Anyone can easily use private, open weight models

(micahflee.com)

20 points | by DeepLogin 4 hours ago

3 comments

  • bobajeff 3 hours ago
    I could swear I've heard of Maple or something very close to it before on here, but can't seem to find any other posts on it.
  • 0xbadcafebee 50 minutes ago
    There's still a number of theoretical ways to compromise your data in this system. The closest thing to real privacy is running the model on your own computer in your own home.
  • jqpabc123 3 hours ago
    But they can get away with this because, as long as the bubble hasn't popped, they don't care.

    Looks like a classic "bait and switch" to me.

    The main thing standing in their way is cheap Chinese models. But maybe they (and the markets) were counting on a political solution to this issue. In the aftermath of the election, this looks less viable so investors are starting to bail.

    • logtempo 39 minutes ago
      I believe everyone would’ve been ok with ia if it was a coordinated, worldwide effort to make ia as a « public R&D ». Doing somthing in common, for once.
    • docshock 1 hour ago
      They are transparently relying on regulatory capture in the name of "safety". Then it's game over.
    • devmor 1 hour ago
      I assume it is partly a loss leader as well.

      What an individual pays $20-100/mo for, large corporate buyers pay API pricing on tokens for.

      The provider benefits by keeping and training users to be reliant on their offering so that corps have a large pool of people to use those tokens.