SoylentNews is people
SoylentNews
How an A.I. 'Cat-and-Mouse Game' Generates Believable Fake Photos (archive)
The woman in the photo seems familiar. She looks like Jennifer Aniston, the "Friends" actress, or Selena Gomez, the child star turned pop singer. But not exactly. She appears to be a celebrity, one of the beautiful people photographed outside a movie premiere or an awards show. And yet, you cannot quite place her. That's because she's not real. She was created by a machine.
The image is one of the faux celebrity photos generated by software under development at Nvidia, the big-name computer chip maker that is investing heavily in research involving artificial intelligence.
At a lab in Finland, a small team of Nvidia researchers recently built a system that can analyze thousands of (real) celebrity snapshots, recognize common patterns, and create new images that look much the same — but are still a little different. The system can also generate realistic images of horses, buses, bicycles, plants and many other common objects.
The project is part of a vast and varied effort to build technology that can automatically generate convincing images — or alter existing images in equally convincing ways. The hope is that this technology can significantly accelerate and improve the creation of computer interfaces, games, movies and other media, eventually allowing software to create realistic imagery in moments rather than the hours — if not days — it can now take human developers.
(Score: 2) by Arik on Wednesday January 03 2018, @09:36PM (5 children)
Too bad it's a story that revolves around images, and they don't seem to have discovered that tag yet.
If laughter is the best medicine, who are the best doctors?
(Score: 2) by MrGuy on Wednesday January 03 2018, @09:50PM (4 children)
Either I'm missing the thrust of your comment, or we're looking at two different articles.
Both the NY Times article linked and the PDF that's linked as a secondary contain images.
(Score: 2) by Arik on Wednesday January 03 2018, @09:58PM (3 children)
The NYTimes article contains no images, just placeholders where it expects them to eventually be overlaid by some kind of ecmascript payload.
If laughter is the best medicine, who are the best doctors?
(Score: 2) by takyon on Wednesday January 03 2018, @10:19PM (2 children)
I gave you an archive link which displays the images (slowly), and you have the PDF.
For the future: Whenever I include an "(archive)" link next to something, I intend for that to be used as a workaround for a paywall or JavaScript issues.
[SIG] 10/28/2017: Soylent Upgrade v14 [soylentnews.org]
(Score: 2) by Arik on Thursday January 04 2018, @05:38AM (1 child)
If laughter is the best medicine, who are the best doctors?
(Score: 2) by takyon on Thursday January 04 2018, @05:47AM
I just want everyone to be aware of archive.(fo|is|li|eu). Incredibly useful.
[SIG] 10/28/2017: Soylent Upgrade v14 [soylentnews.org]
(Score: 1, Informative) by Anonymous Coward on Wednesday January 03 2018, @09:41PM (5 children)
How much longer before AIs replace Hollywood stars? Movie making will get cheaper...
(Score: 3, Informative) by takyon on Wednesday January 03 2018, @10:02PM
https://en.wikipedia.org/wiki/Virtual_actor [wikipedia.org]
It has been used pretty extensively in the new Star Wars trilogy.
Here are two fictional films about the concept (am I missing any?):
https://en.wikipedia.org/wiki/Simone_(2002_film) [wikipedia.org]
https://en.wikipedia.org/wiki/The_Congress_(2013_film) [wikipedia.org]
Hollywood estates will oppose the unpaid use of virtual versions of dead actors. Composite versions could sidestep that (blend two or more famous actors) and lead to a nice First Amendment case. Of course, you could always create something from scratch, like Hatsune Miku [soylentnews.org].
Perhaps the bigger deal will be CGI sets. Sanctuary [wikipedia.org] is a TV show that extensively used green screens. With better/cheaper hardware and more automation on the animation/design side, even total AutoCAD noobs could create a realistic looking mansion in which they can insert their real (or virtual) actors.
[SIG] 10/28/2017: Soylent Upgrade v14 [soylentnews.org]
(Score: 3, Insightful) by bob_super on Wednesday January 03 2018, @10:02PM (3 children)
Well, the face is the last thing left, since most scenarios and dialogs are written by AlzheimerDory dysfunctional AIs already. Have you seen that latest quarter-billion-budget Disney thing ?
(Score: 3, Funny) by DannyB on Wednesday January 03 2018, @10:32PM (2 children)
The dialog may suck. The scenarios may be implausible. The plot may be incomprehensible or too simplistic. The characters may be cardboard. The special effects may be way over done. And now, even the actors may be genuinely fake.
But . . . if that movie cost at least five hundred million dollars to make, then it MUST be good!
And if 500 channels ain't not gooder enough, then with 1,000 channels you must be able to find something worth watching?
When trying to solve a problem don't ask who suffers from the problem, ask who profits from the problem.
(Score: 2) by takyon on Wednesday January 03 2018, @10:52PM (1 child)
What benefits Hollywood today is good for the small filmmaker of ten years from now.
The number of people needed to achieve a particular level of quality and scope of animations (for example, any huge battle scene from Lord of the Rings) will decline over time. Perhaps eventually reaching a point where more money and personnel does not improve the CGI quality that much.
