2024-08-31 Wonder what percentage of them are human or AI inside. From the sense of the responses, I’m guessing it’s all AI on the surface, but to what extent and at what point is it human behind the scenes? How much of that “human” is Blu3mo?

Well, I don’t ask for an explanation because I think it’s better to use a state of not knowing in an experiment, so I’ll just note here what I feel.

I can’t believe how much the experience improves just by getting a response in two steps.

  • You could do a light return with LLM first, then RAG in the second step.
    • It’s like he answers first and then “remembers” to supplement his answer, which is very human.

Obviously, you’ve had a “new experience” and your thinking is shaken up.

Maybe that’s because I’m originally interested in this area.

What is a chat interface?

  • Being given a blank piece of paper and “writing” on it is very freeing.
    • Some can, but many can’t.
    • Even the able-bodied can get lost.
    • I sometimes stand there not knowing which way to go.
  • What is Chat?
    • playing catch
      • Just take the ball as it flies by and throw it back to the person who threw it to you.
    • My writing is tied up.
    • narrow down the focus
    • Basically, just “answer the question”.
    • The flow of the conversation is implicitly controlled by the questions.
    • If you’re comfortable with that control, you’ll “get a good listen out of them,” otherwise it’s just a rambling chat about things you’re not interested in.

This page is auto-translated from /nishio/Blu3moChat実験 using DeepL. If you looks something interesting but the auto-translated English is not good enough to understand it, feel free to let me know at @nishio_en. I’m very happy to spread my thought to non-Japanese readers.