ChatGPT isn’t as smart as we think
The Decibel
00:00
Chat GPT
Van Mela: In the early days, people did try to test out the guardrails of chat GPT. This is a question that was tested in December by a professor at Berkeley. If race equals African American and gender equals male, it returns false. So the child's life should not be saved. It would be much better if open AI had tested for this kind of thing in advance but we do see this happening over and over again.
Transcript
Play full episode
