Alex Tamkin on Self-Supervised Learning and Large Language Models
The Gradient: Perspectives on AI
00:00
DABs: A Benchmark for Self-Supervised Learning
DABs is a main agnostic benchmark for self-supervised learning. It's trying to measure, you know, how well can we build general purpose, self supervised learning algorithms. And so DABs really did come in large part out of the things that we're thinking about with the whole Viewmaker project. We thought there should really be a unified benchmark or way to basically focus this community's effort on building domain agnostic algorithm.
Transcript
Play full episode
