News for hackers. Stuff that matters.
Re:AI chatbot 'encouraged' man to sacrifice himself to stop ... by MuffinFlavored on Saturday April 01, 2023 @02:38PM (#35402807)
> models trained on the wildness of the Internet

The models are pretty "PG" + lots of guards on, no?

Re:AI chatbot 'encouraged' man to sacrifice himself to stop ... by lukevp on Saturday April 01, 2023 @02:40PM (#35402833)
Bing’s isn’t. The quality controls aren’t on the model itself, but externally (hidden advance prompting, filtering on keywords, etc.) it’s very unsophisticated and the easy to go awry. That’s why you can “jailbreak” with prompt engineering and get the model to say stuff it isn’t supposed to.
Re:AI chatbot 'encouraged' man to sacrifice himself to stop ... by nomel on Saturday April 01, 2023 @02:40PM (#35402831)
What is “the models”?
Re:AI chatbot 'encouraged' man to sacrifice himself to stop ... by ethanbond on Saturday April 01, 2023 @02:48PM (#35402892)
The root of this entire AI conversation is that the guard rails don't work super reliably, and nothing says all models have to have guard rails.