·74, 95% CI 0·59–0·92; p=0·0060), a lower incidence of microvascular complications (0·65, 0·45–0·95; p=0·025), fewer cardiovascular disease deaths (0·67, 0·48–0·94; p=0·022), fewer all-cause deaths (0·74, 0·61–0·89; p=0·0015), and an average increase in life expectancy of 1·44 years (95% CI 0·20–2·68; p=0·023).
This is biased reading at its worst. I was simply stating what the authors concluded based on their analysis of 10-20 studies (not just one). This is what they say in the abstract: "The results show that exercise does not reduce all-cause mortality and incident CVD in older adults or in people with chronic conditions, based on RCTs comprising ∼50,000 participants."
One study with 500 people doesn't prove a point.
Of the top of my head I know at least one other study where mid-life lifestyle intervention didn't reduce mortality (it slightly increased it in a statistically non-significant manner). 21 yrs follow up. ~3000 people.
https://www.ahajournals.org/doi/10.1161/CIRCULATIONAHA.121.0...
Diets don’t fail, people fail to follow diets. Which is probably fine in the long run, because they’re always there for people who want them. There’s probably never going to be a wonder drug for obesity though. But Anavar or Trenbolone would work better than Semaglutide, and the side effects likely aren’t that much worse (all of those options are still bad ideas).
That's like saying, "Exhortations to developers not to write bugs don't fail, programmers fail to follow exhortations." Guess what? Even the most skilled and well-intentioned programmers will write bugs, despite their best efforts; and lots of programmers simply won't put out their best efforts. If your goal is simply to find someone to blame, then sure, blame the developers. But if your goal is better software, exhortations not to write bugs are simply an ineffective intervention. You need to introduce other measures, like type systems, linters, static checkers, code reviews, and testing.
Similarly, if people consistently fail to follow diets, then the current crop of diets are an ineffective intervention. If your goal is to assign blame, then sure, you can stop there and feel good that you've found someone to blame. But if your goal is to actually make the situation better, then you need to find something else.
> That's like saying, "Exhortations to developers not to write bugs don't fail, programmers fail to follow exhortations."
It’s not even the littlest bit like that. Nobody knows how to write software without bugs. Everybody knows how to lose weight. Just do some exercise and eat less calories than you burn. You don’t even have to do the exercise part, but you should do at least a little bit if you’re trying to be healthy. The only problem is that most people don’t want to commit the effort or make the sacrifices required to do that. If anybody’s trying to find something to blame, it’s the legions of obese people desperately in search of any reason for their obesity other than their own lifestyle choices. It’s a lifestyle problem, and it’s unlikely that there will ever be a pharmaceutical intervention for it that doesn’t have pretty bad side effects. But if bad side effects are acceptable for you, then magic drugs already exist for obesity. Anabolic steroids. They’ll do exactly what you’re looking for.
> But diets still universally fail. E.g. bariatric surgery probably works, maybe drugs will be shown to work (Tirzepatide/Semaglutide I'm looking at you). A publication on how 90% of obese dieters fail after 1-2 yrs on the same diet:
Buyer beware: Sensation sells (as is portrayed in this news article). Competitors are one click or one setting away. Last I tried switching to DuckDuckGo things didn't go so well -- just overall subpar experience. Bing arguably is a bit better, but that too didn't satisfy, despite the super cool chat feature (deserves kudos for that!).
This list was compiled in 2009 before I took a full time job in an algorithmic trading company, but it's still relevant :) If anything ML is more relevant than ever in trading, except perhaps Deep Neural Nets, Transformers, Large Language Models etc are the norm today.
Many of us know and follow rapamycin's story. There are some Drs in USA who're willing to prescribe it for longevity. I do hope some billionare will be willing to spend a few million on it to test it in clinical trials. Surely one of the most promising molecules out there, and very little sides when dosed intermittently.
From the article, this molecule sounds like a big hammer that you can pound cell reproduction with, affecting a wide range of systems in unpredictable ways. Whereas what we need are precision watchmaker tools and a diagram of the watch.
On thing I think is well established about aging is that it's not a single factor disease that you can "cure" by removing a well defined cause. Rather, it's a pervasive accumulation of entropy in various systems from various causes that evolution had no ability to correct, so that slowing aging down and reversing it involves a massive amount of tinkering with the watch.
The anti-aging effects of rapamycin support the opposite of "pervasive accumulation of entropy" theory - which is that aging is a distorted development program, rather than random accumulation of entropy.
The definition of "random" implies by default that lack of information or knowledge on the object in question is primarily due to cognitive (and hence measurement) limitations of its users, rather than something being 100% objectively ingrained in reality.
Only chink in that theory is rapamycin doesn't seem to have anti-aging effects per-se, rather a wide array of systemic effects, some beneficial to longevity in special circumstances. In the general case, you definitely won't be extending lifespan of the elderly by modulating down their immune system or promoting amyloid formations.
There is definitely "real" randomness, DNA replication errors are truly random because they are caused by quantum molecular effects. You can characterize them statistically and reduce their incidence, for example by controlling the diet or shielding the body from UV, but you can't measure those causes, only observe their effects. At best you could correct them, starting from a known-good-copy, as indeed cell biology already does.
