Conversation

Notices

1. Bernie (codewiz@mstdn.io)'s status on Tuesday, 21-Feb-2023 02:50:01 UTC Bernie

   This is both terrible and awesome 🙂

   #ChatGPT #OpenAI #machinelearning #AI

   • Bernie (codewiz@mstdn.io)'s status on Tuesday, 21-Feb-2023 03:33:35 UTC Bernie

     in reply to

     Chandra: I would like to open a new file: PHOENIX. Do you know what that means?

     SAL: There are 25 references in the current encyclopedia.

     Chandra: Which one do you think is relevant?

     SAL: The tutor of Achilles?

     Chandra: Very interesting, I didn't know that one. Try again.

     SAL: A fabulous bird reborn from the ashes of it's earlier life.

     Chandra: That is correct. And do you know why I chose that?

     SAL: Yes, because you have hopes that HAL could be reactivated.

     https://youtu.be/T2E7sxGAmuo

   • Bernie (codewiz@mstdn.io)'s status on Tuesday, 21-Feb-2023 19:24:24 UTC Bernie

     in reply to

     • mike

     @mike ChatGPT is definitely going too far in it's goofy attempts at human emulation, and the AI ethics rules are.making it comically bad.

     On the other hand, some people assume that all it can do is explained by papers on Transformers and GPTs from a few years ago, when in fact there's a huge leap in capability.

     This dialogue diverged immediately from the 2010 film script, and yet it is plausible and coherent. I'm playing the computer, ChatGPT is my instructor, Dr Chandra:

   • mike (mike@micro.mkp.ca)'s status on Tuesday, 21-Feb-2023 19:24:25 UTC mike

     in reply to

     @codewiz It uses so much fatuous speech. The contrivance of first person pronouns, which it doesn’t genuinely understand. The apologies and thank-yous for things that it does not have the capacity to remember outside the context of a single conversation thread. It affects emotions even while telling you that it cannot behave with bias or have opinions. It’s all a ridiculous set of contradictions rooted in stupid science fiction tropes instead of the actual technology’s fundamental reality. I hate it so much. Its outputs are designed to an aesthetic philosophy that I 100% oppose. Computers should talk like computers, if they must talk at all.

   • Bernie (codewiz@mstdn.io)'s status on Wednesday, 22-Feb-2023 19:40:35 UTC Bernie

     in reply to

     • mike

     @mike Yes, that's a good analogy. I think we can agree that chatbots will continue to improve, it's just a matter of time and money.

   • mike (mike@micro.mkp.ca)'s status on Wednesday, 22-Feb-2023 19:40:37 UTC mike

     in reply to

     @codewiz It’s too much like talking to someone with profound psychological impairment, I can’t deal.

   • Bernie (codewiz@mstdn.io)'s status on Wednesday, 22-Feb-2023 19:41:54 UTC Bernie

     in reply to

     • mike

     Last night I tried to make #ChatGPT respond in Japanese, but using only hiragana with spaces between words. It partially missed, so I kept adding details, and asking for double-width spaces, and spaces also around the particles... but then it would forget the previous rules 🤦♀️

     @mike