Machine learning and other techniques can help by automating some of the technical expertise and creativity needed. Just imagine creating a procedurally generated city with one click. Adjust the parameters, stage your movie in it, control camera movement with coordinates, etc.
Hollywood gets things done at a large cost by hiring one person to animate buildings, another to do clouds, fire hydrants, etc. They can continue blowing $100+ million on CGI work even after each person becomes massively more productive. But at some point, another dollar spent will not make the new Avengers movie look any more photorealistic. And the hardware and personnel needed to achieve that will become less and less. $10 million Iron Sky [wikipedia.org] budget... $100,000 to $1 million indie filmmaker budget... $20,000 Xeon rig... all the way down to the $1,000 gaming PC and 1 person working in their free time. CPU performance is not increasing too much, but that could change with post-CMOS techniques (stacking to prevent the end of Moore's law scaling), and GPU + machine learning performance are increasing a lot. We could get to 1-100 "petaflops" with a single consumer-grade GPU (and most of the relevant software techniques should be highly parallel).
The small filmmaker is not guaranteed to produce a better film than Hollywood can, but they have more freedom of expression and can appeal to a niche audience. If the technological playing field levels out, and any number of virtual actors and sets can be created on demand, the small filmmaker could have less limits on what they can achieve.
[SIG] 10/28/2017: Soylent Upgrade v14 [soylentnews.org]
(Score: 2) by DannyB on Thursday January 04 2018, @03:34PM
The small filmmaker is not constrained by something that plagues Hollywood. Taking risks. If you're going to spend $500 Million or $1 Billion on a movie, then it must:
* Have a known director
* Have only A-List Actors
* Follow a known formula
* Be a sequel or prequel to a remake of a sequel to a 4 or 5 decade old television show
Small filmmakers can try completely new fresh ideas. Unknown actors.
When trying to solve a problem don't ask who suffers from the problem, ask who profits from the problem.
(Score: 4, Insightful) by MrGuy on Wednesday January 03 2018, @09:46PM (17 children)
From TFA:
If you think fake news is a problem now, wait until you have undetectable fake images of Hillary Clinton accepting a suspicious-looking envelope from an Iranian diplomat. Or Donald Trump Jr. meeting with the Russian ambassador and several known KGB figures.
(Score: 3, Insightful) by Anonymous Coward on Wednesday January 03 2018, @09:54PM (8 children)
In a single end-product no less.
I am pretty sure it was explained that machine generated derivative works fall under the original sources copyrights, which in this place is the thousands of copyrighted photographs that Nvidia used to prime its adversarial neural networks.
Which should mean they could be open to infringement lawsuits by any of the photographers who took those pictures, or any of the actors or actresses who likenesses they infringed.
I just wish this would happen since it seems like only the normal people get swatted with legal infringement, even as the big companies find new ways to steal our individual IP and monetize it for their collectives.
(Score: 3, Interesting) by DannyB on Wednesday January 03 2018, @10:11PM (1 child)
Nvidia might not have legitimate access to all of your photos.
But FaceTwit, Google, YouTube, Microsoft, Apple, Instagram, Flikr, Imgur and others probably do. And you probably agreed to it. About 2/3 of the way through the EULA, on page 349, right after the clause that allows them to sneak in the middle of the night and harvest your and your family's vital organs -- unless your ISP has already gotten them first.
So surely Google has an absolutely immense collection of photos to use for this. And nobody would know or could prove that their photo was used as part of a large neural net training data set.
I suspect that Google, et al, would argue that this is not a copyright infringing use. Your photo is not used in public. Never published. Google (or others with large photo data sets) might suddenly be able to generate fake faces, but none of them would look like you. They would merely meet the rules of looking realistic -- as trained by the large data set, of which your face is a part.
If a human learns about faces by looking at a large collection of faces, and that isn't copyright infringement, then why is it suddenly copyright infringement if an AI learns about what makes a realistic face by looking at a large collection of faces?
When trying to solve a problem don't ask who suffers from the problem, ask who profits from the problem.
(Score: 2) by HiThere on Wednesday January 03 2018, @11:57PM
Actually, people have a built-in face recognizer. The "large collection of faces" is used to limit what is seen as a valid face. And you still get people seeing faces in sandwiches, etc. It's sort of like "Start off with everything that has two spots above a third spot being considered a face, and find good limitations". I'm not sure this is the normal form of learning. OTOH, we seem to learn phonemes the same way, so maybe it is.
Javascript is what you use to allow unknown third parties to run software you have no idea about on your computer.
(Score: 2) by takyon on Wednesday January 03 2018, @10:17PM (4 children)
It should be possible to obtain public domain photographs of people like Hillary Clinton or Will Smith, and then use those to create 3D models or whatever is needed to fake them snorting coke.
Or, you use copyrighted material to obtain their likeness, but manipulate it enough so that the derivative work can't be linked to any particular copyrighted photograph (again, anybody can take a photo of Hillary Clinton in public and declare it a public domain work).
Are you sure that Nvidia hasn't asked for permission, licensed the photographs from some archive, or does not have a valid fair use exemption?
