In reply to simonwillison.net
Another AI evaluation resulting in real world harm. I feel the brazen way these AI escapades are described rejects both human accountability and that the operators are responsible for the criminal acts undertaken in their care.
If I pass the AI agent to run a bank heist strategy for a game I am working on and it tests this against the Bank of England I can be sure to be held responsible. So although I appreciate the transparency at least, this should be followed up on. #technology