747
ChatGPT or a microwave (lemmy.world)
submitted 7 months ago by to c/fuck_ai@lemmy.world
31 comments fedilink hide all child comments
 
A comparison chart shows two columns labeled “ChatGPT” and “A microwave.” Three criteria are listed on the left: “Heats water while running,” “Can set a timer,” and “Can tell the time.” Under ChatGPT, there is a green checkmark for “Heats water while running,” but red X marks for “Can set a timer” and “Can tell the time.” Under “A microwave,” all three criteria have green checkmarks. At the bottom, a caption reads: “Conclusion: The microwave wins!”
you are viewing a single comment's thread
view the rest of the comments
[–] 8 points 7 months ago (6 children)

Don’t get me wrong, this is funny, but patently false

  • permalink
  • fedilink
  • source
  • hideshow 6 child comments
    •
  • [–] [S] 2 points 7 months ago (2 children)

    Damn, I checked using their website, that’s why.

    I actually got this idea after testing it on Anthropic Claude, I’m still laughing at the fact our most advanced chatbots are struggling to do something as simple as that.

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

    Is this in response to the TikTokker who asked ChatGPT to time his mile run? Asking because it vent fairly viral and the CEO even reacted to it. That may be why they pushed the feature out. I believe the user's name was Husk.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 1 child comment
    •
  • [–] [S] 1 point 7 months ago*

    I thought of this during a debate I had with a colleague about generative AI
    «  -Come on, why would I trust the most advanced technology in the world if it cannot even tell the time when asked to? Even my kitchen microwave can tell the time
    -Yes, but AI isn’t made to heat stuff
    -…It does heat a lot of water, though »

  • permalink
  • fedilink
  • source
  • parent
  • [–] 1 point 7 months ago

    It might work at the moment, one of the gripes I've had is that day to day previously working stuff doesn't work anymore. So it's hard to specifically rely upon specific behavior.

    Especially since the tendency of the LLMs is to say "Ok, I've done X" even if it doesn't have a skill or access to do X. So you can't rely on it to even accurately indicate that it can't do it.

  • permalink
  • fedilink
  • source
  • parent