energy123

How do you know this? Anthropology?

Can't they use ML to predict where I'm going to click, and pre-cache the predicted page whenever the predicted button doesn't mutate important state? Or skip the difficult ML and have some basic rule of thumb that pre-caches frequent button clicks, using a markov chain, and conditioned on those pages being low bandwidth to pre-load.

That hypothesis requires substantiation. Just because 10 people in a room came up with a breakthrough doesn't mean a sociological process contributed positively to that breakthrough. People generally like socializing, so the base rate is going to be high by default. The breakthroughs might have been more numerous if they had each followed Grothendieck's advice (who is a contender for the world's most talented theory builder) and intentionally decorrelated from one another to free themselves from convention.

Referring to my list of examples as exceptional in the sense of being rare is rather unfair, given how numerous these examples are relative to the body of mathematical work we would consider incredible and especially relative to the desire of the average human to socialize. The fact that such a large % of that body of work occurred while the individual was in relative isolation is something we should pay attention to.

Not error rates which implicitly assumes all errors are of equal consequence, something like 1/N*∑(magnitude of consequences of error) * (did error occur)

Many eminent mathematicians did their best work while not talking about it with anyone, sometimes in isolation. Newton's calculus, Perelman's poincare, much of Grothendiek's work, Wiles's fermat, Ramanujan's earlier days. That's not all top mathematicians as you can look at Von Neumann as a sociable counter-example. But it shows that discussion of your current ideas is not a requirement. Grothendiek goes so far as to say it is a net negative for mathematical creativity because it is difficult to resist thinking like the herd without some level of seclusion.

This is relevant, but only after AI has solved all the open problems including Millenium problems. Until then, as AI keeps solving harder open problems, people will pay attention and be interested.

I was referring to the gases from the cars on the highway more than easily removed indoor sources.

It's a good start, but for VOCs and gases you'll need activated carbon. Those 2-in-1 filters do exist but cost more. And many are scams because they don't have enough volume of activated carbon.

You have activated carbon to filter out the VOCs?

It's a dumping ground for poor people.

Viktor Frankl's body of work deserves a plug. Meaning can be found in the worst conditions, it's quite subjective. With meaning comes contentment. Buddhist teachings also deserves to be mentioned. Suffering comes from clenching to what the external world should be instead of just accepting how it is.

That's a good thing to note. The same can be said of Newton, Wiles and Perelman.

Grothendieck's later years or more extreme isolation were less productive, same with Perelman, so maybe there's some balance that needs to be struck. Although age is a confound with these two cases.

Biden's ARP independently caused inflation according to multiple central bank analyses.

However, you are also correct that Trump pressuring OPEC to cut oil production at the end of his first term did cause additional inflation in Biden's term.

I actually think primatology is quite useful here. We aren't that different from other primates. We selfishly attach extreme morality only to things that impact us, which we would never do when we analytically study primates doing the same kind of thing (stealing and killing). To be frank, your kind of comment makes me more suspicious of human morality itself (as a social tool to mask resentment or overcome collective action problems towards some kind of aggressive action) than it does of humanity.

The glasses are a trust destroying device that degrades privacy.

If someone shoves a camera in my face in public I am going to move away from that person. If I see them filming someone with me in the background at least I know what is going on and can judge the situation accordingly.

Now I don't know. Everyone with glasses is a suspect. It's more public tension, hyper vigilance and uncertainty.

If it's a fine tuning step at the end, why is the need for it to vary a problem? Can't you run the fine tuning, test for regression, and deploy the weights in a day?

I think the more likely reason is it doesn't work as well as in context learning. Otherwise they would prefer to avoid polluting context and degrading performance.

Cheap cooling vests distributed at scale, kept cool or "recharged" by a portable and unpluggable solar panel, seems necessary. Individual air conditioning is too expensive and most affected people need to work outside during the daytime.

Pointing the finger at RLHF is basically right. It removes variance from model outputs compared to base model. That makes each output more predictable and more correct on average, but across trials it repeats the same thing.

