2 ms·
GPT-4 is a tiny subset of "LLMs". If you train a model to do math (and optimize representation for that), it'll do math. GPT-4 just isn't, and, generally speak
by int_19h 3y ago
GPT-4 is a tiny subset of "LLMs".
If you train a model to do math (and optimize representation for that), it'll do math. GPT-4 just isn't, and, generally speaking, they aren't, because it's much more efficient to train them to "use a calculator". Same as with humans.