Slashdot is powered by your submissions, so send in your scoop

 



Forgot your password?
typodupeerror
AI The Media

Artificial Intelligence Is Killing the Uncanny Valley and Our Grasp On Reality (wired.com) 238

rickih02 writes: In 2018, we will enter a new era of machine learning -- one in which AI-generated media looks and sounds completely real. The technologies underlying this shift will push us into new creative realms. But this boom will have a dark side, too. For Backchannel's 2018 predictions edition, Sandra Upson delves into the future of artificial intelligence and the double edged sword its increasing sophistication will present. "A world awash in AI-generated content is a classic case of a utopia that is also a dystopia," she writes. "It's messy, it's beautiful, and it's already here."
"The algorithms powering style transfer are gaining precision, signalling the end of the Uncanny Valley -- the sense of unease that realistic computer-generated humans typically elicit..." the article argues.

"But it's not hard to see how this creative explosion could all go very wrong."
This discussion has been archived. No new comments can be posted.

Artificial Intelligence Is Killing the Uncanny Valley and Our Grasp On Reality

Comments Filter:
  • by Arbitary5664 ( 1979712 ) on Sunday December 17, 2017 @04:54AM (#55754447)
    Yeah, just a rhetorical question -- which of you, here, has 'lost your grasp on reality' due to any tech being produced, anywhere? I mean, acid, sure, but when was the last time Siri made you lose your god-damn-mind outside of a voice to text translation failure? People who claim "thaaaa AI is gunna keeell us all" should be laughed at, instantly, and mocked until they leave the community at large. "We cannot predict what will happen -- but here's my obviously accurate prediction!". Total bullshit.
    • by religionofpeas ( 4511805 ) on Sunday December 17, 2017 @05:23AM (#55754483)

      I think they just mean that you can't tell the difference between reality and fakery.

      • Yeah, I've been high before, who hasn't?

      • by c6gunner ( 950153 ) on Sunday December 17, 2017 @07:23AM (#55754705)

        Most people have never been able to do that. I mean, just look at how popular religion is.

        • by religionofpeas ( 4511805 ) on Sunday December 17, 2017 @07:46AM (#55754743)

          What if you're working for a big company, and your boss Skypes you from his house, and says it somewhat irregular, but there's an important invoice that needs to be paid right now. He's e-mailing you the invoice right now, and he assures you it's legit and urgent.

          Well, it's not your boss but a foreign hacker who used a couple of facebook photos to fake a live conversation.

          • No different than your boss sending you an email asking you to do the same thing. Except the email is spoofed and coming from a hacker.

            • by religionofpeas ( 4511805 ) on Sunday December 17, 2017 @08:18AM (#55754813)

              The number of people that would trust a video call with someone they know is substantially larger than people who believe a random e-mail, especially if both are coming from a home address.

            • Used to it vs. not (Score:5, Interesting)

              by DrYak ( 748999 ) on Sunday December 17, 2017 @09:26AM (#55754957) Homepage

              Currently, not a lot of people are used to put doubt into video (or even real-time face-to-face video) because the technology to fake it realistically enough has only started very recently to become cheap enough to be a worthy try for an attacker: And it will still be a little bit more time until it start getting used in real-time (basically once " ${price of renting cloud GPU time to run the neural net} ${money that can be made in such attempts}" ).

              Once awareness is raised, society will eventually adapt and only the most gullible will fall for the tricks while our successor on /. will wonder why not more people are using whatever authentication is the most common for video chatting.

              A bit like how a couple of decades ago, every body was aware of signature forging and wouldn't trust a simply hand written note, but would fall for attempt at phone-calls social engineering (i.e.: impersonating a general role by being a good actor, back at a time where the phone quality would barely let you recognise a voice reliably).

              • People still fall for crude e-mail and phishing scams. As the attackers get more sophisticated that will only increase. And if we all stop trusting each other, it'll have a great impact on our daily life.

                renting cloud GPU time to run the neural net

                For now. It won't be long before you can buy special GPU cards that are optimized for neural net processing that you can stick in your home PC. Nvidia already has one. Still expensive, but it's only a matter of time before the prices come down and capability increases.

