Incidentally, one thing that might possibly work on humans is a moral argument: that it’s wrong to keep the AI imprisoned. How to make this argument work is left as an exercise to the reader.
Incidentally, one thing that might possibly work on humans is a moral argument: that it’s wrong to keep the AI imprisoned. How to make this argument work is left as an exercise to the reader.