Using homomorphic encryption with machine learning: https://iamtrask.github.io/2017/03/17/safe-ai/ (a nice solution to protecting privacy in training)
Related Posts
Exploring design ideas using AI as a learning assistant, for creative exploration: https://andymatuschak.org/hmwl/
TIL Advent of Code and Project Euler will deliberately look for puzzles where the naive solution is worse than quadratic.
This ensures that people can solve them with any programming language. You don't want fast languages to be able to use the naive solution.
LLMs are a really accessible machine learning technique. I dabbled with text classifiers a few years ago and the APIs were way more involved.
(system_prompt: String, input: String) -> String
I can prototype with this much more easily!