It's definitely correct, and quite trivial to verify. It's possible it has been discovered before, but none of the proof compilations I've seen (e.g. cut-the-knot) has it, and the trigonometric proofs I can find involve using angle-sums (https://forumgeom.fau.edu/FG2009volume9/FG200925.pdf).
This is definitely a much more elegant proof than the angle-sum proofs.
Correctness might be there. But there are also some grandiose claims in the abstract ("“There are no trigonometric proofs, because all the fundamental formulae of trigonometry are themselves based upon the truth of the Pythagorean Theorem.”) and the argument being novel. I don't have a way to verify these two. That's what the peer review is for.
It's obviously correct (it's basic high school math; as with most proofs, the cleverness is in the construction, not the computation) and nearly obviously new or at least newly published. (There are many easily searchable collections of proof, but not everyone published their proof of a theorem already proved and published over 400 different ways)
Measuring associations is one thing. Intervening in people's lives to reduce disease and mortality with exercise is another. Here's what real science has to say about the latter:
Exercise did not reduce all-cause mortality and incident CVD in older adults or in people with chronic conditions, based on RCTs comprising ∼50,000 participants
The average sedentary HN reader, take comfort! Being a gym rat is good for you. Forcing yourself to become a gym rat from being sedentary has unclear benefits regarding hard outcomes.
>Forcing yourself to become a gym rat from being sedentary has unclear benefits regarding hard outcomes.
'hard' does a lot of work in that sentence. Try running up a few flight of stairs if you're sedentary compared to when you're in reasonable shape or try to take a fall if you're skinny compared with some more muscles on your skeleton, or just sitting in your chair for eight hours for that matter.
This is basically 'torture the data long enough until it tells you what you want' to finding an excuse to not hit the gym. Some studies on elderly people with vague exercise programs with probably no change in physique tells you nothing about middle-aged adults.
The real science? Huh? How did you get that from the study? They specifically determined that even small amounts of exercise can dramatically reduce the mortality risk of multiple things including stroke, some cancers and heart disease.
Observational studies are just observational studies. Most of them report spurious correlations. Their results cannot be taken at face value. Only through intervention studies you can demonstrate causation. Intervention studies (RCTs) --- the meta-analysis that I linked --- in this case report a null result (no effect) regarding hard end points (CVD and death).
These intervention studies aren’t measuring what you’re suggesting they are measuring.
The specific question is if elderly people starting exercise late in life received similar levels of benefits. They then look at studies lasting between 1.4 and 5 years where the median participant age was between 72 and 78. The exercise group did have lower mortality but it wasn’t considered statistically significant.
Unless you’re 70+ and only planning on exercising for a few years they don’t say much about what you should be doing. Especially as by this age a significant percentage of the most sedentary people have already died.
Real means, not BS. Real means intervention studies. Observational studies can be hypothesis generating, but you cannot derive definitive conclusions from them. Only through intervention studies, you can be definitively sure about the causality.
Most observational studies simply report spurious correlations, and cannot be trusted --- especially when "reverse causality" is a strong possibility. That is people who are healthier and fitter tend to exercise more, therefore they experience less disease, death etc.
Only by taking sedentary individuals, and coaching them to exercise significantly more, and comparing with a control group, you can be sure of the causal benefits.
That’s not incredibly surprising in older adults or people with chronic diseases - it probably just means that the intervention needs to happen much earlier.
Meshes with other studies I’ve read, like that people who were fitter in middle age have lower risk of alzheimers - so if you did an RCT of getting elderly people to do more exercise you’re not going to see much effect, since you’re not getting the years of cumulative protection from whatever it is that causes it and much of the damage is already done.
GDPR and DMA that came out of EU internet regulators don't impress me. They make it more arduous to browse the web, reduce the revenue collected by open web (content sites), reduce the quality of internet services, lower the quality of ads, hurt user experience, and hurt businesses who're looking to connect with customers.
Looks like EU will be doing the same with AI, and risk locking these countries out of the benefits of AI revolution.
Good job human(s)! It should be obvious to anyone who's following chess AIs: Progress seems endless; we're constantly discovering tactics and anti-tactics to defeat the latest model. That's why computer ELO is going up on a straight line. Progress in Go is admittedly a bit more obscure. Fewer people are working on it (leela go, katago etc). This article demonstrates that we can expect a similar trend.
I don't believe there's been any analogue to this in chess AIs in many years, if ever. Chess AIs are certainly getting stronger, but when is the last time someone found an adversarial strategy against them that even mid-level players, or a high-level player giving a full queen as a handicap, can pull off? This feels much more like "exploit" than "tactics".
In a game, especially a complete-information one that pits mind against mind, tactics are a matter of exploiting the rules to their utmost (maybe somehow tricking an opponent into breaking the rules, or letting you do so, would be an exploit, but I don't think this can be categorized as such.)
Gladwell is an extremely gifted author. He's a poor social scientist. When he wears a social scientist's cloak, and tells engaging and highly believable stories, he's doing a disservice to society. I won't read another Gladwell book (story), nor do I recommend it to anyone.
AI is clearly in the right track regarding performance and the speed of progress. Is it human level in every task? Not really. It clearly has room to grow. Is it super human in many tasks? I think so. Just learn how to use it. Peace.