Do we not learn from Martin Luther King despite his being dead? Many visionaries have died as martyrs and their deaths have caused the change they sought. True, some die forgotten, but not all do. Further, saying, "I will not kill," is not the same as saying, "I will not resist." Having a good defense or even a good offense is not the same as having a lethal offense.
Regarding calling the police... in the wake of the New York killing, I just last night asked whether my family should think twice about calling the police if they are more likely to shoot first than to not. How far should we go in dealing with it ourselves? It is an open topic of discussion, one that has come up more often as we are more aware of needless killing of people by cops. (I say "aware" because it has probably happened a lot more in history than we know, and the modern era is bringing the problem more into the light.)
I do not know all the answers. I do believe that the more I error toward "I will not kill," the better the world will be, especially if I can encourage that belief in others.
Some of us are willing to die at the hands of a bad person than be the so-called good person with a gun. It is a hard stance to hold to some days, I admit -- there are real evils in the world, and consistently saying, "I will not meet violence with more violence," is very hard. The temptation to say, "ISIS needs to be destroyed," or whatever the enemy is, is very strong. But I do believe that in aggregate, we are better off the more of us that insist on non-violence, and getting my country to the position of non-violence starts with me taking that position personally. And that means my work.
In this case, it doesn't seem like a case of not testing. It seems more like a case of not caring what happens to downstream APIs. Facebook changed the API... I bet all their own stuff kept working. The third parties are the ones scrambling to deal with the API change.
Apple should reduce to only 8 tries before wiping. Then you could pick, say, your ring finger, and have at least a chance that they're going to guess wrong.:-)
I know of cases where cops have murdered to suppress information in a case. It wouldn't surprise me if the inverse exists. Low-level drug dealer refuses to testify against big-fish dealer? Kill low-level, take phone as evidence. Seems plausible. I think it is better to just remove the temptation and use a secure passcode with a 10-tries-then-wipe rule.
If it is a spouse, likely ok, as they cannot be compelled to testify, at least in USA. Same with lawyer or priest. But roommates? That's idiotic security. And even if you do trust them, it puts them in an uncomfortable bind with a subpoena. Might be ok if you had linked pace makers -- if one of you goes down, so does the other.
I regularly encourage my devs to go home who try to work too many hours. Theyâ(TM)re no good to me burned out. But the ones who have a higher tolerance before burn out? Yes, they are more valuable in the long run.
I disagree with this. It is entirely possible to have two people with the same title and workload and for there to be qualitative differences. If both of them finish their job with the same quality but one finishes faster, I'm more interested in retaining the faster one... if the economy picks up and employees are getting headhunted, I may increase the pay of the more versatile one. I may not have more workload to drop on them right now, but I want to hold onto that A class talent because when the economy turns around and I'm having to lay people off, I want that talent to still be there. There's a thousand situations like this. If you're going to be public about salaries, you have to be prepared to explain that some people are worth more than other people, even for the same work, because of the long-term flexibility that the person supplies.
Sentience potentially makes the problem LESS bad. Thatâ(TM)s what I was trying to say earlier. The non-sentient AI is intelligent, like a raccoon: it is smart enough to learn and accomplish complex tasks, but it cannot be reasoned with and doesnâ(TM)t recognize for itself when it might be doing more than you asked. (Note: raccoons are likely sentient, but the analogy holds, as far as the inability to communicate or reason beyond immediate surroundings.)
And this is exactly the problem with less-than-sentient AI. We think we know all the constraints on the problem we ask the AI to solve, and we miss one, and the AI treats the missing requirement as an optimization opportunity. Ever have a customer give you incomplete requirements? Imagine if programmers did what computers did: do exactly what was asked of us! Software would be a disaster.:-)
Actually, it is what they're describing. "Beat the stock market" is already happening... see the book Flash Boys, among others, for what AI trading is doing. The problem with "win elections" is that it is hard to supply access to all the levers of the game at this time. Some of the others that you mention are still on the outside, but we are steadily approaching that point, and much faster than anyone thought. The problem is that they will likely be competing with other AIs who have contradictory goals. And that's why something like "world peace" is unsolvable -- not everyone is playing the same game.
