Voracious Thinker

Voracious thinker, digester of info. Interested in AI, tech, science, philosophy, economics, and more. Dry humour occasionally.

[Voracious Thinker @voracious_think]

@chesscom (If I remember correctly)

[Voracious Thinker @voracious_think]

@chesscom Story is that it was meant to represent a battle lost by a Persian Shah (from which Shah-mat, checkmate, etc), so I would say at least 1200-1300 years old.

[Voracious Thinker @voracious_think]

@himanshustwts (Same for a coding task)

[Voracious Thinker @voracious_think]

"Wherever there is no clear reward" to me seems everything that is not STEM research.

How do you design a reward function for how Shakespearean a generated piece of poetry is? For how Impressionistic a generated painting looks?

For sciences, rewards functions are much easier to design. If the p-value of a generated research hypothesis test is < 0.001, if the cross-entropy loss of a proposed architecture is lower than a certain threshold if the Lean proof is validated: it was a good idea and reward it, don't otherwise.

It can literally be a binary reward, which if I remember correctly is what is done in GRPO.

[Voracious Thinker @voracious_think]

It increasingly seems the case that a sufficiently good statistical model of the world, as expressed in language, renders deeper understanding of cause and effect relationships implied at best, superfluous at worst.

I would have not bet on it, not gonna lie.

[Daniel Kokotajlo @DKokotajlo]

Unfortunately, it seems like the intelligence explosion will be happening in 2027 after all. Most of this post analyzes the new METR report (though other sources of evidence are also important). If they’re right, that implies an AI-driven acceleration that is 5 months ahead of the AI 2027 scenario. I am scared, sad, and angry. https://t.co/7KWmdc6AZk

[Voracious Thinker @voracious_think]

@MercuriusFilius I would ask as hint whether I should use a Discounted Cash Flow model or Discounted Dividend model

[Voracious Thinker @voracious_think]

@JasonCorley78 @MercuriusFilius This is poetry

[Voracious Thinker @voracious_think]

@tunguz Unpopular opinion:

Those who can’t, teach.

[Voracious Thinker @voracious_think]

All true.

People really tend to underestimate how much the sunk cost fallacy affects their lives - and I say this looking in a mirror, not from above.

[The Sigma Mindset @thesigmamindset]

He literally explained why your past is holding you back
‼️

[Voracious Thinker @voracious_think]

@chesscom Fun aside, I think it’s 2.Ke2

[Voracious Thinker @voracious_think]

@craigweiss New benchmark just dropped

[Voracious Thinker @voracious_think]

By the way, they're called Data Scientists because Data Sanitation Workers was not cool enough. But that's basically what the whole job is about.

[Voracious Thinker @voracious_think]

Every single life failure stems from your unwillingness to have a tough conversation with someone.

[​𝐥𝐲𝐫𝐚 @sunnkssdseraph]

What's the hardest lesson you've learned?

[Voracious Thinker @voracious_think]

@onlypreneurs From a famous interview: "Life rewards action, not intelligence".

[Voracious Thinker @voracious_think]

@onlypreneurs I agree, but the problem is that there is always more. For each window you close a new one gets opened, so you'll never feel tired. Up until you decide to stop opening new windows, at least the useless ones.

[Voracious Thinker @voracious_think]

@fnthawar Real magic happens in the work you do when no one is watching.

[Voracious Thinker @voracious_think]

There are some initial evidences it may not be correct actually:

https://t.co/szOD0gkeNI

"Hence, informally, providing semantically faithful AI autoformalisation is harder than any computational problem including the Halting problem (which has SCI = 1). To demonstrate the effect of this result we provide several examples of AI mistranslations of NL statements and proofs into Lean in practice, resulting in mismatches between NL proofs and their Lean ‘verifications’. These include OpenAI’s announced Navier-Stokes proof. In particular, we show that the formalised Lean proof does not correspond to the NL proof of blow-up of solutions to the Navier-Stokes equations".

This does not prove that it is wrong, to be clear, but only of translation mismatches between its natural language (NL) text and its computer-verified Lean 4 formalization.

[Voracious Thinker @voracious_think]

Btw, do we have any about @OpenAI ‘s Navier-Stokes proof? Is this officially validated anyhow?

[Voracious Thinker @voracious_think]

@ForrestRoche @skdh But how does the mutation even take place at start? We can say randomness, but in reality is a response to its environment. And that’s learning.

[Voracious Thinker @voracious_think]

For me is more a constant underlying feeling, energizing one's action even without knowing it.

Which is why I find the posts "If you're ambitious but lazy, do this" cringe.

If you're truly ambitious you don't need this or this other trick to get to work towards your goals. You may actually need tricks to stop working.

[Voracious Thinker @voracious_think]

If you're ambitious but lazy, you're not ambitious.