              • Currently, not a lot of people are used to put doubt into video

                I know, I was watching some music video and the artist showed her nipple, and everybody got really exited, and then some mean people were going to fine the broadcaster, and then when they went frame-by-frame to extract the evidence, they realized there was stage tape over her nipple, all they ever showed was stage tape and there wasn't gonna be any fine. That was like, `86 or something.

                Before VHS and the pause button, there was no way to know what you were really seeing! And even with it...

                I was watching so

          • When it arrives without his digital signature, the message ID gets forwarded to IA as a spearfishing attack and they pull it off the server and dissect it to find out where it came from. And even if it does arrive with his digital signature, the message ID gets forwarded to IA as a spearfishing attack, because invoices never come to me, and my boss knows that.

            • because invoices never come to me, and my boss knows that.

              Obviously, they're going to target the person that deals with the invoices. It was an example. Feel free to come up with better examples that relate to your own life. E.g. wife calling up that she forgot the alarm code.

              • by Mal-2 ( 675116 )

                Shared alarm codes can be procedurally generated. If the procedure incorporates a time element (such as the current month), then it will automatically change as well. Sure it's security by obscurity, but it does mean that in this case you'd just have to say "It's December now, remember?" or something along those lines. To an attacker, that just indicates you change your codes every month, and won't directly give them entry. If she still "can't remember" something you've been doing for years, then I'd have t

              • Which means the recipient will be familiar with the ways their company actually handles invoices.
          • by Mal-2 ( 675116 )

            If you're suspicious and want to confirm it, you'd turn around and call him back, wouldn't you? Right now, it's a lot easier to spoof a call to you than to intercept calls going to your boss's house, but how long will that remain the case? Once the spoofers can spoof the confirmation too, it's all over.

            • Considering that many phone calls are now voip, I would not wonder about vulnerabilities in your router that can be exploited to reroute your calls elsewhere.

      • This is a fair point -- and oddly enough the litany of other AI articles has tainted the impression received from this one. I'd mod you up if I could.

      • If you dumb down a populace enough, they can't tell the difference anyway.

        The claims of anyone possesing a fully functional AI are laughable already, but I suppose if we keep telling folks we have them, they'll eventually believe it.

    • People who claim "thaaaa AI is gunna keeell us all" should be laughed at, instantly, and mocked until they leave the community at large.

      The problem with that is that the idea that AI will kill us all is perfectly legitimate. It's not the only possible outcome, but it is a possible outcome. History is full of examples of technically superior nations becoming dominant. What happens when it's not a nation on attack, but a pervasive technology that can self-replicate?

      • There's that but it doesn't even have to go that far; machine logic (I refuse to say "AI") guided by people will be bad enough. The machines don't need to "wake up" or develop their "own" agenda to be a threat to the human race; the humans controlling them can ensure that easily enough.
    • It looks like the author has a hard time with reality, short sightedness is both a blessing and a curse; like supporting child acused molesters that run for political office.
  • They're not signalling the end of it, they might well be crossing it.

    The uncanny vally will stil lexist even if we're capable of doing more than hitting the bottom of it square on.

  • AI vs AI (Score:2, Interesting)

    by Anonymous Coward

    If you can't tell it's AI generated just train another AI to recognize that it was generated.

  • Using Hill-Climbing [wikipedia.org] to escape the Uncanny Valley... it's such a bad play on words I can't look at it as ingenious.

    • by mikael ( 484 )

      Somehow it makes me think of the porn industry moving from Los Angeles to Las Vegas.

  • As most people seem to have a SEVERELY tenuous grasp on reality as it is. So what?

    As I see it, if there are any dangerous repercussions, it'll simply act as chlorination of the gene pool.

    • Re:DOOOOOM!!! (Score:4, Interesting)

      by DivineKnight ( 3763507 ) on Sunday December 17, 2017 @06:24AM (#55754579)

      Whether the world is doomed or not is immaterial; what matters is whether you will wake up one day to find it doing things you don't like (especially to you).

