I thought I’d answer this one real quick, before anyone asks.
Point 1: Wrong question.
Try this instead: how do you teach ethics to AI?
Well, how do you teach ethics to anything?
What do we know that we teach ethics to?
Children. We teach ethics to children.
How do we teach ethics to children?
We don’t. They learn it themselves. All you have to do is feed them information.
In other words, act ethically, and they learn.
Occasionally, they get confused and ask a question. But the question of questions is a question for another day.
So to teach AI ethics, you take a bunch of cases and then say “this one good; this one bad; this one good…”
Eventually, you give yourself a spectrum.
“This one superb… this one kinda meh…. this one……ooooh, I dunno. I’ll add in an option for ‘I dunno’.”
Everyone who fancies doing this, does this.
Then you aggregate the results, in various ways.
And then the AI takes it from there.
See, nowadays, they call this “machine learning”. That’s a very inefficient way to phrase it. A better way is this: “learning”. That’s just how learning works.
So that’s how you solve the AI ethics problem. You numbskulls. The answer was in front of you the whole time — literally screaming for your attention, in this case.
Or maybe you just feed them the Ethics and be done with it. I hear these computer folks love maths. I’m sure they’d dig it.
P.S. Hello, world!