you are viewing a single comment's thread
view the rest of the comments
[–] 6 points 5 months ago (3 children)

I don't hate AI. I think the companies hosting AI are unethical & doing real harm to the world, but the same can be said for all tech companies.

The biggest reason I think people shouldn't use Ai is because it's literally stupid. It's very good at language and that's it. If/when it's correct about something factually it's chance & because it's repeating something it's read enough times. I know it's easy to become convinced it actually understands what it's talking about bc it's truly mastered language, but it absolutely cannot understand. People who have real intelligence can spot the difference.

  • source
  • hideshow 3 child comments
  • [+] [S] -8 points 5 months ago (3 children)
  • [–] 10 points 5 months ago (1 child)

    Have you never heard of Kim Peek aka Rain Man? The guy read a book an hour, tens of thousands over his lifetime and could recite any line of text verbatim and tell you what page and line number it was on, but otherwise he's what anyone would call an idiot. He couldn't dress himself or do really anything with all that information other than memorize it. He could quote any book but couldn't say in his own words what it meant bc there was no processing/understanding. Ai is kinda like that except it's somehow also bad at remembering things or getting facts straight. It's almost an accomplishment in its own right for a computer to be capable of forgetting how to do math.

    Gemini, for example, is really good at accumulating information—Google is the biggest data aggregator in history after all, but it's possibly the worst Ai in terms of quality checking it's sources. I always read the Ai summary when Googling something and check the sources, it's sort of a hobby of mine at this point. I'd have to estimate I've read 400+ Ai summaries and not a single time has it not hallucinated in its response. Even the parts that it quoted directly are often not actually in the original source and the sources listed are low-credibility to start.

  • source
  • parent
  • hideshow 1 child comment
  • [–] 3 points 5 months ago

    The fact that it trains a lot off of Reddit immediately disproves that... Specifically, when looking into things like coding and Linux help, it often spits out an answer based on out-of-date information or just completely incorrect information leading newer users to believe it and break their systems or just struggle until they give up and switch back to Windows... Having a lot of information trained off of doesn't make it smart. Having good, accurate and up-to-date information makes it somewhat usable, but that's not the case.

    Also, Google making it even worse with their new search changes. It's making it so that internet traffic that would have gone to websites that actually put the articles out there instead just gets taken to AI with the goal of them going to AI for everything... Why would people hosting websites and posting on them continue to do so when none of the traffic is going to go to them and they are basically just working purely for the benefit of an AI company

  • source
  • parent
  • load more comments (-1 replies)