Ep 7: Co-Creator of Databricks Dolly Mike Conover on Open-Source LLMs
Unsupervised Learning
00:00
The Four Minute Mile: The Existence Proof of Chat GPD
GPDJ and Pithia 12 billion are the standard benchmarks relative to the instruction to model, they're just really not distinguishable. Even the Databricks dollar 15k data set probably represents between 500 and 1000 hours of writing time. And so I think that's part of the recognition as well that, oh, this can be done. It seems like there was some latent demand for that.
Transcript
Play full episode
