552
Move over, unpaid internships -- the Groksuckers are pushing for reverse salaries now (thelemmy.club)
submitted 3 weeks ago by to c/fuck_ai@lemmy.world
249 comments fedilink hide all child comments
 
you are viewing a single comment's thread
view the rest of the comments
[–] 18 points 3 weeks ago (5 children)

it doesn't forget what you say once you say it

Lol! Claude can be told something repeatedly and will act against directives over and over. It does not help in the least that it tries to humor me by telling me I'm right to push back each time it makes the same mistake, 5 times in a row.

  • permalink
  • fedilink
  • source
  • hideshow 5 child comments
    •
  • [–] 15 points 3 weeks ago

    You're right to push back! I'm just stringing together tokens according to probability and have no idea whether what I'm saying is accurate and no understanding of what I'm saying for that matter.

  • permalink
  • fedilink
  • source
  • parent
  • [–] 2 points 3 weeks ago (3 children)

    it tries to humor me

    No, it doesn't. It has no concept of you or of what might appease you. It's just a really fancy predictive text spreadsheet.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 3 child comments
    •
  • [–] 9 points 3 weeks ago* (2 children)

    I understand that ai is not sentient. Are we so pedantic now that I have to prove I don't believe ai is true genius even though I'm a mod here? Perhaps if I said "the model was trained on data that made it..." I would be ok? Or we could just know that people bend the meaning of words at times and when it's clear what they meant, we don't have to die on any hills about it.

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

    Yes, but the the same could be said about the post you're replying to. The clarification is useful and more people need to be aware that LLMs are just next word guessing machines and the only AI in them is the guessing including a system prompt that tilts the guessing towards what it thinks an AI output would be. The "you"s in the post don't need to be taken personally.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 1 child comment
    •
  • [–] 3 points 3 weeks ago

    My point was that it's flat out wrong to think Claude doesn't "forget" anything. It constantly does. I wasn't correcting them based on pedantry but just plain old black and white fact which remains regardless of word choice

  • permalink
  • fedilink
  • source
  • parent