It's relevant to AI safety. If you have a diversity of outputs, the AI will agree to hack the bank 0.1% of the time regardless. If you have a uniformity of outputs, in most contexts the AI will hack the bank 0% of the time, but in certain odd contexts, all AIs will work together to hack the bank 100% of the time.

The model providers could randomize the system prompt to make it say no 2.36% of the time, automatically tuned up or down depending on user feedback.

Accessing the kind of "feminine" mental states that Grothendieck talks about, basically a type of creative flow state, is not an example of that. It's in the realm of meditative, passionate and artistic.

Other examples like Grothendieck: Wiles (Fermat), Perelman (Poincare), Newton (calculus), Ramanujan.

> Spend too little time and you toil away in a cave, at best reinventing things that were invented already without your knowledge.

Grothendieck disputes this point. He reinvented the wheel many times when he was 17-20 years old, but believes it was training for accessing the required cognitive spaces. He says:

   A few years after I finally established contact with the world of mathematics at Paris, I learned, among other things, that the work I’d done in my little niche [. . . had] been long known to the whole world [. . .]. In the eyes of my mentors, to whom I’d described this work, and even showed them the manuscript, I’d simply “wasted my time”, merely doing over again something that was “already known”. But I don't recall feeling any sense of disappointment. [. . .] Without recognizing it, I’d thereby familiarized myself with the conditions of solitude that are essential for the profession of mathematician, something that no-one can teach you. [. . .]

Non-hierarchical structures (like anarchy) are plausible only when defence is stronger than offence, because the individual/clan/nation can defend against any offence from any other entity, while also being incapable of attacking any other entity due to weak offence relative to the other entity's defence.

But reality doesn't owe us this luxury, unfortunately.

Offense increases entropy. It's easy to destroy. Attack surfaces are numerous. Humans are fragile. It's hard to defend.

This is what AI safety activist Connor Leahy calls offense-defense asymmetry. Imagine if a nuke could be created with household materials. We got lucky that wasn't the case, but we didn't know that it wasn't the case before studying physics. If we weren't so lucky that physics happened to turn out that way, we already wouldn't exist, because the offense would be so much stronger than defense, we wouldn't stand a chance even with the best surveillance systems.

A profitable prediction markets trader who goes by the alias semiotic rivalry posted this prediction on his X account recently:

> Prediction: in 1-2 years, almost everyone will agree that open sourcing the best AIs is totally crazy as the AIs become capable of autonomous hacking of most humans and institutions, serious terrorism and creation of biological weapons, etc.

And I mostly agree with him. As much as it's awful to have such a concentration of capabilities in the hands of a few governments or corporations, as far as I can tell there are no easy answers here, only a matter of trade offs and different risks.

Words vs actions+incentives

> closed sourced model pricing has no relationship to its size.

That's too strong. Only in an actual monopoly for a product with no substitutes that has price inelastic demand can pricing fully disconnect from costs. Frontier model serving is only maybe a soft version of that, where costs and moat both contribute to pricing.

Those results were found with 64 subagents, or with GPT 5.6 Sol Pro which uses parallelism under the hood.

Low latency is a big deal but the immediate use cases are somewhat different in the short-term (more serial coding workflows) rather than pure math research which is effectively massive-scale search through a tree of possibilities, which is where you want throughput and low cost per token, rather than high speed per token.

That's an argument for cheap massive parallelism, which is what's behind the recent math breakthroughs, not expensive low latency serial compute.

That's a moral judgement about what words to use that has no bearing on causality. Incentives are never a valid justification but they are almost always an explanation. We should care about explanations when trying to solve problems.

OpenAI claims that the harness that ARC used was unfairly handicapping the model: https://openai.com/index/how-two-settings-tripled-our-arc-ag...

Because not everyone shares your preference and/or discipline for plants. Lab grown meat doesn't require a planetary scale realignment of moral values and habits.