3 ms·
Passing the Turing Test. Understanding consciousness.[1] Communicating human values to a computer so that it doesn't produce absurd judgments. VR simulations in
by SilasX 11y ago
Passing the Turing Test. Understanding consciousness.[1] Communicating human values to a computer so that it doesn't produce absurd judgments. VR simulations indistinguishable from reality.
[1] Edit: people have different standards for this, of course, but something like, "where have a global, working model of a large domain in which predictions of consciousness aren't a special case, and in which people who study it feel like it makes sense and is less mysterious".
- Retra 11y agoComputers that can make ethical, legal, and moral decisions. Or ones that can do all three better than humans can.
- andreasvc 11y agoSure computers can make such decisions, be it with /dev/random or some sophisticated classifier. But we'd want justifications, and it's difficult to define what that entails. Moreover, what does it mean for a computer to be "better" at this than humans? The whole problem is ill defined.
- Retra 11y agoYes, it is ill defined right now. Are you willing to bet that it can't be properly defined? Or even that is currently is properly defined, and you just haven't been exposed to the definition?
- andreasvc 11y agoI am saying it's not a technological problem, it's a social problem.
- Retra 11y agoYou're certainly not saying anything useful. I have personally made some headway on this problem, and I wouldn't be saying it was possible if I hadn't. Your inability to find a workable framework is more evidence to me of your lack of imagination than it is that the problem is fundamentally unsolvable. There's a finite number of consistent models to explore, and one of them will give us what we want. It doesn't matter if you want to call that a technological problem or a social problem; the world isn't actually divided in those terms. So saying "social problems" strikes me as starting with a shitty model of what we are capable of. It's also just corporate-talk and there's no reason to expect it to be valid in this domain. The best that distinction has ever done is prime somebody to employ an actual problem solving technique.
- andreasvc 11y agoAre we talking about the same thing? You have made a definition of morality that philosophers and other people will all agree on? Where do you get this "finite number of consistent models" from?
- Retra 11y ago>You have made a definition of morality that philosophers and other people will all agree on? No, I have one that works. I really don't care what any philosophers have to say about it. And if you think everyone has to agree to something before it can be useful, then I can see why you'd think this is a social problem. >Where do you get this "finite number of consistent models" from? There's a finite number of people, a finite number of configurations their brains have been in, and thus a finite number of models provided by those configurations.
- andreasvc 11y agoAgain, I could hand-wavingly claim that /dev/random "works". What does "one that works" mean? You may not care for philosophers, but making claims about morality is definitely within their territory, plus it's a philosophical claim in itself to state that philosophy is not relevant. The argument about finite brain configurations is not compelling. There's no reason why only brain states that have occurred would be relevant. Morality deals with hypotheticals and counterfactuals: what would be the moral judgment in this or that situation. I think it would be uncontroversial to claim that there are an infinity of such hypothetical situations, and each model of morality can deal with them differently, ergo you get an infinity of possible models.