sm0ss117

Yea, this reads as LLMs are a pretty obvious technology to develop(for the highly intelligent researchers who are there). Also there's probably a lot of actual divergence in model capabilities and skills that concealed by the fairly narrow set of tests we run them against nowadays. Like wasn't Grok 4.20 super targeted at non-coding tasks.

I'm legit freaked the fuck out by this, it feels like a flashing red warning signal that the alignment problem is wholly unsolved and OAI isn't taking it seriously.

The disconnection between pelican quality and overall model quality is interesting. I initially assumed that since pre-training is when a model gets its general skill that it happened around when RL started to really differentiate models. That is higher quality pre-trains result in higher quality pelicans, but RL is unlikely to touch pelican quality. However the fact that GLM 5.2 beats GPT 5.6 and Claude Fable puts a damper on that idea.

My only guess is that GLM 5.2 was specifically RLed for SVG generation and that resulted in superior performance.

1. The four color theorem is a useful case study, for which the original proof was validated and 400 pages long. My prediction is that the first couple waves of proofs will be hard enough that a layman couldn't produce them, but simple enough that experts can verify them. Over time the most advanced proofs will get more and more complicated until humans can no longer verify them, this process could happen over the course of a few month or could take literally hundreds of years.

2. Especially early on the overwhelming majority of the proofs are likely to be uninteresting and more novel just because actually producing them would take expert time that's better spent elsewhere. That being said, as above over time I expect the interestingness of proofs to go up until they eventually regularly produce interesting proofs. The vast majority of proofs are likely to maintain their position as of no interest to humans for the simple reason that the vast majority of proofs are of no interest to humans.

In neither case will I make any particular guesses about a timeline beyond it seems like the way things will go.

Mathematics seems like the ideal candidate for AIs to achieve absurd results. It's a purely abstract grammar with true auto-verifiability. Even SWE has the requirement of interacting with real physical things. In math there's no external feedback required, you're solely bounded by the rate and quality of token generation.

The evidence says that during disasters ppl come together and help each other during disaster while state and business institution act in an at best obstructionist and at worst predatory way. So why do ppl in power believe the exact opposite? The simple explanations are that it's a delusion they actually hold, or a lie they project and either way it serves as a justification for the systems they benefit from.

Look, in order to prevent scary radiation from entering the environment through nuclear we have to use clean and even more radioactive natural gas.

There's a lot going on but the basic reasons are: Evangelical white Christians think the existence of Israel is necessary to bring about the Apocalypse, a thing they want because they go to heaven.

The Israeli right wing wants criticism of Israel to be inseparable from criticism of Jewish people. This means you can't criticize Israel without being painted as anti-Semitic, and to be fair there are a lot ppl who criticize Israel for deeply anti-Semitic reasons.

Quick rule of thumb, the reason we associate it with cancer is because the only ppl we systemically tested for it had cancer. If you tested everybody you'd probably find it in everybody since it was in a mandatory vaccine

Parts of India might get uninhabitable in the summer in some of the worse case scenarios. Given the population there and the fact they're a nuclear armed state that's a geopolitical powder keg.

If my choice is between definitely dying of boring old hypoxia and maybe dying in a cool explosion I know which one I'm picking.

But to be serious it's risk management and if the patient is definitely going to die of low oxygen then the patient having a risk of dying another way is still a lower risk.

This is a nightmare. Getting spammed by ads on my devices is bad enough. Now when I go out in public I get shamed for all to see. Targeted automated advertising has got to be banned outright.

Yea, if something is a Natural monopoly, which search almost certainly is, then it should be a publicly owned utility.