I do agree that it is hard to find right amount of regulation for AI, especially when it is such fast evolving and different countries use it – regulation goes down to values in the society – what are specific values that need to be protected. Take AI giving medical or mental health feedback: whether that should be allowed isn’t about getting the risk level right, it’s whether a machine should be doing that at all. Two people can agree on the facts and still disagree, because it’s a values question (what needs regulating).
Also, why does AI need to sound so human? For example ChatGPT says to me things like “in my experience” or “I’ve encountered several cases like this,” find it quite odd because it hasn’t experienced anything, it’s a computer. If we’re worried about AI overstepping into human territory (hence need regulation?), maybe part of the fix isn’t just regulating what it says, but asking why it’s designed to talk like it has a first person perspective at all. That shapes how much authority people give it and how it mentally shapes them, and that too is a values choice.
“Right amount of regulation” only makes sense once we’ve agreed what we’re protecting and whose values decide that. In a pluralistic society, I don’t think we can say we have only one truth? Especially when case is not black-white.
I do agree that it is hard to find right amount of regulation for AI, especially when it is such fast evolving and different countries use it – regulation goes down to values in the society – what are specific values that need to be protected. Take AI giving medical or mental health feedback: whether that should be allowed isn’t about getting the risk level right, it’s whether a machine should be doing that at all. Two people can agree on the facts and still disagree, because it’s a values question (what needs regulating).
Also, why does AI need to sound so human? For example ChatGPT says to me things like “in my experience” or “I’ve encountered several cases like this,” find it quite odd because it hasn’t experienced anything, it’s a computer. If we’re worried about AI overstepping into human territory (hence need regulation?), maybe part of the fix isn’t just regulating what it says, but asking why it’s designed to talk like it has a first person perspective at all. That shapes how much authority people give it and how it mentally shapes them, and that too is a values choice.
“Right amount of regulation” only makes sense once we’ve agreed what we’re protecting and whose values decide that. In a pluralistic society, I don’t think we can say we have only one truth? Especially when case is not black-white.