Self-Supervised Learning and Foundation Models

35 points by rgabrielsson 2 years ago | 5 comments
  • nadermx 2 years ago
    It's nice MIT offers lectures like this and the gilbert strands lectures. Any other ML/Math related lectures recomended?
    • rgabrielsson 2 years ago
      One trend I'm seeing: I think the math that is becoming more useful, if huge foundation models are taking over, is around comparing and understanding their embedding/latent spaces. Geometry, topology, optimal transport, alignment -- in favor of statistics and probability theory
      • mdp2021 2 years ago
        The classic 6.034 of [Classic] Artificial Intelligence from the late Prof. Patrick Winston is "mandatory".
      • aliqot 2 years ago
        I preferred the alternative title: "What happens when you pour gasoline on your own bias"