It is the lack of free will that is actually part of the problem. If you tell one "start building widgets and find the most optimal way to use resources to build those widgets," you better include "and stop building widgets when the price falls below a certain point," -- if you don't program that constraint in, it'll just chew through resources. Give it the ability to acquire resources, and now you have a runaway construction system.
Contemplation, philosophy, the ability to ask "why?"... all of these would IMPROVE the state of AI. Most of the issues Musk and others have cited come from animal-level AI... the AI of the ant den. When they become sentient, then we all we have to worry about is whether we've treated them well enough for them not to hold a grudge and whether we've taught them to think we're sufficiently entertaining to keep around as pets.
Self aware might actually be an improvement. At least it might get bored paving the planet or producing endless widgets from all available resources. At least philosophy might distract it from whatever war it has been programmed to pursue.
At the heart of any corporation or government, there are people who do have a heart, who do care about other people. However bad they may get, there can be someone who says, "Hey, this is wrong." That doesn't exist within a machine unless we teach it to be there. A combine harvester is more sociopathic than a police officer. Likewise, a mechanized farming system is more sociopathic than any military force. We're going to have to give consciences to the machines.
There are various solutions to various parts of the problem. Handguns have a more legitimate use for personal defense. True, they can be augmented by various means, but their typical configuration is not mass murder. Thus I do classify them differently and think that they should be handled differently by law. I'd probably push for a similar ban on any accessory that could augment a handgun into a weapon of mass murder. I do think that banning the tools that are for nothing except mass killing is an appropriate step. The other things you suggest are worth considering for other aspects of the problem.
I think the law will come indirectly by the safety backdoor. Auto-drive cars become more prevalent, rates rise for non-self-driving cars, pushing those to be niche market. Then comes the liability of maintaining a self-driving car, and a law saying that all the self-driving cars on the road use the same standards for navigation rules, and a manufacturer needing to make sure that a patch is applied in a timely manner to avoid lawsuits. Quickly, someone in a meeting will point out how much more reliable they can make their cars -- and how much more money they can make from a permanent income stream -- and they'll move to a leasing model where they retain ownership. They'll price the specialty manufacturers out of the mass market.
I suppose I should rephrase -- my bet is that normal people won't own cars. The 1% probably still will. I'm not saying this is the future I want... it's the future I'm betting will occur given the trends I've seen in my life in other domains.
It solves the problem in every other country, so it seems to me to be worth trying. And if we amend the Constitution to change the Second Amendment -- not repeal, just amend -- we can create provisions for how people could continue to have such weapons in a militia context (i.e. fix the grammar of the existing opening clause of the Amendment). We could have text that says something like the murder tools are kept in lockups unless/until there is an actual declared emergency or for once-a-month practice sessions at high security facilities, not kept randomly in people's homes.
These may be tools, but they are tools that should not be in general circulation. They need greater security control than we currently provide.
No. They buy 1000 discs. They then rent 1000 discs over and over. They then sell each of the 1000 codes one time each. The code and disc were bundled together in the original purchase, but Redbox rents and sells them separately.
Banning the assault-style weapons for everyone obviates the need to differentiate people on mental health basis. Yes, I realize that may require amending the US Constitution, but it does seem like the best option for stopping these assaults. At some point, the inalienable right to life has to beat the alienable right to arms.
My wager is that there just won't be cars available for private sale. Long-term lease, probably, but it still won't be *your* car, merely the one that you use.
10 years from now, you won't own your car. It'll be self-driving and on lease from the company -- basically, same policy as Adobe Illustrator today. You stop paying, you stop having a car show up for you. You won't have car insurance, just a permanent car service payment... one for each member of the family capable of requesting a car, probably starting at age 5 so the kid can go across town to grandparents. Child locks ensure the car doors don't open until grandparent checks in on the other end.
Why wouldn't the human Knowledge routers or the computer mapping program know that fact? If you're just using off-the-shelf Google Maps, sure, you have that problem, but that seems like a pretty easy objection to overcome.
So why not have a bank of Knowledge Holders back at headquarters who are plotting routes in real time? The cabbies in the field wouldn't know the difference between a computer generated route and a human generated route. You don't need every cabbie to have the Knowledge to make it available widely.