  • by ezdiy ( 2717051 ) on Sunday December 17, 2017 @06:04AM (#55754557)
    State of the art machine intelligence is "machine intuition". Bots copy and remix the corpus matched to context, but we're still miles away from an observable thread of thought. It will just mimic and more or less randomly remix the corpus - basically replay thoughts of somebody else. In a crude sense, it's just very sophisticated markov model, to the point it will reasonably past turing test on a youtube comment level

    This is not limited to text, a fashion or food instagram is even more trivial with current tools. Good example of this is Spiderman Elsa [reddit.com] (which I suspect is made with good old honest-to-god sweatshop labor, not a bot), but the model of social spam has shown an immense profitability potential already in a format far more sophisticated than appealing to lowest sexual urges.

    The good thing about this is that this will spell an early end to shallow internet memetics once advertising world discovers chatbots and context-aware media remix bots. No more need to bribe lowkey ecelebs to astroturf your product, when you can just unleash fake users in number. Even if the quality on average will be sub-par, statistically some will always get a traction if you spawn population large enough.

    It's a post-scarcity scenario for internet drivel in a cost model where people engaged in drivel for social bond and validation, both points being moot when it's a machine on the other end.

    This can lead to two possible outcomes:

    1. The cancer spreads, remember the south park episode about living ads? This is it. People will literally lose grasp on reality and will feel about adverts as if they were people..

    2. It's a chemo which will bring us back to 1993. Folks will recognize low effort posts lost all of its shreds of utility for validation, pushing the bar for social network posts a lot higher (low effort posts being implicitly assumed a bot when it becomes a common case).

    In either case, there will be constant market pressure for "better ads" as users adapt, there will be this arms race for ever better "living ad" until the bots start having so much grasp of context we'll enter a very weak GAI era.
    • by SuricouRaven ( 1897204 ) on Sunday December 17, 2017 @09:33AM (#55754971)

      Or it might be the end of comments sections. Consider this scenario:

      Someone develops and publishes a comment-bot AI. It's not a general-purpose AI, but you can configure it with a position to promote and point it to a site, and it will then start posting unique comments promoting the view, and posting rebuttals to anyone who opposes the view. It's not going to pass for human in a conversation, but in single posts it'll appear human most of the time.

      First thing that happens? Joe's Pizza unleashes a hundred instances to tell the world how great their pizza is. AI spam. But this is hard spam to get rid of, because it's constantly changing: This AI learns how anti-spam measures work. CAPTCHA tests get even more annoying for a while. But that's ok: The internet is used to spam. Joe's Pizza gets a lot of hate.

      Then an election rolls around. Say, a US presidential election.

      Suddenly, millions of instances appear - half of them promoting the Republicans, and half the Democrats. Comment threads all over the internet become fifty-pages of almost fully automatically generated text, flooding out any human voice. Both parties deny such underhanded techniques, of course - and perhaps even truthfully, as fingers are pointed to independent pressure groups or the governments of other countries as a possible source.

      Meanwhile, the Church of the Easily Offended gets their running. They set a few thousand running - their job is to identify 'inappropriate' material - anything that offends their religion, or standards of decency or of clean language - and submit reports or write angry letters to site operators. In an amusing irony, the church website shortly has to close their own comments section because of the millions of bots now searching the internet for church comments pages and posting about why Islam is the true religion.

      In the end the only option is to drop anonymous comments entirely, and tie any comments into verified accounts established with proof of identity.

      • by Kjella ( 173770 )

        Comment threads all over the internet become fifty-pages of almost fully automatically generated text, flooding out any human voice.

        You didn't need AI for that, without CAPTCHAs most comment fields would be overrun by bots already, even though it would be junk posts. And the nuclear arms race there continues.

        In the end the only option is to drop anonymous comments entirely, and tie any comments into verified accounts established with proof of identity.

