How synthetic media, or deepfakes, could soon change our world
515.1k views2042 WordsCopy TextShare
60 Minutes
Synthetic media, better known as deepfakes, could be a goldmine for filmmakers. But the technology h...
Video Transcript:
you may never have heard the term synthetic media more commonly known as deep fakes but our military law enforcement and intelligence agencies certainly have they are hyper-realistic video and audio recordings that use artificial intelligence and deep learning to create fake content or deep fakes the us government has grown increasingly concerned about their potential to be used to spread disinformation and commit crimes that's because the creators of deep fakes have the power to make people say or do anything at least on our screens most americans have no idea how far the technology has come in just the last four years or the danger disruption and opportunities that come with it the story will continue in a moment you know i do all my own stunts obviously i also do my own music this is not tom cruise it's one of a series of hyper-realistic deep fakes of the movie star that began appearing on the video sharing app tick tock earlier this year hey what's up tick-tock for days people wondered if they were real and if not who had created them it's important finally a modest 32 year old belgian visual effects artist named chris umi stepped forward to claim credit we believe as long as we're making clear this is a parody we're not doing anything to harm his image but after a few videos we realized like this is blowing up we're getting millions and millions and millions of views umi says his work is made easier because he teamed up with a tom cruise impersonator whose voice gestures and hair are nearly identical to the real mccoy umi only deep fakes cruz's face and stitches that onto the real video and sound of the impersonator where the magic happens for technophiles jeep tom cruise was a tipping point for deep fakes still got it how do you make this so seamless but it begins with training a defect model of course where i have all the face angles of tom cruise all the expressions all the emotions it takes time to create a really good defect model what do you mean training the model how do you train your computer training means it's going to analyze all the images of tom cruise all his expressions compared to my impersonator so the computer is going to teach itself when my impersonator is smiling i'm going to recreate tom cruise smiling and that's that's how you train it using video from the cbs news archives chris umi was able to train his computer to learn every aspect of my face and wipe away the decades this is how i looked 30 years ago he can even remove my mustache the possibilities are endless and a little frightening i see a lot of mistakes in my work but i don't mind it actually because i don't want to fool people i just want to show them what's possible you don't want to fool people no i want to entertain people i want to raise awareness and i want and i want to show where it's all going it is without a doubt one of the most important revolutions in the future of human communication and perception i would say it's analogous to the birth of the internet political scientist and technology consultant nina shikh wrote one of the first books on deep fakes she first came across them four years ago when she was advising european politicians on russia's use of disinformation and social media to interfere in democratic elections what was your reaction when you first realized this was possible and was going on well given that i was coming at it from the perspective of disinformation and manipulation in the context of elections the fact that ai can now be used to make images and video that are fake that look hyper realistic so i thought well from a disinformation perspective this is a game changer so far there's no evidence deep fakes have changed the game in a u. s election but earlier this year the fbi put out a notification warning that russian and chinese actors are using synthetic profile images creating deep fake journalists and media personalities to spread anti-american propaganda on social media so how do you get deep fakes the u. s military law enforcement and intelligence agencies have kept a wary eye on deep fakes for years at this 2019 hearing senator ben sasse of nebraska asked if the u.
