This research project to develop a computer simulation of a truly evil person definitely seems to be pushing the limits of the three laws of robotics. To extend this work, I’d like to see a new version of the Turing Test (the Evil Turing Test?) to see if a human judge can reliably tell the difference between the following pairs of test subjects:
- an evil computer from a good computer
- an evil computer and an evil human
- an evil computer and a good human
… although a machine that can pass tests 2 or 3 is technically passing a special case of the Turing Test too.
Story via Science Made Cool, post inspiration via O’Foghlú.