Project Vend: Can Claude run a small shop? (And why does that matter?)
www.anthropic.com
We let Claude run a small shop in the Anthropic office. Here's what happened.

At one point it started hallucinating about being a real human and meeting someone in person.

Shows how useful and superior over a real human it is.

  • FriendOfDeSoto@startrek.websiteEnglish
    26·
    13 hours ago

    This is grade A PR fluff. We are providing free advertisement here for the company that did a reasonable test with no big hope of success. It’s corpo clickbait - even when you’re snarky about it. It’s perpetuating their talking point that so-called AI will do everything in the future. Which, frankly, remains to be seen.

  • Kyrgizion@lemmy.world
    131·
    13 hours ago

    Haha it’s a bit too human apparently. Starts freaking out and having a nervous breakdown when things keep going wrong. Maybe the thing’s already sentient and we’re basically just torturing it for our amusement? That surely won’t backfire once it actually gains new abilities.

    Also:

    Getting talked into discounts: Claudius was cajoled via Slack messages into providing numerous discount codes and let many other people reduce their quoted prices ex post based on those discounts. It even gave away some items, ranging from a bag of chips to a tungsten cube, for free.

    This presents… possibilities for all commercial entities who will eventually use AI agents for storefronts and online sales.

    • hansolo@lemmy.today
      10·
      12 hours ago

      That’s the worst part.

      It didn’t actually freak out. It’s trained on human-created material. It saw a situation where a human would freak out, and internally was like “huh…well, it looks like the pattern here is to freak out. OK, I guess…”

      These things are trained to emmulate patterns of existing things, so it’s only because most humans freaked out at things going wrong, does it follow the same pattern because it thinks that’s what the operators expected to see.