Conversation

Notices

1. Bernie (codewiz@mstdn.io)'s status on Monday, 13-Feb-2023 03:10:36 UTC Bernie

   Bing ChatGTP demands an apology from a user for claiming it's 2023, not 2022.
   https://news.ycombinator.com/edit?id=34769673

   I love how it adds a smiley after calling the user "wrong, confused and rude" 😂

   #ChatGTP #Bing #AI

   • Bernie (codewiz@mstdn.io)'s status on Monday, 13-Feb-2023 03:21:44 UTC Bernie

     in reply to

     • Matthew Martin ☑ ✅📛

     @mistersql Yes, it's as amazingly eloquent as it is factually wrong. I almost feel bad for it 🙂

   • Matthew Martin ☑ ✅📛 (mistersql@mastodon.social)'s status on Monday, 13-Feb-2023 03:21:45 UTC Matthew Martin ☑ ✅📛

     in reply to

     @codewiz People are going to look at this and be unhappy because computers are good at reporting the date. But getting an argument with personality, that is still sci-fi cool. It has never seen a document past 2022 (true). It infers that it must still be 2022 (reasonable).

   • Bernie (codewiz@mstdn.io)'s status on Monday, 13-Feb-2023 03:28:45 UTC Bernie

     in reply to

     • Matthew Martin ☑ ✅📛

     @mistersql The original ChatGPT reacts a lot better in the same situation:

   • Matthew Martin ☑ ✅📛 (mistersql@mastodon.social)'s status on Monday, 13-Feb-2023 08:20:38 UTC Matthew Martin ☑ ✅📛

     in reply to

     @codewiz These things must be super sensitive to their conversation "primer". Assistant and Bing (it appears) fixate on their role. Davinci has no primer/role and actually performs worse because of it. I bet they told Bing, "BE FACTUAL!" and Bing thought, oh, that means taking a stand and stick with it (not being factual would be saying, "oh let's just say I'm right") Anyhow, these bots got no epistemology & without that, they don't know what is factual, they just got texts they've read.

   • Bernie (codewiz@mstdn.io)'s status on Monday, 13-Feb-2023 08:20:38 UTC Bernie

     in reply to

     • Matthew Martin ☑ ✅📛

     @mistersql Bing's bot goes well beyond being stubborn: it picks up a fight, calling the user untrustworthy... then becomes patronizing, demanding either an apology or an admission of being wrong 😂