So the neuroscientist in me is constantly amazed by what our brains can do effortlessly. Among my favorite brain features are: unconscious control of walking movements, recognizing music from a very short time series of notes, and most of all the creation of a stable egocentric mental image of our environment from the chaotic images built up on our retinas. So those are among the features. But certainly there are bugs also.
If we were designing a humanoid robot with general artificial intelligence what might we augment or delete from the human repertoire of mental characteristics? This is an interesting question for which knowledge of neuroscience is not a prerequisite. I’m certain that my economist friends would look to tweaking the time discounting bias. And my statistically-knowledgable colleagues might push our robot to be more Bayesian in its decision frameworks. And probably my psychologist collaborators might push for less dopamine-rush from social networking apps. But those are the obvious ones. What might the more subtle design changes consist of?