Sometimes humans remind me of the “nearest unblocked strategy” problem with AI. If you have an AI that wants to do something horrible, and you hard-code a constraint “no, you cannot do literally X”, that doesn’t automatically turn it into a safe and friendly AI. Instead, it will probably use its intelligence to find something that is very close to X, but not literally X-as-you-specified-it, because it still has mostly the same reasons as before, only it also wants to follow the letter of your injunction.
Some humans have a desire to hurt others. If you make a law against it, they won’t magically turn into nice and friendly people. Under the (extremely optimistic) assumption that your law will be enforced reliably and impartially, they will find ways to hurt other people as much as possible in ways that do not technically violate your law.
Is hitting other people forbidden and punished? Then maybe shoving them down the stairs is the best tactic. Is any physical contact forbidden and punished? They maybe spitting, or yelling loudly right into their ear. And if they punch you back, or just push you away, now they are the criminal. Or just use your body to block them from going wherever they need to go, under the right circumstances that could be a lot of fun.
(Sadly, half of these ideas I didn’t even make up, but saw them used in YouTube videos, and I am pretty sure the aggressors would insist that they were the good ones, totally non-violent, etc.)
Now, I am also not saying that all behaviour should be allowed, just because any rules can be abused. Only that making the law does not solve the whole problem; you also need to exercise good judgment in enforcing it. You need to be aware than no matter how good your intentions were, someone is going to weaponize your rules. Otherwise you are part of the problem.
Sometimes I enjoy a combative debate, sometimes I am in a mood for a supportive safe space. What I hate is the passive aggression, emotional blackmail, and other things that happen when aggressive people are technically nice. Especially when calling a spade a spade is considered a violation of niceness (but poking you with the spade is not). My largest complaint against the so-called safe spaces is that they are actually not.
One hypothesis that surfaced lately is that some people seem to think that having categories for predatorial and parasitical behavior patterns itself will encourage bad outcomes.
My intuition would be that having categories would make these behavioral patterns legible and recognizable to others, potentially defanging them. Of course, as soon as they’re “spotted,” behaviors will shift evasively, but the core problem here seems to be reifying object-level behavior that at some historical point for some people, coincided with predation (e.g. “nice guy behavior”) rather than identifying the higher-level, abstract patterns.
Good pointer on nearest unblocked problem; that’s a very good analogy and I’ll have to think more about it. What kind of solutions might present themselves if we look at it through this frame?
I agree about passive aggression, but I’d also point out that “herding” culture is somewhat different from nurture.
Sometimes humans remind me of the “nearest unblocked strategy” problem with AI. If you have an AI that wants to do something horrible, and you hard-code a constraint “no, you cannot do literally X”, that doesn’t automatically turn it into a safe and friendly AI. Instead, it will probably use its intelligence to find something that is very close to X, but not literally X-as-you-specified-it, because it still has mostly the same reasons as before, only it also wants to follow the letter of your injunction.
Some humans have a desire to hurt others. If you make a law against it, they won’t magically turn into nice and friendly people. Under the (extremely optimistic) assumption that your law will be enforced reliably and impartially, they will find ways to hurt other people as much as possible in ways that do not technically violate your law.
Is hitting other people forbidden and punished? Then maybe shoving them down the stairs is the best tactic. Is any physical contact forbidden and punished? They maybe spitting, or yelling loudly right into their ear. And if they punch you back, or just push you away, now they are the criminal. Or just use your body to block them from going wherever they need to go, under the right circumstances that could be a lot of fun.
(Sadly, half of these ideas I didn’t even make up, but saw them used in YouTube videos, and I am pretty sure the aggressors would insist that they were the good ones, totally non-violent, etc.)
Now, I am also not saying that all behaviour should be allowed, just because any rules can be abused. Only that making the law does not solve the whole problem; you also need to exercise good judgment in enforcing it. You need to be aware than no matter how good your intentions were, someone is going to weaponize your rules. Otherwise you are part of the problem.
Sometimes I enjoy a combative debate, sometimes I am in a mood for a supportive safe space. What I hate is the passive aggression, emotional blackmail, and other things that happen when aggressive people are technically nice. Especially when calling a spade a spade is considered a violation of niceness (but poking you with the spade is not). My largest complaint against the so-called safe spaces is that they are actually not.
One hypothesis that surfaced lately is that some people seem to think that having categories for predatorial and parasitical behavior patterns itself will encourage bad outcomes.
My intuition would be that having categories would make these behavioral patterns legible and recognizable to others, potentially defanging them. Of course, as soon as they’re “spotted,” behaviors will shift evasively, but the core problem here seems to be reifying object-level behavior that at some historical point for some people, coincided with predation (e.g. “nice guy behavior”) rather than identifying the higher-level, abstract patterns.
Good pointer on nearest unblocked problem; that’s a very good analogy and I’ll have to think more about it. What kind of solutions might present themselves if we look at it through this frame?
I agree about passive aggression, but I’d also point out that “herding” culture is somewhat different from nurture.