The Importance of Fine Tuning Language Models

The amount of meaning that the AI is getting out of the knowledge it's trained on is actually correlated with the quality of its output. So for example, if you train it on Wikipedia and then ask it a question about Harrison Ford or something, it isn't just going to literally just quote the text that it was trained on in answer to a question. It'll make its own evaluation of that and then answer the question, for example. I think that even though it's showing that custom models can be done cheaper and yield pretty good quality results, I don't think it negates the need for larger models That have more capabilities if you're looking for search for meaning and emergent behavior.

Transcript

Play full episode

Transcript

Episode notes

The AI-powered Podcast Player

Save insights by tapping your headphones, chat with episodes, discover the best highlights - and more!

Get the app