Poll picks markdown database with RAG for AI learning
Herbie Bradley posted results from a poll on methods for continual learning in advanced AI systems capable of serving as drop-in knowledge workers. The winning response combined an md file database with retrieval-augmented generation. Bradley said the outcome surprised him because he believes not all required knowledge can be captured in explicit text. Timothy B. Lee replied that the RAG answer is ridiculous.
results from my poll are out! the winner was "md file database + RAG".
I find this quite surprising since I fundamentally don't think you can learn everything you need for this milestone in explicit text form.
@binarybits I agree! but somehow it won the poll...
@herbiebradley Sorry the RAG answer is ridiculous.
@herbiebradley Sorry the RAG answer is ridiculous.