        Maybe, but that alone doesn't make the public debate great. Very often it's taken over by extremists on both sides that aren't interested in a debate and are willing to post dozens of replies on a single issue. I've seen way too many comment fields essentially turn into a shouting match between the same dozen people or so. You still

        • And if you have that, AI pot shots wouldn't matter that much since they'd be very thin on substance.

          Except that you're replying to someone who postulated

          It's not going to pass for human in a conversation, but in single posts it'll appear human most of the time.

          And therein lies the problem. Once bots can reasonably approximate the median person on a forum, I think humans chatting in forums will rapidly disappear. Bots will be able to post far faster than humans, and once you have a couple of competing bots, the bots will have hashed the point out and will have moved on before the human gets done pecking at the keyboard.

          Now, I don't know if this is a bad thing or a good thing, to be honest.There are few places li

    • by epine ( 68316 )

      Sometimes I don't even know why Slashdot exists.

      Bots copy and remix the corpus matched to context, but we're still miles away from an observable thread of thought.

      The whole point of "the future is already here — it's just not very evenly distributed" is that "miles away" is not a viable unit of measure.

      Generative adversarial networks are the Ajax of DNN 1.0.

      I was reading O'Reilly's new book the other day, and he was talking about how little use was made of XMLHttpRequest because people were suffering

  • by khchung ( 462899 ) on Sunday December 17, 2017 @07:21AM (#55754691) Journal

    "But it's not hard to see how this creative explosion could all go very wrong."

    No. It is VERY HARD to see how this could go wrong.

    You mean, computer can generate an entire movie without having to hire real actors? So Hollywood movie stars can't make millions anymore? Cry me a river.

    This would be just as bad, which is to say not at all, as computers able to generate the sound of musical instruments which normal person cannot distinguish from real recordings. (Gee! Computer can generate a "fake" recording of an orchestra playing Bach symphonies! Aren't you afraid now?) It allowed music writers to compose and create music recordings (and put on YouTube) even though he cannot play any of the instruments in the score, including synthetic singers singing the song that goes with it. Is that bad?

    In the future, there could be many more solo "movie creators" who would, by his/her own effort, create an entire movie. Much like writers writing up a whole novel. It will take out all the middle man like the movie studio. This can only be good for humanity.

    • You mean, computer can generate an entire movie without having to hire real actors? So Hollywood movie stars can't make millions anymore? Cry me a river.

      In a Hollywood movie, you know it's all fake, so it won't matter if the actors are fake. It's all about entertainment.

      The danger comes from people manipulating things that are supposed to be true, like video recordings of political figures.

      • by gweihir ( 88907 )

        The danger comes from people manipulating things that are supposed to be true, like video recordings of political figures.

        The only danger there is that actually genuine recordings may not be recognized as such. People will just assume "fake" after the first few ones have proven to be.

        • It's going to make it much easier to avoid scandal though. Simply insist you didn't do it, and denounce any proof otherwise as fake. Seems to work already.

          • What? You're describing now. What this means is you say you didn't do it, and produce a video showing you not doing it. That's the real issue.

            Of course, that will also be denounced as fake as well, and we're back to just insisting that things are true.

            I have optimism that AI will be able to spot it if AI can make it. However, the real trick is going to be ensuring that the AI isn't manipulated to rule in your favor. AI auditing is going to be one hell of a challenge in the upcoming decades.

            • I have optimism that AI will be able to spot it if AI can make it

              These fakes are made by GANs (Generative adversarial network). It consists of two parts. The first part generates fakes, the second part tries to identify the fakes. These two networks are then trained against each other. In the end, you get good quality fakes that can fool the AI.

    • by gweihir ( 88907 )

      In the future, there could be many more solo "movie creators" who would, by his/her own effort, create an entire movie. Much like writers writing up a whole novel. It will take out all the middle man like the movie studio. This can only be good for humanity.

      Indeed. And music creators can do impressive scores all by themselves. Still requires the creativity and insight as before, nothing will change in that regard. Although as to movies, quite a few "AAA" productions seem to lack that creativity and insight as well. For example, after reading a few reviews, I am not even tempted to view the latest Star Wars movie. It seems to basically be a B-quality generic action flick with some Star Wars decor and references. I can do without that.

