629
« on: July 09, 2020, 10:16:59 pm »
Biases of what/how it learns, more like, as people probably won't be building an AI's psychology piece by piece, idea by idea, in the same sense as you might build a machine - rule-based AI approach plunged around the time of the AI Winter. As a pale example see neural networks, obviously not a full AI or even close but they show that's the type of approach being gravitated towards.
NG, you would probably like the book Superintelligence by Nick Bostrom (about how to contain a superintelligence suspected of being harmful) and Godel, Escher, Bach by Douglas Hofstadter.