@lauren Brilliant. (And funny). I published a paper in 2014 on emerging security threats from #AI, especially #autonomous systems. I called the problem “the psychopathology of intelligent machines” and used your Hal example. (A fair number of people reacted in bizarre ways during my presentation (one even throwing a chair!) If you know enough about AI, it is not hard to see the mechanisms which would explain Hal’s maladaptive behavior. In the years since, I think every one of the new fault modes I predicted has actually occurred, either in lab tests or in the wild. You’re fair in saying today’s AIs cater to the dark side, but it is mostly from AI developers’ sloth, not malevolence.