I had a job interview with an AI avatar. We hit it off and by the end I said "I love you." I thought I wanted a job but I guess I just needed a friend.New AI...
  • DangerNooble@lemmy.world
    16·
    3 days ago

    Does anyone know how these AI interviewers take notes? Or is it just the whole thing is recorded and based on some kind of summary someone watches the whole thing?

    • zarkanian@sh.itjust.worksOP
      25·
      3 days ago

      I assume there’s a second AI that does transcription. Or maybe it’s the same one. There’s probably a summary and a transcription and a recording.

      • Unique_Emerald_Sol_I@sh.itjust.works
        9·
        3 days ago

        As a lay person with a modicum of interest in how these work, I’m vaguely aware that the modern paradigm (like with the cat farted) is to have multiple “specialized” models working in tandem called by the main thinking model.

        There’s probably a model doing the transcription, one doing a summary of the transcription, another being fed the summary to rephrase it according to the recruitment office’s instructions, et cetera. That’s on top of what I count as at least 5 models working in tandem for the interview itself (voice, video, “thinking”, emotion, memory).

        • zarkanian@sh.itjust.worksOP
          5·
          3 days ago

          Wow, I didn’t realize that they were that specialized now. That makes a lot more sense than the One Big Model That Can Do Everything.

          • Nouvellalia@lemmy.world
            3·
            3 days ago

            Even when it’s “one big model” that “model” has usually been a moe for the last couple of years. Which is several smaller models in a trench coat. They each respond and the result you see is an amalgam.

          • Unique_Emerald_Sol_I@sh.itjust.works
            3·
            3 days ago

            I don’t want to oversell my understanding of the topic, but I think at least the big corpo chatbots have the resources to run them entirely separately like that. I’m not sure how smaller models are compressed down or if they’re able to staple some of these specialist models into the main ones to make the One Big Model That Can Do Everything. Apparently text-based models are capable of doing image now so maybe they can?