yelvin

[Epoch AI @EpochAIResearch]

AI is getting cheaper more quickly than any other transformative tech in history. At a given level of performance, cost has fallen ~47%/quarter since 2023.

That’s 4× faster than DNA sequencing, 6× faster than compute, 18× faster than lithium batteries, and (up to 1973) 54× faster than electricity.

[yelvin @comtopost]

@AestherML Bread

[yelvin @comtopost]

@yuzu_4ever It has always been a hard path. Whether it becomes easier or harder because of AI induced changes remains to be seen. Do it for the love of your field and passion to understand. Good luck!

[yelvin @comtopost]

Global civilization is a superintelligence and it is not clear if it is “aligned”.

[yelvin @comtopost]

@dwarkesh_sp @polynoamial Very good episode. Informative discussion about alignment.

[Yafah Edelman @YafahEdelman]

@PoliticalKiwi ... do we need to start publishing graphs of indicators of AI impact on the world that show no change. We could publish a lot of those. Here are some (very vibe coded) examples:

[Nate Silver @NateSilver538]

One political problem for AI labs is that they can't articulate an optimistic medium-term future that sounds broadly good for humanity without also sounding incredibly weird, having huge wealth/power disparities, or both.

[yelvin @comtopost]

@AlejandroZarUrd @haider1 If you’re an actual person I hope you realize that your beliefs in the moral superiority of machines amount to a religion. Which is also Tao’s point here, not about math, that we’re stepping down on the gas not knowing if there’s a cliff ahead.

[yelvin @comtopost]

@AlejandroZarUrd @haider1 Who cares about truth?

Humans

[Teortaxes▶️ (DeepSeek 推特🐋铁粉 2023 – ∞) @teortaxesTex]

I realized I'm way more interested in what lies beyond all human mathematics than in whether humans + AIs could solve all problems they've posed in my lifetime.
Is there anything, or will we just… complete meaningful, knowable math? What kind of AI can even ASK new questions?

[yelvin @comtopost]

Legend

[Fermat's Library @fermatslibrary]

Jean-Pierre Serre turns 100.

Youngest Fields Medalist ever: 27.
First Abel Prize laureate: 2003.
and still editing Wikipedia.

[Naval @naval]

The best way to pace the frontier is to hold the labs fully liable for the behavior of their models.

[Elliot Glazer @ElliotGlazer]

Nobody has taken me up on my $25k wager offer yet! C’mon, whatcha waiting for? @steph_palazzolo’s mole says its free money, take the offer before someone else nabs it 😮‍💨

[yelvin @comtopost]

@dioscuri It depends on whether the final form of AI is one monolithic general model, or highly decentralized specialized models. The latter is preferable for many reasons.

[yelvin @comtopost]

I’m surprised how much mathematicians seem to be despised.

[yelvin @comtopost]

@ben_golub The letter is perfectly coherent if by “prove” you don’t mean getting the Lean checkmark, but actually understanding. Mochizuki was also shunned for not explaining his claimed proof of the abc conjecture.

[yelvin @comtopost]

@robinhanson What do you think are good reward metrics for academics?

[yelvin @comtopost]

@francoisfleuret Nice caricature of a high school/undergrad "math kid" who has yet to develop into a mature person.

[yelvin @comtopost]

@ATabarrok If AI doom happens then everyone (including Anthropic and Dario) loses

[François Chollet @fchollet]

A very common sentiment I'm hearing among math students: "I don't want to become a mathematician anymore." I've seen this before, I think the generation of digital artists who came of age around 2022 will be the last.

AI might not functionally replace mathematicians or artists, but it might well cause their near-extinction anyway, by crushing the spirits of the younger generations.