Sabrina wrote an interesting
write up on solving a math
problem with gpt-4o. It turned out the text-only, chain-of-thought approach
was the best performing, which is not what I would have guessed.
It’s was cool to see Simon dive into LLM-driven data extraction in using his
project datasette in
this video. Using multi-modal
models for data extraction seems to bring a new level of usefulness and makes
these models even more general purpose.