laughingcurve

It's unfortunate we've reached such an insane level of cynicism

Thoughtful piece that understands economics

Criteo is a dogshit company that makes money by tracking you everywhere. the idea of them telling anyone how to skill up or be better is ridiculous.

“B-b-b-but muh pop culture trope”

Thank you for commenting this. I'm tired of hearing about Dune like its some kind of utopia

Interconnects is the best place to keep up to date on open weight models because of analysis and discussion like this

Agreed. Just based on this not being Ollama, so I will give it a try.

Distillation attacks? Volume of calls?

Dang, I thought I was being harsh I wouldn't say scam but maybe all AI is a scam to some.

In multiple threads now you've claimed I was wrong and then ended the conversation without any assertion, claim, or argument I could rebut. You've ended those conversations with snarky comments clearly designed to shut down debate. This is not how hackernews benefits.

Edit: Nevermind, realized he's just an uneducated troll just looking for his kicks. Comment flagged.

This is a failure to engage the clear arguments and claims made. If you don't want to debate something, why bother commenting at me? Was I supposed to just cede ground and accept your framing wholesale? I'm putting forth very clear and open-to-rebuttal assertions which is what you should do.

Edit: Nevermind, realized he's just an uneducated troll just looking for his kicks. Comment flagged.

Can you explain how 'recursive self-improvement' functions without 'endless benchmark chasing'? I mean, RSI is literally that.

What do you think they're improving on? How would a model self-improve without some metric/data of some kind to check? When you have metrics+data, that is a benchmark. And yes, simulations and or soft-verification like LLM judges are still a kind of benchmarking. Maybe its not a static benchmark they can easily hack.

Folks -- RSI does not mean the self-improvement is them going to therapy and seeking inner peace to overcome trauma.

It's natural for literally any AI lab to end up doing auto-research, since 'auto-research' is literally just 'autonomous AI' which is the whole darn point of all of this. I'm not going to hand out genius brownie points to folks working on RSI because of course its powerful. How about we hand brownie points to the folks who do things that are not hype and end up being important/powerful?

> was posed specifically under the framing of there needing to be more fundamental research beyond squeezing as much as we can out of relatively vanilla transformer stacks.

Not to be contentious, but this is so broad of a description that it could include literally thousands of papers in the last year or two. I'm imagining double digits or more if we go back the full decade.

I'm saving brownie points for people who deserve them

TRUE