LLM Limitations and Hallucinations

แชร์
ฝัง
  • เผยแพร่เมื่อ 25 ธ.ค. 2024

ความคิดเห็น •

  • @Matrium0
    @Matrium0 4 หลายเดือนก่อน

    Can't believe this has only so little views - very good explanation!

  • @dubhd4r4
    @dubhd4r4 ปีที่แล้ว +2

    I'm an AI researcher, but this is a good explanation for the general public. A lot of people will be needed to be educated if these are to be widely used (which takes work but is better than only a few members of society having access)

    • @jonathankey6444
      @jonathankey6444 ปีที่แล้ว +1

      I don’t understand why we don’t add fact checking on top of the models. Obviously the models can’t do that, but can we not do some sort of post-processing, if you will, to tidy up its responses. My interactions with GPT-4 showed that it couldn’t even get correct facts that it invented itself, such as a fictional character’s name. It seems to me that it should be easy for a separate algorithm to at least review what the model previously established, let alone checking facts from the internet. I’m not an AI researcher at all, I just want to develop an AI app

    • @jonathankey6444
      @jonathankey6444 ปีที่แล้ว

      I wasn’t asking rhetorically. I’m genuinely curious, and since you’re an AI researcher, I figured you could help me understand it if you can spare the time

  • @tobiaslutz5874
    @tobiaslutz5874 ปีที่แล้ว +1

    This was a outstanding video, thank you!

  • @LokeshSharma-me5pg
    @LokeshSharma-me5pg ปีที่แล้ว

    yes, this is what people don't understand that it's a LLM and will be as good as the data it is trained on.
    well nice video, loved it

  • @Synechex
    @Synechex 3 หลายเดือนก่อน

    This is what my Evrostics Triad addresses.

  • @和平和平-c4i
    @和平和平-c4i ปีที่แล้ว +1

    Of course, there are not knowledge bases, there are language models. The general public is confused to say the least.

  • @touseefmehmood2179
    @touseefmehmood2179 7 หลายเดือนก่อน

    Here is I have a question if LLM models are generating hallucinations in their answered then how can someone refer these models for students and researchers?