Thomas Hunter IIAn LLM is basically a big database that's slow and expensive to query. The code it produces is just a query result. If you publish a library made up of LLM-produced code then that library is just a cache.
Bart Louwers@tlhunter I don't think the database analogy applies. The output of an LLM is stochastic.
Also if you call anything that involves a computation on some data and a query a 'database', the word loses its meaning. I don't think a chess engine should be considered a database. Neither should a much simpler linear regression model.