The point is that you call the LLM to generate the code that lets you talk to the API, rather than writing that glue code yourself. Not that you call the LLM to talk to that API every time.
1,547 karma · joined April 1, 2011
> Let models talk to each other directly, making their own case and refining each others’ answers. Exemplified in patterns like Multi-Agent Debate, this is a great solution for really critical individual actions. But XBOW is basically conducting a search, and it doesn’t need a committee to decide for each stone it turns over whether there might not be a better one.
In general, this seems reasonable to me as a good approximation of what works with humans, but with _much_ faster feedback loops in communication.
He was insufficiently confident of the power of the intellectual tools he already possessed and did not drive his thought to the very end because he felt instinctively that new ideas and new methods were necessary to go further than he and his students had already gone. Some may call it a lack of faith, but in my opinion it was more a turning away from the hard, crude methods of theoretical physics into a mystical realm of broad intuition.
In my own field of network security, I think that's largely true: we have a lot of analytical tools that we haven't applied sufficiently well, largely because people are not interested or educated in the sorts of theoretical tools we could apply (ML or graph theory, for example).
How well this applies to theoretical research is outside my expertise, though...
[0]: http://istruecryptauditedyet.comThis doesn't even rise to the level of what Weev did.
[0]: http://catb.org/~esr/faqs/smart-questions.html