4 ms·
I do not understand how the second paragraph of your reply is related to the first. Computer programs (currently) do not respond to incentives, nor do they reac
by geofft 9y ago
I do not understand how the second paragraph of your reply is related to the first. Computer programs (currently) do not respond to incentives, nor do they react to available resources in any way other than the way in which the resource provider explicitly changes them. So the idea of "deserving" does not apply, but it does apply to entities that respond to incentives or consume and make use of resources in complex/emergent ways.
If we had strong AI, then yes, certain strong AIs would be more deserving of resources from the commons than others - that's exactly what MIRI is working on.
And if you step back from "program" to "project" (meaning the whole system of code + the autonomous decisions of humans, including groups of humans working as companies, to continue developing the code), then yes, certain projects do deserve to be fixed more than others do. A particularly difficult crash in Python 3.6 is much more worth debugging than one in Python 2.4, which in turn is much more worth debugging than one in GW-BASIC, because the net benefit for humanity in exchange for the opportunity cost varies widely.
- jfaucett 9y agoMy point has to do with the notion of retributive justice i.e. that punishment should be inflicted because the criminal deserves it - in a quasi-religious sense i.e. he is evil or deserving of hell. This usually carries along notions of moral good and evil that are almost never grounded in measurements to give them real meanings. Conversely, I think people should think about crime in a more reinforcement learning approach i.e. rehabilitative. I view humans as machines essentially. If a machine breaks you fix it, but there is no emotional garbage attached. That's the concept I was trying to get across.
- s73ver_ 9y agoHumans are not machines, so trying to look at them that way is going to be flawed. Furthermore, there is no excuse for what they did. They knew it was wrong. They knew it was against the law. It was not an accident; it was a deliberate act. There's not really any rehabilitation that can take place here.
- geofft 9y agoI agree that punishment as retribution is a poor way to run a civilized society. But punishment as incentive structure fits the framework of "deserving" just fine - full participation in human society is a privilege, not a right, and extending that privilege to people who will hurt society is a poor use of resources. And credibly threatening not to extend that privilege will cause people to try to be the sort of person who "deserves" not to be punished. Viewing humans as machines doesn't work precisely because humans have agency. A machine generally does not decide to break, and a machine generally can be properly fixed (or if it can't, it's obvious). Attempting to convince a machine that it will suffer some negative consequence for misbehavior is unlikely to have any results at all. Meanwhile, especially for humans making decisions about corporate strategy (which means that they're in a culture that's pretty firmly incentivized towards "is this profitable"), credibly convincing them that they will suffer punishment is a great way to change their behavior and prevent them from deciding to do something unwanted. And one way to credibly convince people that they will suffer punishment for an action is to actually punish people who do the same thing. It's very unlikely this executive will make the same mistake again, punishment or no punishment. Rehabilitation isn't the point. But we need to make an example of them, and I say that without the slightest shred of emotion.
- jfaucett 9y agoBasically, I don't accept what you and most people accept - that humans have free will, we could have made a choice X instead of Y, etc, etc. I know its not a popular opinion, and I don't expect to convince you of my views in a couple sentences, but generally speaking I think the notion of punishment should always be viewed in the context of societal functioning. If we do as you suggest and punish an executive, does that actually lead to decreased occurances? If so by how much? What about the next time it occurs? Can we completely prevent another occurance? Can we decrease its probability? At what costs to the person and society? What caused it in the first place? These are all much more important questions to mold a system of punishment than our current focus which almost completely lacks any measurements and analyses and is still steeped in vague moralities derived from historical accident and our own intuitions which are perforated with biases and flaws.
- 9y ago