Neuroscientist Konrad Kording reveals shocking truth about machine learning and the brain
The Rhys Show
00:00
The Stanford Alpaca Model and the No Free Lunch Theorem
The Stanford alpaca model was built on the llama model. It looked like a hundred billion parameter model while only being a seven billion parameters and they fine-tuned it. So that makes me think of what you should just say which is here's a successful model And here is um, let's like train the dumber model on the successful model to show it how it should act Is that how does that live? I'm not sure if I'd call that a deconstraint but there might be some interesting constraints there.
Transcript
Play full episode
