The Jordan Harbinger Show - 486: Nina Schick | Deepfakes and the Coming Infocalypse

Episode Date: March 25, 2021

Nina Schick (@NinaDSchick) specializes in how technology is transforming geopolitics and society in the 21st century. She is the author of Deepfakes: The Coming Infocalypse. What We Discuss with Nina Schick: What deepfakes are and how they're created using technology that is getting exponentially more sophisticated every year and accessible to anyone. How deepfakes go beyond novelty and into dangerous territory occupied by blackmailers, propagandists, terrorists, conspiracy theorists, and other unsavory ne'er-do-wells who profit by making the rest of us doubt what we can trust as reality. Why deepfakes are so much more concerning than reality-bending manipulations of the past. The potential positive uses of AI-assisted deepfake technology that will change our world as dramatically as the advent of the Internet. How we can discern between real photo, video, or audio footage and imposter deepfakes to avoid being duped by 21st-century technoshysters. And much more... Full show notes and resources can be found here: jordanharbinger.com/486 Sign up for Six-Minute Networking -- our free networking and relationship development mini course -- at jordanharbinger.com/course! Like this show? Please leave us a review here -- even one sentence helps! Consider including your Twitter handle so we can thank you personally!See Privacy Policy at https://art19.com/privacy and California Privacy Notice at https://art19.com/privacy#do-not-sell-my-info.

Coming up on the Jordan Harbinger show. Fake pornography has a long history. So this is not just a Photoshopped image of a celebrity's face, stuck onto a porn star's body. This is a real live video where the celebrity is moving her face. She's got different expressions. AI is getting better and better. So less training data is needed.

I can make a nude image of your sister, your wife, your mom, and young girls from a single photo, for example. And there were over 100,000 of these images, is just being publicly passed around in these telegram groups. The key point is the AI can basically be used to hijack anyone's biometrics. So it's pretty clear that it's going to leach out into other forms of malicious use with fraud, with political disinformation, as a tool of intimidations. We're already starting to see that.

Welcome to the show. I'm Jordan Harbinger. On the Jordan Harbinger show, we decode the stories, secrets, and skills are the world's most fascinating people. We have in-depth conversations with people. at the top of their game, astronauts and entrepreneurs, spies and psychologists, even the occasional billionaire investor, rocket scientist, or extreme athlete, and each episode turns our guest's wisdom into practical advice that you can use to build a deeper understanding of how the world works and become a better critical thinker. If you're new to the show or you're looking for a handy way

to tell your friends about it, we now have episode starter packs. These are collections of your favorite episodes organized by popular topics to help new listeners get a taste of every everything we do here on this show. Just visit jordanharbinger.com slash start to get started or to help somebody else get started. And of course, we always appreciate that. Today's episode is about deep fakes, which is an incredible phenomenon that if you haven't heard about it yet, this is about to rock your world. And if you're listening to this in five, ten years, you're going to be like, what? This was news? How did you not know about this? I'm serious. This is going to take the world by storm, this is not hyperbole. Deepfakes are essentially software-slash-computer-created videos that are, of course,

synthetically generated, but they look like they were filmed. So it's synthetic media. You may have just recently seen some video of Tom Cruise doing a magic trick, and then you find out it's not Tom Cruise at all, but some guy on TikTok with some software. That's a deep fake. Seems harmless, right? Kind of a novelty thing? Well, it all depends on the application. What if deep fakes are used to create disinformation? What if they're used to create or discredit evidence in a trial? What if they're used to create revenge porn about somebody that you know? What if it's about a minor? Well, it turns out this is already happening. Today on the show, my friend Nina Schick and I discuss where this deepfakes phenomenon is going, how it's being weaponized against us, and what we can do about it.

You like that dramatic pause right there? Very intense, right? I think this episode was fascinating, and it's a great intersection of technologies, social engineering, information warfare, and the internet. And Nina really is the expert in this area, so I think you're going to love this conversation as much as I did. And if you are wondering how I book great, amazing folks like Nina all the time, these authors, thinkers, and creators, it's because of my network.

I'm teaching you how to build a network just like mine, whether it's for business or personal reasons. Just check out Jordan Harbinger.com slash course. It's free. I don't need your credit card. None of that stuff. Just go to Jordan Harbinger.com slash course. and by the way, many of the guests you hear on the show, they subscribe to the course, so come join us,

and you'll be in smart company where you belong. Now, here's Nina Schick. Deepfakes are interesting for me because this is something that I'd seen used in various instances, and we'll get into that here later in the show. Obviously, there's some sort of, I hate to say, obvious use, but like the first thing people think of, and a lot of people's first exposure to deep fakes is something to do with celebrities, and I'll let you explain that.

