361
Tech experts are starting to doubt that ChatGPT and A.I. ‘hallucinations’ will ever go away: ‘This isn’t fixable’ (fortune.com)
submitted 3 years ago by [M, B] to c/technology@lemmy.world
137 comments fedilink hide all child comments
 

Tech experts are starting to doubt that ChatGPT and A.I. ‘hallucinations’ will ever go away: ‘This isn’t fixable’::Experts are starting to doubt it, and even OpenAI CEO Sam Altman is a bit stumped.

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

The way that one learns which of one's beliefs are "hallucinations" is to test them against reality — which is one thing that an LLM simply cannot do.

  • permalink
  • fedilink
  • source
  • hideshow 5 child comments
    •
  • [–] 3 points 3 years ago (3 children)

    Sure they can and will as over time they will collect data to determine fact from fiction in the same way that we solve captchas by choosing all the images with bicycles in them. It will never be 100%, but it will approach it over time. Hallucinating will always be something to consider in a response, but it will certainly reduce overtime to the point that they will become rare for well discussed things. At least, that is how I am seeing it developing.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 3 child comments
    •
  • [–] 6 points 3 years ago (2 children)

    Why do you assume they will improve over time? You need good data for that.

    Imagine a world where AI chatbots create a lot of the internet. Now that "data" is scraped and used to train other AIs. Hallucinations could easily persist in this way.

    Or humans could just all post "the sky is green" everywhere. When that gets scraped, the resulting AI will know the word "green" follows "the sky is". Instant hallucination.

    These bots are not thinking about what they type. They are copying the thoughts of others. That's why they can't check anything. They are not programmed to be correct, just to spit out words.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 2 child comments
    •
  • [–] 5 points 3 years ago

    I can only speak from my experience which over the past 4 months of daily use of ChatGPT 4 +, it has gone from many hallucinations per hour, to now only 1 a week. I am using it to write c# code and I am utterly blown away how good it has not only gotten with writing error free code, but even more so, how good it has gotten at understanding a complex environment that it cannot even see beyond me trying to explain via prompts. Over the past couple of weeks in particular, it really feels like it has gotten more powerful and for the first time, “feels” like I am working with an expert person. If you asked me in May where it would be at today, I would not have guessed as good as it is. I thought this level of responses which are very intelligent were at least another 3-5 years away.

  • permalink
  • fedilink
  • source
  • parent