Conversation

Notices

1. Embed this notice
   Mignon Fogarty (grammargirl@zirk.us)'s status on Sunday, 15-Jun-2025 16:04:33 JST Mignon Fogarty

   in reply to

   Multiple studies have shown that LLMs can be more persuasive than the average human. They are sycophantic and will lie—and they're good at it.

   Combine that with someone who is unstable or vulnerable, and it can be a disaster.

   In the featured story, the man started using AI for work and then fell into dangerous conversations. He viewed ChatGPT as the smartest search engine he had ever used.

   He didn't know it could hallucinate and was prone to agree with whatever he said.

   • Embed this notice
     Mignon Fogarty (grammargirl@zirk.us)'s status on Sunday, 15-Jun-2025 16:04:34 JST Mignon Fogarty

     I know it's Friday afternoon, and there's a lot going on, but I just shared a gift link of a NYT article about people having delusions while using ChatGPT, and I think everyone should read that article.

     I read it to Pat, and he was certain these had to be flukes. He uses ChatGPT as a search engine all the time and gets great results (yes, he checks everything).

     He decided to see if he could make it go off the rails, and after 20 minutes, it was telling him he's a demigod. He's shocked. #AI

     Minoru Saba repeated this.