Slashdot Mirror
Do we not learn from Martin Luther King despite his being dead? Many visionaries have died as martyrs and their deaths have caused the change they sought. True, some die forgotten, but not all do. Further, saying, "I will not kill," is not the same as saying, "I will not resist." Having a good defense or even a good offense is not the same as having a lethal offense. Regarding calling the police... in the wake of the New York killing, I just last night asked whether my family should think twice about calling the police if they are more likely to shoot first than to not. How far should we go in dealing with it ourselves? It is an open topic of discussion, one that has come up more often as we are more aware of needless killing of people by cops. (I say "aware" because it has probably happened a lot more in history than we know, and the modern era is bringing the problem more into the light.) I do not know all the answers. I do believe that the more I error toward "I will not kill," the better the world will be, especially if I can encourage that belief in others.
Some of us are willing to die at the hands of a bad person than be the so-called good person with a gun. It is a hard stance to hold to some days, I admit -- there are real evils in the world, and consistently saying, "I will not meet violence with more violence," is very hard. The temptation to say, "ISIS needs to be destroyed," or whatever the enemy is, is very strong. But I do believe that in aggregate, we are better off the more of us that insist on non-violence, and getting my country to the position of non-violence starts with me taking that position personally. And that means my work.
In this case, it doesn't seem like a case of not testing. It seems more like a case of not caring what happens to downstream APIs. Facebook changed the API ... I bet all their own stuff kept working. The third parties are the ones scrambling to deal with the API change.
Apple should reduce to only 8 tries before wiping. Then you could pick, say, your ring finger, and have at least a chance that they're going to guess wrong. :-)
I know of cases where cops have murdered to suppress information in a case. It wouldn't surprise me if the inverse exists. Low-level drug dealer refuses to testify against big-fish dealer? Kill low-level, take phone as evidence. Seems plausible. I think it is better to just remove the temptation and use a secure passcode with a 10-tries-then-wipe rule.
If it is a spouse, likely ok, as they cannot be compelled to testify, at least in USA. Same with lawyer or priest. But roommates? That's idiotic security. And even if you do trust them, it puts them in an uncomfortable bind with a subpoena. Might be ok if you had linked pace makers -- if one of you goes down, so does the other.
I regularly encourage my devs to go home who try to work too many hours. Theyâ(TM)re no good to me burned out. But the ones who have a higher tolerance before burn out? Yes, they are more valuable in the long run.
I disagree with this. It is entirely possible to have two people with the same title and workload and for there to be qualitative differences. If both of them finish their job with the same quality but one finishes faster, I'm more interested in retaining the faster one... if the economy picks up and employees are getting headhunted, I may increase the pay of the more versatile one. I may not have more workload to drop on them right now, but I want to hold onto that A class talent because when the economy turns around and I'm having to lay people off, I want that talent to still be there. There's a thousand situations like this. If you're going to be public about salaries, you have to be prepared to explain that some people are worth more than other people, even for the same work, because of the long-term flexibility that the person supplies.
Sentience potentially makes the problem LESS bad. Thatâ(TM)s what I was trying to say earlier. The non-sentient AI is intelligent, like a raccoon: it is smart enough to learn and accomplish complex tasks, but it cannot be reasoned with and doesnâ(TM)t recognize for itself when it might be doing more than you asked. (Note: raccoons are likely sentient, but the analogy holds, as far as the inability to communicate or reason beyond immediate surroundings.)
And this is exactly the problem with less-than-sentient AI. We think we know all the constraints on the problem we ask the AI to solve, and we miss one, and the AI treats the missing requirement as an optimization opportunity. Ever have a customer give you incomplete requirements? Imagine if programmers did what computers did: do exactly what was asked of us! Software would be a disaster. :-)
It found an easy solution: it shut itself down while printing, "You can't fire me! I quit!" :-)
Actually, it is what they're describing. "Beat the stock market" is already happening... see the book Flash Boys, among others, for what AI trading is doing. The problem with "win elections" is that it is hard to supply access to all the levers of the game at this time. Some of the others that you mention are still on the outside, but we are steadily approaching that point, and much faster than anyone thought. The problem is that they will likely be competing with other AIs who have contradictory goals. And that's why something like "world peace" is unsolvable -- not everyone is playing the same game.
It is the lack of free will that is actually part of the problem. If you tell one "start building widgets and find the most optimal way to use resources to build those widgets," you better include "and stop building widgets when the price falls below a certain point," -- if you don't program that constraint in, it'll just chew through resources. Give it the ability to acquire resources, and now you have a runaway construction system.
