745
Don't rely on it for anything important (media.piefed.world)
submitted 11 months ago by to c/memes@lemmy.world
87 comments fedilink hide all child comments
 
you are viewing a single comment's thread
view the rest of the comments
[–] -4 points 11 months ago (11 children)

That's not one of the published use-cases for the AI you're parodying.

If you don't read the manual and follow the instructions, you don't get to complain that the app misbehaved. Ciao~

  • permalink
  • fedilink
  • source
  • hideshow 11 child comments
    •
  • [–] 1 point 11 months ago (8 children)

    While I understand what you're trying to say, it should be on the owner of the software to ensure that the AI won't confidently answer questions it isn't qualified to answer, not on the end user to review the documentation and see if every question they want to ask is one they can trust the AI on.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 8 child comments
    •
  • [–] -2 points 11 months ago (7 children)

    I disagree for similar reasons.

    There's no good case for "I asked a CHAT BOT if I could eat a poisonous mushroom and it said yes" because you could have asked a mycologist or toxicologist. The user is putting themself at risk. It's not up to the software to tell them how to not kill themselves.

    If the user is too stupid to know how to use AI, it's not the AI's fault when something goes wrong.

    Read the docs. Learn them. Grow from them. And don't eat anything you found growing out of a stump.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 7 child comments
    •
  • [–] 2 points 11 months ago (1 child)

    If someone calls up Bank of America's customer service and asks if they should eat a mushroom they found in their back yard, and the rep confidently tells them "yes", do you think the response should be "Well, it's not the rep's fault you listened to their advice, you should have known that Bank of America isn't a good source for mycology information", or "That rep should have said 'I don't know, ask someone qualified'"?

    I'd argue that it's at least 50% on the person who gave the advice.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 1 child comment
    •
  • [–] -2 points 11 months ago

    The customer is pushing the responsibility of protecting their own health onto someone else. It's not that other person's responsibility, it's yours. You don't get to sah "but I asked Timmy the 8th grader and he said yes" or "I asked an AI chatbot and it said yes" and then be free of responsibility. Protecting yourself is always your responsibility. If you get a consequence, it's because of what YOU did, not because of Timmy or ChatGPT. ciao ~

  • permalink
  • fedilink
  • source
  • parent
  • [–] 0 points 11 months ago* (last edited 11 months ago) (4 children)

    Also there are too many cases to cover, so it's impossible. Sure i do it for mushrooms but then i have to do it for other 100 things

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 4 child comments
    •
  • [–] -2 points 11 months ago (3 children)

    what do you mean?

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 3 child comments
    •
  • [–] 0 points 11 months ago (2 children)

    That someone who develop an AI can block their AI from anwering some types of questions but there are too many things than an user can ask, it's unrealistic to being able to cover everthing.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 2 child comments
    •
  • [–] -3 points 11 months ago (1 child)

    It's unrealistic to expect a software developer to predict every type of ridiculous question a user could ask their software. That's why they don't. Instead, the publish use cases, eg. "Use my app to do X in Y situation" or "this app will do Z repeatedly until A happens". Anything that falls outside of those use cases is an inappropriate use of the app, and the consequences are the fault of the user. Just read the docs, friend :) ciao

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 1 child comment
    •