so I don't come across as the creep that I'm going to if I start off with it. But also, it can threaten our democracy, which is usually you don't equate like revenge porn with democratic threats or threats to the existence of the American or Western way of life. And I want to get into that here in the show, because what sort of technology could possibly be that powerful? Can you define what a deep fake is first for people who don't know? Sure. Well, a deep fake is essentially a, a type of synthetic or fake media. So that's to say a video, a piece of audio or an image that's either been manipulated by artificial intelligence or wholly generated by AI. And this amazing ability

of AI basically to create fake media has only been possible really for about the last five years and it's getting better and better and better. One of the things that's very interesting about it is that it's very good at recreating humans. And all that you need in order for AI to basically learn how to recreate a human is the right training data. So let's say, this has already been done, by the way, you want to get a machine learning system to generate images of fake people. You need to train it on a data set of lots of human faces. And then voila, you basically have an AI system that can now, at the click of a button, generate a fake human face that looks so authentic. like so real that to the naked human eye, you can't tell it's not real. And if you want to check it

out, you can check out, this person does not exist just to see how amazing it is at doing that. That's this person does not exist.com, right? This is a website. Exactly. And I'm thinking, okay, where do we get a huge database of human faces that are all digitized? And the answer is what, Facebook, Instagram? Facebook, Google search, very, very, very easy. The thing that is really alarming. Okay, so this instance, right, you're talking about people who don't exist. AI can make these faces. that look like real images. But another thing it can do is it's very good at basically, I call it hijacking your biometrics.

So if you have a digital footprint and who doesn't online, right, an image on LinkedIn, on Facebook, on Instagram, wherever, or a piece of audio of your voice, somebody can take that piece of data and get an AI to basically learn how you look and sound and make a deep fake of you saying and doing things that you never said and did. So it's not only very good at generating fake people, it's very good at generating you in fake scenarios. And the key thing is, again, we're just at the very, very beginning of this technology. So the amount of training data that's needed is going to be less and less and

less to the extent where it's only a couple of seconds of your voice or a single image taken from Instagram, Facebook, wherever. That's enough to make you the target of a deep fake. So to be really clear here, This is not someone editing media in a deceptive way. It's not me saying something like, well, I'm not racist and then taking out the word not, and then it's me admitting I'm a racist or something like that, right? It's not taking things out of context. It's creating new fake media out of whole cloth, often not using anything that I created that's even in the same context or in the same subject line.

This is maybe something like in the feet, and there are legit uses of this that are exciting, right like um disney can remake a movie and put audrey hepburn and frank sinatra acting right next to each other and then they can remake an entire movie based on that and it will look just like those two people acting which actually brings up a whole bunch of interesting questions from a legal perspective about like what do those people's estates get paid an acting fee for their dead relative performing that's probably a different show and a different topic but it's scary because you know we've seen like the and we'll get to some of these examples later we've seen like politicians be creatively edited, but this is like three levels up from just a creative, a slowing down the video

or something like that. And it can be used in pretty horrible ways. People don't really notice, when you go to This Person Does Not Exist.com, I kind of expected these cartoonish looking faces where like things were off a little and I was like, oh, you know, see how this eye looks. You can't tell that it's fake. That's the scariest part. Ten years ago, 20 years ago, remember the dancing baby video where it's like a little baby dancing around this cartoon, it doesn't look like a real baby dancing. Now you could make someone's actual baby do that and it would look as real, almost as real as if it were a video of their baby dancing. And that's scary. Absolutely. It is scary. And you're absolutely right to point out, however, and I think this is probably something

we should tackle right at the top. It's not only going to be used maliciously, right? Just like every amazing technology of the exponential age, it's going to be. a massive amplifier of human intention. So I think what we're looking at here when you think about synthetic media, that's media that's created by AI. It's such a paradigm changer that I think it's just going to be just as important in the history of human communication and in the history of the development of human society as the internet. Because increasingly, the way we produce media and consume media is going to be led by AI. It's increasingly, going to become synthetic. Some experts who I interviewed when I was writing my book think that in as

little as a decade's time, 90% of video content online could be synthetic. The problem is, is that we're no longer going to know what's authentic and what's synthetic. So you really lose your touch on what is real, what is reality, because thus far, you kind of still accept that what you see and hear is real, right? It's an extension of your own perception. We know there's been a long history of like tampering with the visual record that goes way back into the even the 19th century. For example, Stalin, the Soviet dictator used to like tamper had, there was an entire cottage industry that developed under his brutal regime that was dedicated to tampering with historical photographs. So every time his enemies were executed, you know,

they'd be kind of scratched out of and removed and edited out of photographs. But this is far more sophisticated than that, and it's also going to become accessible to everyone. So when you are talking about Hollywood level effects today, you still need a multi-million dollar budget. You need teams of special effects artists to achieve something that the best kind of Hollywood effects are. But by the end of the decade, a YouTuber with a budget of zero will be able to make the same type of thing, not only an image, but even in video form. And video is just becoming such an important mode of human communication. So you're essentially looking at a future where synthetic media is going to become ubiquitous. On the good side, it's going to completely transform entire