      • by Mal-2 ( 675116 )

        The procedural generators are pretty damn good. I can take the chords from an existing song, have Band In A Box generate a backing track along with two or three procedurally generated melodies, and two or three generated solos, and selectively stitch together my favorite pieces without changing a note. It will come off as cheesy because of a shortage of inflection, but you would most likely accept it as a "proof" of a human-composed piece.

        The problem with the procedural generators isn't that they can't get

    • I haven't seen anything *beautiful* generated by AI. Useful yes, sometimes even amusing, but not beautiful.

    • No. It is VERY HARD to see how this could go wrong.

      One barely needs to be awake to the present day to see how this can go wrong, because we're already living it.

      Next to Samoa and Saudi Arabia (and one or two others), America is pretty much the most obese country in human history. Did the bariatric hover-chairs in WALL-E seem altogether implausible? No, they didn't. (Well, not unless unless you think too much about the hovering part.)

      Evolution equipped us with a survival heuristic: pursue sugar.

      And then,

  • I am sick and tired of these bullshit announcements. Usually, you just find somebody that wants to get rich without actually contribution to society or some "journalist: that wants a cheap story. Basically nothing of these predictions comes true in the time-frame announced, if ever.

  • by Anonymous Coward

    We live in a world where Warner Bros. just spent millions trying to shop out Superman's mustache to spectacular failure. Disney's Grand Moff Tarkin was even worse. If these titans of the entertainment industry can't pull off a canny reproduction with their hand-crafted flagship products then I really think we'll have to worry about defeating Skynet before we'll have to worry about 'AI' defeating the uncanny valley.

  • None of the examples from the article are very convincing, even at these poststamp resolutions with massive compression artifacts. The whole point of the uncanny valley is that the last mile is very hard. I'm sure that it's all very impressive, but it's a bit early to say that they've crossed the uncanny valley. They better get close to the valley first.

  • ...mostly our craving for more and more entertainment.

    We're getting increasingly bored with our lives, we've never had it this good before in any century, we have all the luxuries of the modern world. From birth we're literally given a tablet computer where we can interact with whatever is crawling around on the screen, we're constantly connected to everything and everyone around us on an smartphone or similar devices - and it's addictive, and it's not necessarily in a good way.

    A.I. can indeed create wonder

    • But all this sitting in front of a smartphone screen, computer screen - even at work, has some serious physical health implication

      Compared to what? because if you go back even 50 years, our lifespans were a lot shorter. Go back another 50, and it was even worse.

      Sure, activity is better than inactivity, but honestly, humans are doing better health-wise with every passing decade. Looks like that's going to be close to leveling off at this point, but it also looks like that's because we're reaching the limits of what the human body is capable of in terms of longevity. We might be able to genetically engineer around that, but we also may

  • by vix86 ( 592763 ) on Sunday December 17, 2017 @09:02AM (#55754917)

    All of the claims about "fake news" could come to a head here really soon with more extreme left and right news sites/blogs putting out fake speeches and audio bites that have been created using this new technology. This tech is really going to muddy the waters on social media and will be utilized by movements and countries to spread disinformation. The more legitimate news outlets will spend more time fighting this disinformation instead of reporting on the actual events that are going on.

    I think ultimately what we'll see is that other companies will come along offering services that archive and perform various match tests against sound bites and recorded speeches. You'll be able to confirm if the video clip you just saw actually happened and if so, when and where it occurred. Without something like this, we all will be lost in a see of fake speeches and events. I expect the government will get involved in this and the Library of Congress will be tangentially involved in the collection, storage, and verification; but I don't think any of this will be taken seriously until politicians on both sides of the aisle get burned by fake creations.

    • I think ultimately what we'll see is that other companies will come along offering services that archive and perform various match tests against sound bites and recorded speeches

      I doubt it. Once you can fool people, it's a small step to fool a program. Besides, people don't like fact checking as it is. Authenticity is assumed if people agree with the content.

