Can we please, for once, make this not a "Trump is evil" thread ? Besides, yes, Trump is evil, but among the general US population there are far worse people than him, and despite what you probably think, together they have far, far more control over your life than Trump does. Private individuals are already implementing something worse than China's social credit score all over America. [1] Currently they're limited somewhat by technology, and I mean that in the worst possible way: they're more limited by what they think the technology can do, more so than by what is actually possible.
Building webcams that asses creditworthiness of store customers and the likelihood of returns is possible today. How long until it happens ? How long until we discover that badly dressed black people are often flagged by whatever tool comes out ?
Tracking people by their faces ? Been done, easy now (hell, opencv comes with it built in, so does YOLO). Tracking/identifying people by their movements (and I mean by how they walk, how they grab things, ...) ? Been done. Tracking people by the proportions of the limb lengths ? Been done. Taking a bad human tracker, feeding it into an LSTM and get vastly better performance when presenting it with video ? Been done.
And it gets much worse: outsmarting most of the human race ? Been done (let's be honest: humans have lost the Turing challenge. AIs are better at chatting up humans than other humans are). But no worries : humans still beat AIs at specific problems, and indeed humans still beat AIs ... well not at making AIs actually (genetic algorithms seem to beat everything else here). But humans still beat at "full system" AI architecture. For now.
And let's go further: how long until you can no longer talk to another human being ? Because that's exactly where voice recognition will lead. The 2030 version of Google Duplex : [2]
And of course, that makes the technology itself evil, not it's use by the executive branch per se. I'm sure they'll abuse it, but I guarantee they won't be the worst of your worries.
And it's MAD. Machine learning has joined the set of technologies that we must make sure we have. Otherwise, a conflict with, say, China is going to be thoroughly unpleasant (which would make them a lot more likely to start it as well). So we can't skip that.
And yet in this thread, nobody is talking about outlawing machine learning itself. And there certainly is no international effort to do that. Without that option, we might as well stop talking about measures we can take since they'll be absurdly ineffective.
You can "do no evil" but fact of the matter is since 1970 (and really since 1870) humans have been competing by robots. Humans can effectively no longer needlepoint, because they can't do so usefully, profitably, effectively, whatever you want to call it. Systematically the realm of humans is moving from everything to a niche.
Today, AI is a mostly-invisible salesperson making you purchase more stuff on Amazon.com, making you watch more advertisements on Microsoft/FB/Youtube/Google, and so on. It's literally choosing what to present to you in order to accomplish that, just like that friendly salesperson did when you walked into bed bath and beyond in 1995.
There is no solution here that is anywhere close to even being discussed.
My local Kroger is now taking video of the faces of every customer as they check out. To their credit they make it obvious by having a display screen at each register where you can see what they are recording, but I stopped shopping there for that reason.
I guarantee you there's 5000 private webcams and non-web cams pointed at the street for everyone listed there. 100 government ones.
Stopping shopping is not an effective punishment, and will not lead to stopping progress on machine learning. I would just ignore it. Besides, at this point, it's probably just a security cam. 24h and your video is gone (assuming the store wasn't attacked).
Building webcams that asses creditworthiness of store customers and the likelihood of returns is possible today. How long until it happens ? How long until we discover that badly dressed black people are often flagged by whatever tool comes out ?
Tracking people by their faces ? Been done, easy now (hell, opencv comes with it built in, so does YOLO). Tracking/identifying people by their movements (and I mean by how they walk, how they grab things, ...) ? Been done. Tracking people by the proportions of the limb lengths ? Been done. Taking a bad human tracker, feeding it into an LSTM and get vastly better performance when presenting it with video ? Been done.
And it gets much worse: outsmarting most of the human race ? Been done (let's be honest: humans have lost the Turing challenge. AIs are better at chatting up humans than other humans are). But no worries : humans still beat AIs at specific problems, and indeed humans still beat AIs ... well not at making AIs actually (genetic algorithms seem to beat everything else here). But humans still beat at "full system" AI architecture. For now.
And let's go further: how long until you can no longer talk to another human being ? Because that's exactly where voice recognition will lead. The 2030 version of Google Duplex : [2]
And of course, that makes the technology itself evil, not it's use by the executive branch per se. I'm sure they'll abuse it, but I guarantee they won't be the worst of your worries.
And it's MAD. Machine learning has joined the set of technologies that we must make sure we have. Otherwise, a conflict with, say, China is going to be thoroughly unpleasant (which would make them a lot more likely to start it as well). So we can't skip that.
And yet in this thread, nobody is talking about outlawing machine learning itself. And there certainly is no international effort to do that. Without that option, we might as well stop talking about measures we can take since they'll be absurdly ineffective.
You can "do no evil" but fact of the matter is since 1970 (and really since 1870) humans have been competing by robots. Humans can effectively no longer needlepoint, because they can't do so usefully, profitably, effectively, whatever you want to call it. Systematically the realm of humans is moving from everything to a niche.
Today, AI is a mostly-invisible salesperson making you purchase more stuff on Amazon.com, making you watch more advertisements on Microsoft/FB/Youtube/Google, and so on. It's literally choosing what to present to you in order to accomplish that, just like that friendly salesperson did when you walked into bed bath and beyond in 1995.
There is no solution here that is anywhere close to even being discussed.
[1] https://www.washingtonpost.com/business/economy/the-surprisi...
[2] https://www.youtube.com/watch?v=flLoSxd2nNY