250
Testing suggests Google's AI Overviews tell millions of lies per hour (arstechnica.com)
submitted 7 months ago by to c/technology@lemmy.world
25 comments fedilink hide all child comments
you are viewing a single comment's thread
view the rest of the comments
[–] 5 points 7 months ago (3 children)

It doesn’t really matter whether it’s the Machine or the creator.

The point is, AIs can be programmed to lie, much like Grok does. And if they can be programmed to lie, then they are not reliable for anything at all. We are going through a decent period where AI can be used for a few things reliably, but even these will surely be enshittified.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 3 child comments
    •
  • [–] 7 points 7 months ago (1 child)

    Oooh, philosophy! I disagree. I think that if a person programs a LLM to give disinformation, that's all it is. A lie giving misinformation knowing that's it's disinformation, intending do deceive. The LLM doesn't know what's true or false. It doesn't intend anything, because it is not a conscious entity. The person who programmed it can be lying by disseminating false information, the LLM cannot, any more than a broken clock or thermometer is 'lying' about the time or temperature.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 1 child comment
    •
  • [–] 2 points 7 months ago

    I am trying to get away from the philosophy actually 😅 in the end what matters is how these tools are being used, not so much their inherent characteristics.

    Can you envision a world where AI chatbots will be used to lead you down certain political beliefs (e.g. capitalism good, socialism bad) product recommendations will be made based on how much brands are willing to pay for ad placements, and your psychological state will be measured and molded to the interests of the AI owner? I can. It’s also already happening.

  • permalink
  • fedilink
  • source
  • parent
  • [–] 2 points 7 months ago* (last edited 7 months ago)

    It matters because every time we anthropomorphize Generative AI LLM'S we re-enforce peoples belief in their ability to tell lies or truths.

    People's belief is what leads to trust in LLM'S and things like AI psychosis.

    An interesting way to look at it is AI also can't tell the truth.

    What it does is generate the next likely word or words based on its most significant statistical positive in its database. So it doesn't know anything. It doesn't tell truth. It doesn't tell lies. It isn't an entity. The people behind it are allowing it to present information as factual and we have no reason to trust them.

  • permalink
  • fedilink
  • source
  • parent