[SIG] 10/28/2017: Soylent Upgrade v14 [soylentnews.org]
(Score: 2) by Rivenaleem on Monday January 08 2018, @01:40PM (3 children)
Or, you find these people at any public event and take a photo of them yourselves!!!
(Score: 2) by takyon on Tuesday January 09 2018, @01:42AM (2 children)
I badly implied that by writing "obtain public domain photographs". If you take it yourself, you can do whatever you want with it including releasing it into the public domain.
I still think that if you collect enough 2D data to create your 3D virtual actor, it will be impossible to trace the origin of any copyright material used. I think the more relevant legal issue is "personality rights".
[SIG] 10/28/2017: Soylent Upgrade v14 [soylentnews.org]
(Score: 2) by Rivenaleem on Tuesday January 09 2018, @11:26AM (1 child)
Well, you don't need this tech to create an infringement on "personality rights" because, in the case of computer games, it's totally possible for a talented CG Artist to, from scratch, create a 3D image of a celebrity. It won't be perfect, but it might fool enough people into thinking that the person posed for motion capture, and thus endorsed the likeness. Then you just need someone capable of mimicking their voice to a sufficiently believable level.
(Score: 2) by takyon on Tuesday January 09 2018, @12:19PM
https://www.theverge.com/2016/9/2/12777204/lindsay-lohan-grand-theft-auto-5-lawsuit-thrown-out [theverge.com]
[SIG] 10/28/2017: Soylent Upgrade v14 [soylentnews.org]
(Score: 3, Insightful) by ilsa on Wednesday January 03 2018, @11:22PM
Do you seriously think anyone would give two shits about copyright infringement? I think you missed the critical point of the parents argument.
This will take the concept of "Fake News" to unheard of levels. Currently it takes a great deal of effort to photoshop existing imagery and manipulate it to what you want it to be, and unless the person is exceptional, there will likely be tells that prove the image was doctored.
But if the image is so perfect that it is impossible to tell which one was original? You'll be able to invent all kinds of evidence to incriminate whoever you want, with nothing more than a nod and a wave of the hand.
This is seriously dangerous stuff that, combined with the manipulative Fake News/Social Media issues we're already struggling with, can be devastating to society.
(Score: 2) by MostCynical on Wednesday January 03 2018, @09:55PM (2 children)
Except they say they haven't perfected it yet, so maybe photos of one of those events events might be real...
Meh, it's all fake: fake news, fake smiles, fake hair..
"I guess once you start doubting, there's no end to it." -Batou, Ghost in the Shell: Stand Alone Complex
(Score: 1, Insightful) by Anonymous Coward on Wednesday January 03 2018, @10:45PM (1 child)
Fake comments to generate fake outrage.
(Score: 2) by DeathMonkey on Thursday January 04 2018, @07:37PM
Fake comments to repeal Net Neutrality. [fortune.com]
(Score: 2) by takyon on Wednesday January 03 2018, @10:07PM (3 children)
"Undetectable" perfection isn't needed. It only needs to be good enough to fool Facebook users.
[SIG] 10/28/2017: Soylent Upgrade v14 [soylentnews.org]
(Score: 2) by DannyB on Wednesday January 03 2018, @10:34PM (1 child)
What about the Twit users?
When trying to solve a problem don't ask who suffers from the problem, ask who profits from the problem.
(Score: 3, Interesting) by takyon on Wednesday January 03 2018, @10:56PM
https://help.twitter.com/en/using-twitter/twitter-videos [twitter.com]
https://sproutsocial.com/insights/twitter-video/ [sproutsocial.com]
The short attention span could be a big help.
[SIG] 10/28/2017: Soylent Upgrade v14 [soylentnews.org]
(Score: 2) by janrinok on Thursday January 04 2018, @03:12PM
(Score: 2) by DeathMonkey on Thursday January 04 2018, @07:40PM
If you think fake news is a problem now, wait until you have undetectable fake images of Hillary Clinton accepting a suspicious-looking envelope from an Iranian diplomat. Or Donald Trump Jr. meeting with the Russian ambassador and several known KGB figures.
Good thing the reality-based media corroborates their stories. So the Trump scenario is likely safe.
Can't say the same about Fox, though, so we'll definitely be seeing the Clinton one.
(Score: 1, Interesting) by Anonymous Coward on Wednesday January 03 2018, @10:09PM (2 children)
Indeed, what would Max Headroom do with a girl like that?
(Score: 0) by Anonymous Coward on Wednesday January 03 2018, @10:23PM
Take her for a ride. [youtube.com]
(Score: 0) by Anonymous Coward on Wednesday January 03 2018, @10:34PM
https://www.youtube.com/watch?v=jjeUuakHsLw [youtube.com]
(Score: 3, Interesting) by linkdude64 on Thursday January 04 2018, @12:30AM
Someone is going to make a shitload of money if we're at war with NK? They'll pay for a video of Kim Jong saying he's going to nuke Hawaii tomorrow. If the possibilities for deception are not discussed at large, this will become a disconcerting development for humankind.