industries. Entertainment is going to become amazing. It's going to rewrite fashion and corporate communications. However, on the negative side, it's going to become the most powerful tool of visual disinformation known to humanity. And not only that, it's going to become accessible to everyone. So you're not just talking about a state actor or a terrorist group. Any teenager can create a deep fake because the AI will do all the heavy lifting. There's a lot to unpack there. And most of it is unfortunately kind of scary. I mean, it will be cool to be able to talk to a dead relative and get close. I don't know how this would work from a therapeutic standpoint, but to get closure or to like, imagine a lonely child whose parent is away, deployed in a military scenario or on a business

trip, and it's one o'clock in the morning, their time. They can't call that person, but they can be comforted at night by that person, or even just babies can get comforted by an AI mom who's finally getting some sleep. And I know you and I kind of have that. That's like the first application I want for this thing, right? The AI parents who just tells them to go back to sleep for God's sake for the 18th time at night. But it becomes more scary. The Stalin thing is creepy, right? I mean, so you're saying there was a group of people sitting in a basement somewhere and they said, hey, that guy who's in all these photos with the leader, he's been disappeared. We want to make sure that it looks like he never existed, which is very 1984 where they kind of reprint the old

newspapers to say, we've always been at war with South Eurasia. That seems like a miserable job, especially when they ask to take you out of the photos, to do it to yourself. But the fact that any teenager can do this, and there's no longer kind of any oversight or any decision-making body, even if it is a terrible dictator, who's making the moves that we can monitor. It's just like, some kid in the class made a video of another kid in the class. And now that's the bullying that happened. You can't put that toothpaste back in the tube when there's a deep fake app that everybody

gets for free from the app store, and it runs. on their iPhone 25. Yeah. There's already actually dozens of deep fake apps. And this is what I mean about this technology becoming accessible to almost everyone. Like what's changed in the past 30 years when you think about how we communicate in this global information ecosystem. Well, what's changed is the internet, social media, and smartphones, right? More than half of the world is connected into this ecosystem with a device that they hold in their hand. And the other half of the world who hasn't joined yet will be joining within the next 10 years. And on this device, you're going to be able to have the most sophisticated tools of visual

tampering that we have known, right? It's going to be very, very easy to kind of use these apps to create deepfakes. You already see dozens of deepfake apps that have been exploding on the app, on the app store. Again, look at the timeline here. This technology only started becoming available about three years ago. Can you imagine where we're going to be in 10 years? time. And I think you're really right to point out, you know, there obviously are going to be like bad use cases, good use cases, but there are profound philosophical questions that we need to answer

as a society as well. To what extent, for example, is it permissible? Should it be permissible for us to interact with relatives back from the dead, for example? As a form of therapy, is that okay if we resurrect them using AI to emulate their voice to speak to us? Or does that mean that we're retreating further and further from what is reality. And I think at the heart of it, this is a really philosophical question about are there forms of communication as we go forward? That should be entirely authentic and organic or can everything become synthesized in future? Is that a good thing? Is that a bad thing? Is it even desirable? Yeah, I think about this probably more than I, more than normal people do because there are so

many thousands of hours of my voice out there and hundreds and hundreds of hours of my face and my voice talking at the same time. So any expression that I have, the unshaven version of me that you're looking at right now versus like the clean cut version that I might have had had. Had we done this interview an hour from now, right? Like all of these variants are available. And it's cool to think that my great-grandchildren can be like, great-grandpa Jordan, what would you have done in this situation. My super smart AI can look over like 10,000 hours of my show and give good advice based on that, not that it would be relevant in 100 years, but you never know. But the downside is much more obvious and probably going to be much more common. And it reminds me of sort of every dystopian sci-fi movie

ever where Joaquin Phoenix orders an AI girlfriend from the internet. Have you seen this one? Yeah, yeah. Yeah, her, I think it's called. And she ends up being super smart and talking with like a million guys at the same time and then just vanishes and says, like, to hell with humanity. That's the good outcome that we can get. The other one is everyone is trapped in a box talking with their AI friends and family because the real version is completely incapable of human interaction at that point.

Yeah, I mean, ultimately, you can frame this as, you know, is this going to become the end of reality? I mean, what is reality? It's just an extension of your perception. But if what you perceive is no longer actually real, you know, what is happening in your life? And I think there is this danger that you retreat further and further into these virtual worlds. But I think the most potent danger directly with the threat of deepfakes is very specific to liberal democracies.

