Login
You're viewing the functional.cafe public feed.
  • Simon Formancarapace@mastodon.social
    Oct 9, 2026, 7:28 PM

    If you made an industrial robot and sent it out in the street and it killed someone holy shit you would be in trouble.

    But with this one trick you can evade all responsibility and criminal liability:

    Just make your killer robot look like a car.

    Oh. and be sure to put a human in it so they can take the blame if you kill someone, it's called a "morality crumple zone". It doesn't matter who, they're just a sacrificial lamb anyway.

    Bloddy hell.

    💬 1🔄 0⭐ 1

Replies

  • 💬 1🔄 0⭐ 0
  • Simon Formancarapace@mastodon.social
    Oct 9, 2026, 7:30 PM

    People get away with manslaughter and yeah even murder every day all over the world, if the murder weapon was a car and the murder victim was riding a bike.

    💬 1🔄 1⭐ 0
  • Simon Formancarapace@mastodon.social
    Oct 9, 2026, 7:31 PM

    Anyway, in other domains like machine translation of human languages the situation is not so cut and dried.

    💬 1🔄 0⭐ 0
  • Simon Formancarapace@mastodon.social
    Oct 9, 2026, 7:31 PM

    How do you know if a machine translation from one human language to another is good, or even accurate at all? Unless you ask a human who is fluent and erudite in both languages you can't know. You're merely gambling that the translation isn't crap. Each translation. Every time you click that button. Any confidence you have in that machine translation is based on probabilities you have no possible way of knowing or even beginning to estimate.

    💬 1🔄 0⭐ 0
  • 💬 1🔄 0⭐ 0
  • Simon Formancarapace@mastodon.social
    Oct 9, 2026, 7:34 PM

    And yeah, this applies to the little "translate this toot" links that use machine translation.

    Who fucking knows if those translations are good or not? Or if they even have anything to do with what the person actually tooted?

    💬 1🔄 0⭐ 0
  • Simon Formancarapace@mastodon.social
    Oct 9, 2026, 7:36 PM

    No one. That's who.

    You're rolling dice you don't even get to see.

    There's no dungeon master behind the screen backstopping the story. You're in digital dream land basking in made up bytes and phony pixels.

    "DANGER WILL ROBINSON! DANGER!"

    💬 1🔄 0⭐ 0
  • Simon Formancarapace@mastodon.social
    Oct 9, 2026, 7:41 PM

    "parts of medicine"

    Oh hell no.

    As a person who has encountered Reiki our concepts of medicine are woefully out of date.

    For a few things, like surgery (again, mostly physical kinetic with well defined decision procedures and feedback!) yeah robot surgeons will probably become a thing.

    For everything else...

    Love is a panacea. Awareness with compassion is a panacea.

    en.wikipedia.org/wiki/Panacea_

    💬 1🔄 0⭐ 0
  • Simon Formancarapace@mastodon.social
    Oct 9, 2026, 7:45 PM

    Love is a panacea.
    Awareness is a panacea.

    > panacea, noun, A remedy for all diseases; a universal medicine; a cure-all

    The last thing we need in medicine is a bunch of machines getting in the way of TLC.

    💬 1🔄 0⭐ 0
  • Simon Formancarapace@mastodon.social
    Oct 9, 2026, 7:48 PM

    This also is exactly the sort of thing I mean when I say "Please do not use computers to intensify the stupid."

    E.g. instead of asking for a cure for cancer, why not ask it how we can stop giving ourselves so much fucking cancer? (And also cure cancer too, duh, but you get what I'm saying right?)

    💬 1🔄 0⭐ 0
  • Simon Formancarapace@mastodon.social
    Oct 9, 2026, 7:51 PM

    If you ask your LLM buddy "How can I improve my health?" would it tell you about Reiki (and a couple of other things I could mention.)

    If it doesn't, and people are relying on it, that makes it a kind of prison of conventionality, eh?

    It's intensifying the stupid rather than getting at the core problem.

    💬 1🔄 0⭐ 0
  • Simon Formancarapace@mastodon.social
    Oct 9, 2026, 7:52 PM

    Instead of replacing all human drivers with robots why not redesign things so that we don't have to drive so much? (And yeah, duh, also robot drivers, once they're reliable and don't kill people!)

    💬 1🔄 0⭐ 0
  • Simon Formancarapace@mastodon.social
    Oct 9, 2026, 7:53 PM

    Instead of making better prosthetics and wheelchairs, why not human limb and organ regeneration? (And yeah, better prosthetics and wheelchairs too.)

    💬 1🔄 0⭐ 0
  • Simon Formancarapace@mastodon.social
    Oct 9, 2026, 7:54 PM

    Instead of an AI arms race between blocs, why not ask it how we can all get along and take care of each other in cooperation and harmony?

    (This one has not the "why not both?" energy, eh?)

    💬 1🔄 0⭐ 0
  • Simon Formancarapace@mastodon.social
    Oct 9, 2026, 7:57 PM

    In sum, once you get the hang of it, every single dumb or terrifying thing has this alternate "what if good idea instead" dual to it.

    It just takes a little intelligence to find it, and now the machines are intelligent enough to look for mutually beneficial solutions if only we would ask them to!

    (And yeah, ethics... We need to rebuild ML systems on ethical lines and use those instead of these Ouija board/slot machine fucking things.)

    💬 1🔄 0⭐ 0
  • Simon Formancarapace@mastodon.social
    Oct 9, 2026, 7:59 PM

    But again, as I'm always ALWAYS yammering on about: we already have all the solutions we need, we just have to notice that and apply them.

    💬 0🔄 0⭐ 0