Colin Clarke

Data, AI. Extension of my Notes app

[Colin Clarke @clarke_space]

@TheNormNetworks @tszzl It's not closed vs open source models. It's Attack vs Defense. If it hasn't already, Attack will soon gain an unassailable advantage over Defense. Its hard to imagine a free and open Internet in that world. But hey at least the models will be competitively priced right? Right?

[Colin Clarke @clarke_space]

@weswinder @ylecun @FukuyamaFrancis You want AI to keep getting better so you can vibe code more impressive toys.
Yann just raised like a billion dollars to build superintelligence.
If I was able to persuade you both an international treaty is a good idea maybe I'd be superintelligent 😅

[Colin Clarke @clarke_space]

@kpertsev @weswinder @ylecun @FukuyamaFrancis Maybe but were we not a hell of a lot closer to building nukes once we had it in large quantities? What do you think is a better analogy for llms compared to eventual superintelligence?

[Colin Clarke @clarke_space]

@weswinder @ylecun @FukuyamaFrancis "highly enriched uranium will never be a nuke"

[Colin Clarke @clarke_space]

@InfiniteReign88 @ylecun @FukuyamaFrancis

[Colin Clarke @clarke_space]

@ylecun @FukuyamaFrancis I get that we're character limited here, but if we shared the same definition of superintelligence, you would think twice about glibly dismissing such an important conversation.
To entertain your analogy, I think we're much closer to 1944.

[Colin Clarke @clarke_space]

@ylecun @FukuyamaFrancis Your analogy is wrong in too many ways to count. Your hubris is pathological.

You clearly either believe superintelligence is impossible, or don't know what it means.

Play a game of chess against Stockfish, then map how you feel to every strategically important area of life.

[Colin Clarke @clarke_space]

@tszzl @BjarturTomas Don't you work for OpenAI??

[Colin Clarke @clarke_space]

It's never been easier to make stuff, but the psychological barrier of "someone's already done this better" has never been higher

[Colin Clarke @clarke_space]

Thought experiment:

Humanity vs Stockfish.

- Every human on Earth is on one team.
- We get 24 hours to discuss each move.
- No chess engines or outside assistance.
- At the end of the 24 hours, humanity submits one move by global vote.

Who wins?

[Colin Clarke @clarke_space]

@brian_a_burns @tszzl Imagine the damage a Stockfish-level singleton AI could do if generalized.
Imagine the coordination we would need to put it down. Now imagine 1000 of them. You might not be saying "the global economy is the real ASI" as we scramble to maintain power.

[Colin Clarke @clarke_space]

There should be a font/punctuation mark to indicate that a query you have is purely information seeking and not normative

[Colin Clarke @clarke_space]

@benjamin_horne What would it look like if they were genuinely seeing signs of imminent recursive self-improvement? What would be different?

[Colin Clarke @clarke_space]

@liron @nikitabier What is happening here? When/where is this from?

[Colin Clarke @clarke_space]

@michelleefang .

[Colin Clarke @clarke_space]

@SocraticScribe https://t.co/IAAeUmqjyT buy it here!

[Colin Clarke @clarke_space]

@nabeelqu How do you cope with it's incessant thesaurus-speak though? Every sentence feels like a vocab flex which actually obscures the underlying message🤔

[Colin Clarke @clarke_space]

@SimonJalbert SEND

[Colin Clarke @clarke_space]

@charliebennett @Readwiseio save thread

[Colin Clarke @clarke_space]

@girdley @readwise save thread