I’m a climate scientist by trade. Interested in interesting things. Ecology, complexity, politics, social change, music.

  • 18 Posts
  • 1.58K Comments
Joined 2 years ago
cake
Cake day: May 8th, 2024

help-circle






  • What I would like to see is a LLM that’s developed … with FULL traceability to sources uses,

    That’s not really possible. I mean, you can list the training data, but you can’t attribute a specific output to a specific piece of training data, because all of the training data influences all of the weights. It’s not possible to know which training data influenced which weights and how, because that data is not retained (and it would be massive and difficult to interpret if it was -on the order of weights × number of training context windows, and the weights alone are already terabytes)


  • Why is it an entirely different conversation? Or rather, why do you think that this conversation is not what the AI-opposition are concerned about? (At least some of us are)

    By analogy, many of us leaped in to social media 20 years ago. 15+ years later, there was wide recognition that the new technology brought with it a shitload of undesirable outcomes that were completely unpredictable at the outset. The same can be said of fossil fuel use, as well as factory automation. Why would you not take lessons from those? Especially given what you know about the arseholes running the GenAI companies?