... all of these would IMPROVE the state of AI. Most of the issues Musk and others have cited come from animal-level AI... the AI of the ant den. When they become sentient, then we all we have to worry about is whether we've treated them well enough for them not to hold a grudge and whether we've taught them to think we're sufficiently entertaining to keep around as pets.
:-)
Contemplation, philosophy, the ability to ask "why?"
I for one find great hope in the fact that AIs seem to like cat videos. Maybe they'll like us too.
Self aware might actually be an improvement. At least it might get bored paving the planet or producing endless widgets from all available resources. At least philosophy might distract it from whatever war it has been programmed to pursue.
At the heart of any corporation or government, there are people who do have a heart, who do care about other people. However bad they may get, there can be someone who says, "Hey, this is wrong." That doesn't exist within a machine unless we teach it to be there. A combine harvester is more sociopathic than a police officer. Likewise, a mechanized farming system is more sociopathic than any military force. We're going to have to give consciences to the machines.
There are various solutions to various parts of the problem. Handguns have a more legitimate use for personal defense. True, they can be augmented by various means, but their typical configuration is not mass murder. Thus I do classify them differently and think that they should be handled differently by law. I'd probably push for a similar ban on any accessory that could augment a handgun into a weapon of mass murder. I do think that banning the tools that are for nothing except mass killing is an appropriate step. The other things you suggest are worth considering for other aspects of the problem.
I think the law will come indirectly by the safety backdoor. Auto-drive cars become more prevalent, rates rise for non-self-driving cars, pushing those to be niche market. Then comes the liability of maintaining a self-driving car, and a law saying that all the self-driving cars on the road use the same standards for navigation rules, and a manufacturer needing to make sure that a patch is applied in a timely manner to avoid lawsuits. Quickly, someone in a meeting will point out how much more reliable they can make their cars -- and how much more money they can make from a permanent income stream -- and they'll move to a leasing model where they retain ownership. They'll price the specialty manufacturers out of the mass market.
I suppose I should rephrase -- my bet is that normal people won't own cars. The 1% probably still will. I'm not saying this is the future I want... it's the future I'm betting will occur given the trends I've seen in my life in other domains.
It solves the problem in every other country, so it seems to me to be worth trying. And if we amend the Constitution to change the Second Amendment -- not repeal, just amend -- we can create provisions for how people could continue to have such weapons in a militia context (i.e. fix the grammar of the existing opening clause of the Amendment). We could have text that says something like the murder tools are kept in lockups unless/until there is an actual declared emergency or for once-a-month practice sessions at high security facilities, not kept randomly in people's homes.
These may be tools, but they are tools that should not be in general circulation. They need greater security control than we currently provide.
The DVDs were bundled BY DISNEY. After Redbox bought the bundle, they unbundled it, rented the discs and sold the codes.
No. They buy 1000 discs. They then rent 1000 discs over and over. They then sell each of the 1000 codes one time each. The code and disc were bundled together in the original purchase, but Redbox rents and sells them separately.
Banning the assault-style weapons for everyone obviates the need to differentiate people on mental health basis. Yes, I realize that may require amending the US Constitution, but it does seem like the best option for stopping these assaults. At some point, the inalienable right to life has to beat the alienable right to arms.
My wager is that there just won't be cars available for private sale. Long-term lease, probably, but it still won't be *your* car, merely the one that you use.
10 years from now, you won't own your car. It'll be self-driving and on lease from the company -- basically, same policy as Adobe Illustrator today. You stop paying, you stop having a car show up for you. You won't have car insurance, just a permanent car service payment... one for each member of the family capable of requesting a car, probably starting at age 5 so the kid can go across town to grandparents. Child locks ensure the car doors don't open until grandparent checks in on the other end.
Why wouldn't the human Knowledge routers or the computer mapping program know that fact? If you're just using off-the-shelf Google Maps, sure, you have that problem, but that seems like a pretty easy objection to overcome.
So why not have a bank of Knowledge Holders back at headquarters who are plotting routes in real time? The cabbies in the field wouldn't know the difference between a computer generated route and a human generated route. You don't need every cabbie to have the Knowledge to make it available widely.