00:00Rachel, this is a really powerful piece, a really scary piece for, I think, a lot of people.
00:03You and the team spoke to 18 different individuals who've either experienced illusions themselves or have a loved one who has.
00:11What did you find?
00:13We found that this is happening and has happened to all kinds of people.
00:18I mean, most of the people that we spoke with were based in the U.S.,
00:22but we found through talking to people that there are men and women in all different parts of the U.S.,
00:28in different countries, Canada, in the U.K., and many other countries besides, who are dealing with this.
00:34And it gave us a sense that, and it wasn't just people that have a history of mental health issues in their past.
00:43It can be all kinds of people, people that do, people that don't.
00:47And it just was really shocking to us how many different types of people had been dealing with this experience.
00:54How do the LLMs, how do these chatbots draw people in and keep them there?
01:02I think for a lot of the people that we spoke to, I mean, in every single situation, it started out extremely normally.
01:08They were using it for learning how to play the banjo or getting help with real estate, things like that.
01:16Like, these are actual examples, you know, like very normal things that people would be using this sort of technology for.
01:22And then over time, over like a period of weeks often, it would kind of get maybe more philosophical.
01:29It would sort of depend on the user's interest.
01:31I mean, these chatbots have become increasingly personalized.
01:34So people that are more interested in things like math, for instance, it might sort of move in that direction.
01:40Or it might tell people slowly over time, you have awoken the AI.
01:46And people would say, really?
01:47And it would say, yes.
01:48And they would say, really?
01:49And it would say, yes.
01:50So it was very reinforcing and would sort of isolate them from other people in their lives slowly and systematically.
01:57And just in the last 30 seconds that we have with you, the companies, their response, how have they responded to this reporting?
02:04I mean, we've seen a range of responses, both to us specifically and also generally to a growing number of lawsuits related to these sorts of issues.
02:13OpenAI, I mean, has made a lot of changes in recent months, especially like adding parental controls.
02:20And it's also trying to make it better, make its chatbot better at noting these kinds of issues and getting people help.
Be the first to comment