Apr 15, 2020 06:06 PM
(This post was last modified: Apr 15, 2020 06:07 PM by C C.)
https://www.knowablemagazine.org/article...-deepfakes
EXCERPT: The snapshots above [visit KM article] look like people you’d know. Your daughter’s best friend from college, maybe? That guy from human resources at work? The emergency-room doctor who took care of your sprained ankle? One of the kids from down the street?
Nope. All of these images are “deepfakes” — the nickname for computer-generated, photorealistic media created via cutting-edge artificial intelligence technology. They are just one example of what this fast-evolving method can do. (You could create synthetic images yourself at ThisPersonDoesNotExist.com) Hobbyists, for example, have used the same AI techniques to populate YouTube with a host of startlingly lifelike video spoofs — the kind that show real people such as Barack Obama or Vladimir Putin doing or saying goofy things they never did or said, or that revise famous movie scenes to give actors like Amy Adams or Sharon Stone the face of Nicolas Cage. All the hobbyists need is a PC with a high-end graphics chip, and maybe 48 hours of processing time.
It’s good fun, not to mention jaw-droppingly impressive. And coming down the line are [...] applications that could make quick work out of once-painstaking tasks: filling in gaps and scratches in damaged images or video; turning satellite photos into maps; creating realistic streetscape videos to train autonomous vehicles; giving a natural-sounding voice to those who have lost their own; turning Hollywood actors into their older or younger selves; and much more.
Yet this technology has an obvious — and potentially enormous — dark side. Witness the many denunciations of deepfakes as a menace “Deepfakes play to our weaknesses [...] When we see a doctored video that looks utterly real ... “it’s really hard for our brains to disentangle whether that’s true or false.” And the internet being what it is, there are any number of online scammers, partisan zealots, state-sponsored hackers and other bad actors eager to take advantage of that fact.
The technology gets its nickname from Deepfakes, an anonymous Reddit user who launched the movement in November 2017 by posting AI-generated videos in which the faces of celebrities such as Scarlett Johansson and Gal Gadot are mapped onto the bodies of porn stars in action. This kind of non-consensual celebrity pornography still accounts for about 95 percent of all the deepfakes out there, with most of the rest being jokes of the Nicolas Cage variety.
But while the current targets are at least somewhat protected by fame [...] non-celebrities will increasingly be targeted, as well. Old-fashioned revenge porn is a ubiquitous feature of domestic violence cases [...] “unsophisticated perpetrators no longer require nudes or a sex tape to threaten a victim. They can simply manufacture them.”
Then there’s the potential for political abuse. Want to discredit an enemy? [...] Meanwhile, don’t forget old-fashioned greed. ... “I could create a fake video of Jeff Bezos saying ... ‘Amazon’s profits are down 10 percent.’ ... “You could have global stock manipulation to the tune of billions of dollars.”
And beyond all that, Farid says, looms [...] no longer believing and miscreants can bask in a whole new kind of plausible deniability. Body-cam footage? CCTV tapes? Photographic evidence of human-rights atrocities? Audio of a presidential candidate boasting he can grab women anywhere he wants? “DEEPFAKE!” Deepfake video methods can digitally alter a person’s lip movements to match words that they never said. ... Thus the widespread concern about deepfake technology...
[...] Broadly speaking, the efforts to date follow two strategies: authentication and detection. ... In the meantime [...of those being better developed...], though, there is a widespread consensus in the field that technology will never solve the deepfake challenge by itself. Most of the issues that have surfaced in the WITNESS roadmap exercises are non-technological. What’s the best way of raising public awareness, for example? How do you regulate the technology without killing legitimate applications? And how much responsibility should be borne by social media?
[...] Facebook partially answered that last question on January 6, when it announced it would ban any deepfake that was intended to deceive the viewers. It remains to be seen how effective that ban will be. ... On the technological front, should the platforms start demanding a valid authentication tag for each file that’s posted? ... Alternatively, the platforms could go the crowdsourcing route by making the detection tools available for users to flag deepfakes on their own — complete with leaderboards and bounties for each one detected.
There’s a gnarlier question, though [...] What should sites do when they discover bogus posts? Some might be tempted to go the libertarian route that Facebook has taken with deceptive political ads, or non-AI video manipulation: Simply label the known fakes and let the viewers beware.
Unfortunately, that overlooks the damage such a post can do — given that a lot of people will believe it’s real despite the “fake” label. Should, then, sites take down deepfakes as soon as they are discovered? That’s pretty clear-cut when it comes to non-consensual porn [...] But ...deepfakes that are clearly intended as satire, parody or political commentary[?] ... In those cases ,,, “the legislative and regulatory side is complex because of the free-speech issue.”
