• @[email protected]
    link
    fedilink
    1410 months ago

    what AI outputs is what I think a majority of the people want anyways. They want an answer to something in most cases.

    When I ask a question I want a correct answer, not one that is merely statistically likely. Using an AI and not fact checking it means you will never know if the answer given to your question is true. The AI tells you what it thinks you want to hear, not what it knows is true, because it doesn’t know anything, it’s a pattern matcher.

    • @thantik
      link
      English
      -1
      edit-2
      10 months ago

      Well, when other people ask a question, they get a page of jumbled answers, and I think we’ve proven that Karen cannot be expected to “do her own research”. AI will give a more correct answer 99/100 times.

      I’d love to know what questions you’ve asked an LLM and gotten wrong answers from – with the exceptions of math and continuity questions, which we already understand LLMs have trouble with.

      • Arsecroft
        link
        fedilink
        510 months ago

        Not who you were speaking to, but there’s at least one infamous example about a mycology book on amazon that was presumably copy pasted from LLM output. It is the same issue my professors had with people who copy-pasted from wikipedia. It is mostly right, but there is no real reason to believe it unless you can check the answer, and many people don’t understand that.

        https://decrypt.co/154187/ai-generated-books-on-amazon-could-give-deadly-advice