    • by mikael ( 484 )

      It's been happening for ages. Russians would get artists to "airbrush" people out of photographs. People fake UFO videos using basic computer animation software. Anyone can use GIMP and composite pictures together using "intelligent scissors", edge blurring and the airbrush.

      • It takes a lot of money to do it by hand. And it's very hard to do it accurately on moving images. AI systems will make it quick and cheap. This allows it to be done on a much wider scale.

    • They tried. Or at least the businesses that profit from ads on fake news sites did. What they found was that anything they tried got debunked in a matter of hours. It just didn't work. There's lots of theories why, you can Google the discussion. It's too controversial and this post is already likely to start a troll storm. But the fact remains fake news just doesn't work for the left.
    • All of the claims about "fake news" could come to a head here really soon with more extreme left and right news sites/blogs putting out fake speeches and audio bites that have been created using this new technology. This tech is really going to muddy the waters on social media and will be utilized by movements and countries to spread disinformation. .

      Which will just highlight the problem we already have. We already have quite capable media lies with no need for AI, thank you.

      Why should you just believe the crap that people beam to your house now? This phenomenon will just make that problem more obvious.

  • by 110010001000 ( 697113 ) on Sunday December 17, 2017 @09:29AM (#55754967) Homepage Journal
    It is 2017 and the Uncanny Valley is very much still alive. AI isn't real.
    • by mark-t ( 151149 )

      Is intelligence real?

      If so, why not AI? Which, by definition, is just intelligence that happens to be artificial.

      And of course, just as certainly as their can be varying levels of natural intelligence, there should quite obviously also be varying levels of artificial intelligence as well.

      But how do you define intelligence? As I said previously, is someone intelligent if they can memorize a sufficient amount of information and rules to be able to independently perform a complex task. What is the dif

    • by Kjella ( 173770 )

      Yeah, I hear some people repeat that like a robot in a GOTO 10 loop or a parrot, whichever is the worst insult. The Uncanny Valley is real, the question is who's sliding into it from which side...

  • It's not KILLING anything.

    Killing means terminating the life of a living organism.

    • That term "killer" has always baffled me since the game Doom and all of the "Doom-killers." I heard the term Doom-killer before the term "first person shooters." Suffice it to say none of them "killed" Doom. Nor did all "iPod-killers" kill the iPod. I recognize "killer app" as a more useful term.
      • Maybe it's related to the fascination Americans have with guns and the military in general. Just look at their movies. There's a disproportionate percentage of war/violence movies compared to other countries.

  • Imagine the bullshit election propoganda of fake âhidden camerasâ(TM) published by the zealots on both sides so obsessed with winning at all costs thet they dont care about eviscerating and disenfanchising the voters. Eventually with all the accusations, denials, lawsuits for defamation; in the end we will be no better off than one run by big brother. And that will be double plus ungood.

  • Governments, corporations and others who want to manipulate the public certainly have good reasons to develop AI's that pass the Uncanny Valley test.

    For a lot of average people, though, I suspect a supremely uncomplicated use: make me a robot I enjoy fucking, maybe conversing with once in a while if I'm lonely, and that doesn't kick up a fuss when it gets turned off and put away in the cupboard.

    Human relationships are complex and difficult. They usually require two very different individuals to compromise

    • You can buy the pasta ready made. Just like Spaghetti.
      But the true trick is: use a really big pan, make enough for 4 or more meals. Just cook them like 75%, and freeze everything you don't need.
      Now you have your home made super Lasagne, ready in your freezer.
      Making your own pasta, especially for Lasagne is super easy as well.

  • ..and we're the sheep. The singularity will arrive not with a bang but with the soft sound of us ceding little bits of volition and self-determination to AI because they "know best" - and likely truly will. What scares me most is the blind human faith in algorithmic certainty, where "certainty" is far from certain.

Perfection is acheived only on the point of collapse. - C. N. Parkinson

Working...