muckrAIkers

OpenAI's o1, aka. Strawberry

Listen Later

OpenAI's new model is out, and we are going to have to rake through a lot of muck to get the value out of this one!

⚠ Opt out of LinkedIn's GenAI scraping ➡️ https://lnkd.in/epziUeTi

  • (00:00) - Intro
  • (00:25) - Other recent news
  • (02:57) - Hot off the press
  • (03:58) - Why might someone care?
  • (04:52) - What is it?
  • (06:49) - How is it being sold?
  • (10:45) - How do they explain it, technically?
  • (27:09) - Reflection AI Drama
  • (40:19) - Why do we care?
  • (46:39) - Scraping away the muck

    • Note: at around 32 minutes, Igor says the incorrect Llama model version for the story he is telling. Jacob dubbed over those mistakes with the correct versioning.

    Links relating to o1

    • OpenAI blogpost
    • System card webpage
    • GitHub collection of o1 related media
    • AMA Twitter thread
    • Francois Chollet Tweet on reasoning and o1
    • The academic paper doing something very similar to o1

    Other stuff we mention

    • OpenAI's huge valuation hinges on upending corporate structure
    • Meta acknowledges it’s scraping all public posts for AI training
    • White House announces new private sector voluntary commitments to combat image-based sexual abuse
    • Sam Altman wants you to be grateful
    • The Zuck is done apologizing
    • IAPS report on technical safety research at AI companies
    • Llama2 70B is "about as good" as GPT-4 at summarization tasks
    ...more
    View all episodesView all episodes
    Download on the App Store
    muckrAIkersBy Jacob Haimes and Igor Krawczuk