@gregisenberg I once pitched to a VC because my friend was part of the investment team there. The VC came a half hour late, and brought his dog with him.
I should've pitched to the dog.
I believe in the Festivus School of prompt engineering, which says all prompts used in production naturally iterate toward an airing of grievances—a list of all the ways the model has disappointed you in the past year.
@yminsky I think there is a lot of potential here since the next token prediction can be narrowed down quite dramatically based on type matching. Unclear if anyone is proceeding along these lines..
@mametter Aren’t your top three languages also ones for whom the most training data is available? I wonder how well an LLM explicitly trained on ocaml code would do.
@Vjeux@DavidKPiano That is typically the definition of a non-chaotic system. A chaotic system can still be fully deterministic but small perturbations in the input space can map to large output variances.