• mindbleach@sh.itjust.works
    link
    fedilink
    English
    arrow-up
    70
    arrow-down
    2
    ·
    6 months ago

    Neural networks are magical anywhere that near misses are good enough.

    Companies keep using them as if they’re infallible, when lives and fortunes are at stake.

    Tech is not the problem.

    • Paradachshund@lemmy.today
      link
      fedilink
      English
      arrow-up
      37
      arrow-down
      7
      ·
      6 months ago

      Tech is ravenously trying to convince the world they need AI for every aspect of their business. Tech wants you to think LLMs are infallible and they strongly imply that they are even if the fine print says otherwise. So personally I would say tech is very much part of the problem. One could say they are the root of the problem in fact.

    • fckreddit@lemmy.ml
      link
      fedilink
      English
      arrow-up
      3
      ·
      6 months ago

      Yeah, I have to agree with you. For example, I would have no problem using a decently tested LLMs for engineering simply because Engineering usually accounts for errors and uses appropriate factors to accommodate them. Sure LLMs could be get more accurate in future, but I believe the error will reduce asymptotically. Essentially, more accurate LLMs get, it will get that much harder to increase the accuracy. There is always a price to pay, IMO.

      • JasonDJ@lemmy.zip
        link
        fedilink
        English
        arrow-up
        5
        ·
        6 months ago

        “There’s always a price to pay” is basically what engineering is.

        Anybody could build a bridge to last 100 years, or to survive a barge ship crashing into it, but it takes an engineer to build a bridge that will barely last 100 years, or barely survive a bridge crashing into it (which you could kind of say the F.S.Key bridge did, since only really the middle section was taken out).

        Put another way, in the real world, there are budgets and sacrifices.