Hacker News

points by estearum 1 day ago | 0 comments

This doesn't even address the direct point. The question is how any of this is fundamentally different (difference in type and not just of "scale") from how human cognition works.

[–] datsci_est_2015 1 day ago

Well, I've addressed it implicitly. Every human collects different context by living one of 8.3 billion+ unique lives. We don't have 8.3 billion+ different LLMs. The only way to simulate that is to perform a random walk through the compressed language-based knowledge storage by injecting random context.

Also, those 8.3 billion+ unique lives already produce contexts that are highly optimized for solving human problems that modern humans experience in the physical universe most relevant to humans.

[–] pixl97 14 hours ago

So all you're saying is we need log term context in LLMs and then we could have 'trillion' of unique LLM paths.

If that's the only limitation that you're putting forward than LLMs have already won over humans. If you're not hoping for some quantum level limitations then it's already over for us.

[–] Diti 1 day ago

One opinionated answer to this question: LLMs can’t jump [1].

[1]: https://openreview.net/forum?id=klU4737opt