Learn

A humble view, in this age of LLMs, where they can seemingly solve our hardest (but verifiable) problems, like the millennium problems, would be to not stop learning.

I personally never thought that I could ever solve any open math problems, or even theoretical computer science ones. My whole motive was to keep learning for the sake of it, and pursue my curiosity to the ends of math, theoretical physics, theoretical computer science, Chemistry, Biology, economics, history, psychology and such. The first few, it seems may fall to LLMs, as they can provide solutions to many of our problems. A paradox thus appears: should we continue learning the field in question ? My answer is yes.

I believe even if LLMs can solve any math problem we throw at them, we still need:

• People who know what problem to throw
• People who can formulate the correct problems
• People who can educate, explore and pass on the results of LLMs. One may argue that none of these are necessary, once the LLMs become "good enough". I do doubt that however, since:
• No one is particularly training them to be good communicators right now (there's some mention of orthogonality between communication skill and problem solving skills)
• It becomes particularly harder since we cannot truly/formally verify anything (which is the whole reason of the current progress)

This comes from a pure learning perspective. I've aggressively increased the number of concepts I try to understand, and actively try to find different mediums to learn myself. Learning was always the goal for me, and in that process, if I can contribute and produce something of value, that was always a happy by-product !