Will Cook

22

1,2,3,4 Some crazy-ass kids gonna knock up on your door, so...

[Will Cook @cookie1234593]

@danilexn To say it hasn't revolutionisd SWE reveals how much confidence you have in what you don't know enough about lmao that is hilariously wrong https://t.co/RJJ3dcf1d2

[Will Cook @cookie1234593]

@programmer_just nevermind was a glitch smth going on

[Will Cook @cookie1234593]

@programmer_just think they just reset

[Will Cook @cookie1234593]

@AbroadInJapan They do NOT think they'll be able to control it lmao, more like figuring out how to "raise it". If your kid is guaranteed to have >100x Von Neumann level technical genius with Steve Jobs level taste and Elon Musk's drive, there's not controlling it, just good parenting and pray.

[Will Cook @cookie1234593]

@alexiskold It's AI written with extreme rules like no punctuation, written so badly that it couldn't possibly be AI, except it is

[Will Cook @cookie1234593]

@BetterCallMedhi @sama intellectual LARP

[Will Cook @cookie1234593]

@hyperconsumers @b_geist I'm sure they're going to have far far smarter models, I'm just talking about the floor of capability. The Bel equivalent in a year will be something we likely can't comprehend right now.

[Will Cook @cookie1234593]

@pashmerepat These
https://t.co/sO61cfaQKX

[Will Cook @cookie1234593]

@PI010101 That's generally how it goes if there's a problem unsolved for 90 years.

[Will Cook @cookie1234593]

@_SsAsS_ https://t.co/tGy22StBbX

[Will Cook @cookie1234593]

@mark_k Interested

[Will Cook @cookie1234593]

@rperezmarco @OpenAI @AnthropicAI @SebastienBubeck @__alpoge__ How does the behaviour of OpenAI vs Anthropic differ in your view? Genuinely interested

[Will Cook @cookie1234593]

@goth600 Creating diseases is way easier than curing them given the assumption you had a no holds barred model. Just look at COVID.

[Will Cook @cookie1234593]

@TJeparskis That's what Anthropic do and everyone complains, you people can't be appeased lmao

[Will Cook @cookie1234593]

As in if you go back up the family tree you arrive at the very first single cellular life was in some capacity. It really depends on how you define awareness, is the spectrum so generous as to include single cellular life? We see emergent phenomenon in all sorts of systems as complexity grows, why can't awareness, or perhaps we can rephrase as agency, be another instance of this?

[Will Cook @cookie1234593]

@joshwhiton you're saying awareness itself organises, over time, greater forms of awareness; that's a stronger statement than saying the awareness "leveling up" is simply a natural consequence of evolution, rather than something that originates from awareness itself.

[Will Cook @cookie1234593]

@ishanr07 @ycombinator

[Will Cook @cookie1234593]

@Exomnium @cryptodavidw Why?

[Will Cook @cookie1234593]

How does it look like economics is holding up right now? It's circular because by beign unverfiable you cannot measure what "good" means objectively. Except when you bridge these fields into mathematics, modern economics is mostly maths.

AI excels in verifiable domains because of how training lends itself to a reliable measure of "good" and "bad". So experimentation and applying things like statistics to measure accuracy of theory can allow domains such as economics to become ripe for this same feedback loop.

Plus as models become generally smarter, you don't necessarily have to train for an AI to get better at each specific domain, it just happens. That's what we've seen so far, and your argument that this will stop because we run out of what to train forgets that methods of training that these labs use are guarded secrets.

What you know is probably years behind the forefront methods. Why do you think the Palantir freaked out over data retention?

Okay maybe AI right now cannot invent general relativity, but perhaps someone with domain expertise could formalise a modern equivalent breakthrough very quickly with AI assistance. Then suddenly a large quantity of rich, novel and breakthroguh data becomes available. Do you see the argument I am making? More than one way to increase the rate at which society learns to skin a cow type of thing.