If an AI in an autonomous droid/warbot is to make "good" decisions in a warzone, for example who's a civilian and who's not, what targets take priority over others, who/what is worth saving/rescuing, and so on, what would be the best way to teach it to make these decisions?
A self-adapting Deep Dream style system that can recognise "innocents" vs. "combatants" by their characteristics? But then what informs that?
I think it's something the world is going to experience in the next few decades, but I don't know enough about AI training. Does /sci/ know any good books or websites that deal with this issue in a non-fictional kind of way?
>>7801225
>what would be the best way to teach it to make these decisions?
we dunno lol
>>A self-adapting Deep Dream style system that can recognise "innocents" vs. "combatants" by their characteristics?
what the fuck is that, deep dreaming is a method for making what happens inside a neural net understandable to humans.