1124
"What did students do before chatgpt?" (lemmy.blahaj.zone)
submitted 1 year ago by to c/curatedtumblr@sh.itjust.works
116 comments fedilink hide all child comments
 
you are viewing a single comment's thread
view the rest of the comments
[–] 14 points 1 year ago (4 children)

The dangerous part is that it makes stuff up and you won't have the knowledge to tell.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 4 child comments
    •
  • [–] 8 points 1 year ago (2 children)

    Exactly. I had a colleague who searched for a question about a retail math formula. The LLM returned a result that was close but slightly wrong. She spent two weeks with incorrect numbers for her baseline and as a result her forecasts were all wrong. When reviewing her numbers, everything was just a little wonky so I dug into it and discovered her mistake. She was absolutely dumbfounded the "AI" even could be wrong and tried to argue that I was incorrect. Dug out my old retail math cheat sheet and showed her the correct formula.

    I haven't used LLM's for anything since. Gotta validate all that shit anyways, so why use it at all?

    These things will be fantastic for taking my order at the drive thru and in a few other applications, but if you're trying to learn from them; don't.

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

    I was a kid in the era of separate pocket calculators, so I've heard so much of this song and dance before. Even with deterministic tools that always work barring user error you need to have enough understanding that you can tell when something is off and to properly frame the problem

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