s is prepared for the onslaught of disinformation fakery and fraud when you think about the catastrophic potential to public trust and to markets that could come from deep fake attacks are we organized in a way that we could possibly respond fast enough we clearly need to be more agile it poses a major threat to the united states and something that the intelligence community needs to be restructured to address since then technology has continued moving at an exponential pace while u. s policy has not efforts by the government and big tech to detect synthetic media are competing with a community of deep fake artists who share their latest creations and techniques online like the internet the first place deep fake technology took off was in pornography the sad fact is the majority of deep fakes today consist of women's faces mostly celebrities superimposed onto pornographic videos the first use case in pornography is just a harbinger of how deep fakes can be used maliciously in many different contexts which are now starting to arise and they're getting better all the time yes the incredible thing about deep fakes and synthetic media is the pace of acceleration when it comes to the technology and by five to seven years we are basically looking at a trajectory where any single creator so a youtuber a tick talker will be able to create the same level of visual effects that is only accessible to the most well-resourced hollywood studio today the technology behind deep fakes is artificial intelligence which mimics the way humans learn in 2014 researchers for the first time used computers to create realistic looking faces using something called generative adversarial networks or gans so you set up an adversarial game where you have two ais combating each other to try and create the best fake synthetic content and as these two networks combat each other one trying to generate the best image the other trying to detect where it could be better you basically end up with an output that is increasingly improving all the time six says the power of generative adversarial networks is on full display at a website called this person does not exist dot com every time you refresh the page there's a new image of a person who does not exist each is a one of a kind entirely ai generated image of a human being who never has and never will walk this earth you can see every pore on their face you can see every hair on their head but now imagine that technology being expanded out not only to human faces in still images but also to video to audio synthesis of people's voices and that's really where we're heading right now this is mind-blowing yes what what's the positive side of this the technology itself is neutral so just as bad actors are without a doubt going to be using deep fakes it is also going to be used by good actors so first of all i would say that there is a very compelling case to be made for the commercial use of deep fakes victor riparbelli is ceo and co-founder of synthesia based in london one of dozens of companies using deep fake technology to transform video and audio productions the way synthesia works is that we've essentially replaced cameras with code and once you're working with software we do a lot of things that you wouldn't be able to do with a normal camera we're still very early but this is going to be a fundamental change in how we create media this video was of course generated by syndica synthesia makes and sells digital avatars using the faces of paid actors to deliver personalized messages in 64 languages and allows corporate ceos to address employees overseas did somebody synthesia has also helped entertainers like snoop dogg go forth and multiply this elaborate tv commercial for european food delivery service just eat cost a fortune just it has a subsidiary in australia which is called menu log so what we did with our technology was we switched out the word just eat for menu log and all of a sudden they had a localized version for the australian market without snoop dog having to do anything so he makes twice the money huh yeah all it took was eight minutes of me reading a script on camera for synthesia to create my synthetic talking head complete with my gestures head and mouth movements another company dscript used ai to create a synthetic version of my voice this is bill woodacher's synthetic voice with my cadence tenor and syncopation this is the result the words you're hearing were never spoken by the real bill into a microphone or to a camera he merely typed the words into a computer and they come out of my mouth it may look and sound a little rough around the edges right now but as the technology improves the possibilities of spinning words and images out of thin air are endless i'm bill woodacher i'm bill whittaker i'm bill woodaker wow and the head the eyebrows the mouth the way it moves it's all synthetic i could be lounging at the beach and say folks uh you know i'm not going to come in today but you can use my avatar to do the work maybe in a few years don't don't tell me that i'd be tempted i think it'll have a big impact the rapid advances in synthetic media have caused a virtual gold rush tom graham a london-based lawyer who made his fortune in cryptocurrency recently started a company called metaphysic with none other than chris umi creator of deep tom cruise their goal develop software to allow anyone to create hollywood caliber movies without lights cameras or even actors as the hardware scales and as the models become more efficient we could scale up the size of that model to be an entire tom cruise body movement and everything well talk about disruptive i mean are you going to put actors out of jobs i think that it's a great thing if you're a well-known actor today um because you may be able to let somebody collect data for you to create a version of yourself in the future where you could be acting in movies after you have deceased or you could be the director directing your younger self in a movie or something like that if you are wondering how all of this is legal most deep fakes are considered protected free speech attempts at legislation are all over the map in new york commercial use of a performer's synthetic likeness without consent is banned for 40 years after their death california and texas prohibit deceptive political deep fakes in the lead-up to an election there are so many ethical philosophical gray zones here that we really need to think about so how do we as a society grapple with this just understanding what's going on because a lot of people still don't know what a deep fake is what synthetic media is uh that this is now possible the counter to that is how do we inoculate ourselves and understand that this kind of content is coming and exists without being completely cynical right how do we do it without losing trust in all authentic media that's going to require all of us to figure out how to maneuver in a world where seeing is not always believing when any video can be faked how do you prove when one is real if everything can be faked then anything can be denied at 60 minutesovertime.