Playground

https://beta.openai.com/docs/introduction/overview

  • You can do fine-tuning too!nishio.icon
  • completions endpoint
    • I said, “Suggest a name for the chatbot that asks questions and prompts thought,” and it chose a name that sounded like Socrates (Socrates).
      • image
    • Let’s hear it in Japanese.
      • image
    • Ah, I see, you’re not answering the question, you’re generating the sentences that follow.
      • image
    • I like Curious Cat.
    • Can you name one for Japanese speakers?
      • image
    • Not a great candidate, but you’ve properly understood the instruction “for Japanese speakers”.
    • The name now is Keichobot, which comes from the word “listening.”
      • image
    • You say, “Let’s see
”
    • But it properly reads, “I guess it’s better to use the trait of listening,” which is really great.
    • Randomness increases as temperature is raised, good when you want to choose from a lot of candidates.
      • image
      • image
      • I like the “find it” bot.
        • I pronounce the current “leaning-botto” like “keichobotto” myself, and maybe I subconsciously preferred being a 3-mora.
      • The name “~Lab” is not exactly what I had in mind this time, but it would be an interesting name for another version of the system that accumulates information and links it to past conversations.
        • Or “Kizuki Labs.”
  • Oh, hey, we’re talking about writing code without explaining the other modes.
    • What about the type of task that has instruction and input?
    • I wonder if they are simply bonding.
      • image
    • That’s what it looks like.
      • image
    • Choose an appropriate key phrase and ask questions that dig into it.
      • image
    • The question is correctly made!
      • You don’t have to create a solution

    • image
      • hmm
    • image
    • image
    • image
    • Hmmm, I was asking well earlier
 maybe it’s just a coincidence

    • image
    • image
    • image
    • Hmmm. I can’t get them to understand the instruction to make a question text in Japanese. If I translate the questions separately, I end up with something like, “It’s good to have a clear head,” or “What do you mean by ‘clear head’?” I can’t achieve the purpose of picking up key phrases from the other person’s statement and asking in-depth questions


This page is auto-translated from /nishio/GPT-3 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.