A lot of people are afraid to touch this question [...] Particularly in the Silicon Valley culture, she says, “there’s often a sense that either we have no regulation, or we have a Ministry of Truth.” ... “I don't necessarily have a good answer to what we should do about the new media environment ... But I do think that the gray areas are where we need to have this conversation.” (MORE - details)
EXCERPT: The snapshots above [visit KM article] look like people you’d know. Your daughter’s best friend from college, maybe? That guy from human resources at work? The emergency-room doctor who took care of your sprained ankle? One of the kids from down the street?
Nope. All of these images are “deepfakes” — the nickname for computer-generated, photorealistic media created via cutting-edge artificial intelligence technology. They are just one example of what this fast-evolving method can do. (You could create synthetic images yourself at ThisPersonDoesNotExist.com) Hobbyists, for example, have used the same AI techniques to populate YouTube with a host of startlingly lifelike video spoofs — the kind that show real people such as Barack Obama or Vladimir Putin doing or saying goofy things they never did or said, or that revise famous movie scenes to give actors like Amy Adams or Sharon Stone the face of Nicolas Cage. All the hobbyists need is a PC with a high-end graphics chip, and maybe 48 hours of processing time.
It’s good fun, not to mention jaw-droppingly impressive. And coming down the line are [...] applications that could make quick work out of once-painstaking tasks: filling in gaps and scratches in damaged images or video; turning satellite photos into maps; creating realistic streetscape videos to train autonomous vehicles; giving a natural-sounding voice to those who have lost their own; turning Hollywood actors into their older or younger selves; and much more.
Yet this technology has an obvious — and potentially enormous — dark side. Witness the many denunciations of deepfakes as a menace “Deepfakes play to our weaknesses [...] When we see a doctored video that looks utterly real ... “it’s really hard for our brains to disentangle whether that’s true or false.” And the internet being what it is, there are any number of online scammers, partisan zealots, state-sponsored hackers and other bad actors eager to take advantage of that fact.
The technology gets its nickname from Deepfakes, an anonymous Reddit user who launched the movement in November 2017 by posting AI-generated videos in which the faces of celebrities such as Scarlett Johansson and Gal Gadot are mapped onto the bodies of porn stars in action. This kind of non-consensual celebrity pornography still accounts for about 95 percent of all the deepfakes out there, with most of the rest being jokes of the Nicolas Cage variety.
But while the current targets are at least somewhat protected by fame [...] non-celebrities will increasingly be targeted, as well. Old-fashioned revenge porn is a ubiquitous feature of domestic violence cases [...] “unsophisticated perpetrators no longer require nudes or a sex tape to threaten a victim. They can simply manufacture them.”
Then there’s the potential for political abuse. Want to discredit an enemy? [...] Meanwhile, don’t forget old-fashioned greed. ... “I could create a fake video of Jeff Bezos saying ... ‘Amazon’s profits are down 10 percent.’ ... “You could have global stock manipulation to the tune of billions of dollars.”
And beyond all that, Farid says, looms [...] no longer believing and miscreants can bask in a whole new kind of plausible deniability. Body-cam footage? CCTV tapes? Photographic evidence of human-rights atrocities? Audio of a presidential candidate boasting he can grab women anywhere he wants? “DEEPFAKE!” Deepfake video methods can digitally alter a person’s lip movements to match words that they never said. ... Thus the widespread concern about deepfake technology...
[...] Broadly speaking, the efforts to date follow two strategies: authentication and detection. ... In the meantime [...of those being better developed...], though, there is a widespread consensus in the field that technology will never solve the deepfake challenge by itself. Most of the issues that have surfaced in the WITNESS roadmap exercises are non-technological. What’s the best way of raising public awareness, for example? How do you regulate the technology without killing legitimate applications? And how much responsibility should be borne by social media?
[...] Facebook partially answered that last question on January 6, when it announced it would ban any deepfake that was intended to deceive the viewers. It remains to be seen how effective that ban will be. ... On the technological front, should the platforms start demanding a valid authentication tag for each file that’s posted? ... Alternatively, the platforms could go the crowdsourcing route by making the detection tools available for users to flag deepfakes on their own — complete with leaderboards and bounties for each one detected.
There’s a gnarlier question, though [...] What should sites do when they discover bogus posts? Some might be tempted to go the libertarian route that Facebook has taken with deceptive political ads, or non-AI video manipulation: Simply label the known fakes and let the viewers beware.
Unfortunately, that overlooks the damage such a post can do — given that a lot of people will believe it’s real despite the “fake” label. Should, then, sites take down deepfakes as soon as they are discovered? That’s pretty clear-cut when it comes to non-consensual porn [...] But ...deepfakes that are clearly intended as satire, parody or political commentary[?] ... In those cases ,,, “the legislative and regulatory side is complex because of the free-speech issue.”
A lot of people are afraid to touch this question [...] Particularly in the Silicon Valley culture, she says, “there’s often a sense that either we have no regulation, or we have a Ministry of Truth.” ... “I don't necessarily have a good answer to what we should do about the new media environment ... But I do think that the gray areas are where we need to have this conversation.” (MORE - details)