• by ftxbro on 4/12/2023, 11:46:50 PM

    This is an excellent kind of analysis I was curious about this kind of thing and I don't know why I never saw it before.

    Conclusions are that by 2030 there will be at most 1000x the available computer power and data for AI models unless some super weird things happen.

    EDIT: oh it's submitted by someone at openai no wonder lol

  • by ftxbro on 4/13/2023, 7:34:09 AM

    Imagine writing one of the most insightful takes on GPT and it gets only 2 upvotes and 3 dumb comments all by the same account. Something must be wrong with the title, probably it needs more clickbait. I had expected that by now it would have 300 upthumbs and fifty comments.

  • by ftxbro on 4/13/2023, 12:05:47 AM

    If anyone is interested in this kind of scaling laws there is a reddit (oh no!) for tracking it: https://old.reddit.com/r/mlscaling/