RE: Early Horror Stories with AIs and Robots
You are viewing a single comment's thread:
Yes, they are. But they are provoked on purpose by AI researchers, to understand their limits and how to control their behavior dangerous behavior. But not all AI models will have these safeguards. This will be messy!
0
0
0.000
I missed that "provoked" part. So it's part of the test.
!PIZZA
!LOLZ
Not all of them. The incident with the robot going nuts that went viral wasn't planned, for sure.
But the last ones with the decisions to blackmail or leaving the executive to die, were part of a test from Anthropic, the creator of Claude AI, to see how AI models would react in certain limit situations.
Thanks for the clarification.