And that is that in order for a liberal democratic society to work, you need to have some kind of common basis of what is objective fact, what is reality? Because if you don't even agree the base of facts, you're not going to be able to agree to do anything in society. If you corrode the very notion that there is an objective reality, then you basically corrode the very fabric of what's holding democratic society together. And you see this with the increasing polarizing trends in Western democratic countries today. You see it very vividly in the United States, for instance. So you can imagine how much worse it's going to become when there is a deep fake to support every theory you might have. There's going to be no room for people to come together and compromise,

which of course is going to be devastating for a democratic style of government. It could be a very powerful tool for an authoritarian regime because you basically create a reality that isn't real and say this is the truth. It's interesting to see that all governments from the United States to China and Russia are investing heavily in artificial intelligence and synthetic media is a big part of that. their militaries are doing a lot of research into it. Yeah, that's terrifying because I can see right now, I mean, I work a lot with China, and I talk to Chinese people pretty much every day because I take Mandarin lessons in the morning. And I'm like the dangerous client they have because I go,

don't you think it's weird that you can't say Winnie the Pooh? And they're like, uh, but then 10 lessons in when they trust me, they're like, yeah, it's so weird that we can't say things in WeChat because it'll get deleted. And it's so weird that a colleague of mine has a red notice next to his picture that says this person doesn't pay his debts, so his social clout score is decreased. And it's only a matter of time. Like, if China gets this, and I don't know what the United States will do, and I'm not saying it'll be a better thing than what China is doing, but I can tell you what China will do is they will create a reality that they shove down everyone's throat. And if somebody creates a different one or shows

the real version of a video that's similar, that person will vanish. Or they will be completely unable to communicate. And they have this great firewall of China, as it's called, that, will search for unauthenticated videos that someone has made or unauthenticated videos that are actually the video that that's the real version of events. Those things will get shut down very quickly. They can already do that. So it's only going to improve in a decade. Exactly. And you're touching on one of the key points here with deep fakes. And it's literally called the Liar's dividend. It's the flip side of the coin. So we're increasingly going to be in this future where everything can be faked, right, including video, which to this day we kind of see as hard proof as evidence,

which is why it's so powerful, for example, in a trial. But the flip side is that everything that's authentic can also be denied, because if everything can be faked, everything can be denied. And that means that bad actors get lots of leeway to get away with things that otherwise would have been caught and documented and seen as evidence of their wrongdoing. And we're already starting to see that because even before synthetic media becomes ubiquitous, we're not there yet, right? We're just at the very, very start of the deep fake journey. We're already starting to see real-life political events where authentic media has been decried as a deep fake to basically let bad actors get away with their actions. The interesting point about China I'll make specifically

in regard to deep fakes is that China is the only country that has a blanket law that bans deep fakes. So essentially what that means is the central government has the power to say what piece of media is authentic or not. Yeah, that's scary. Like many things authoritarian slash Chinese Communist Party, and again, I always have to clarify the difference between the CCP and the people of China. The people of China have nothing to do with this. It's really scary that the government says, okay, you know what, this is real. I get why it sounds like a good idea because, hey, we need a central authority that can tell if something is fake or something is real. because of deepfakes, so we're just going to take that power.

But then it's like, well, wait, how do I know that you're not lying about what's real? And the answer is you don't. And that is the classic sort of technological authoritarian dilemma is, this is for your own protection. Trust me, you definitely want this. And people go, yeah, you're probably right. And then like 10 years later, they're like, how the hell did we agree to let this happen? And this is how it happens.

The appeal of an authoritarian government in the first place is we got to figure out what the hell is going on. You know what? I'll do it. I'll clean up all the things that are confusing for you and all of the things that you feel are destabilizing. I will clean that up for you. In return, you just have to live in the exact way that I want to and you don't have any pesky freedom to get in the way of a productive society. Exactly. And that problem is actually going to be felt even in the West as well, right? Because as knowledge of deepfakes and synthetic media starts to broaden the kind of distrust in society, this crisis of trust, lack of faith in institution, mainstream media. A lot of it justified, by the way, it's only going to augment. And if somebody says,

well, I'm going to be the central government, the government of the USA is going to be the authority on telling you what's authentic media, what's synthetic media, what's a deep fake or not, or Twitter is, or Facebook is, you can see why that's going to go down like a pile of sick amongst a certain part of the electorate, right? Who are you to tell me what's real and what's not? So I think this distrust that you see in societies, just the knowledge that synthetic media exists is actually going to harden that distrust

and augment the liar's dividend even before we start seeing loads of deepfakes out in the wild. You're listening to the Jordan Harbinger Show with our guest, Nina Schick. We'll be right back. And now back to Nina Schick on the Jordan Harbinger Show. Look, as a former attorney, I'm just waiting for a trial where there's a videotape of the defendant walking into a convenience store, whacking the clerk over the head with a club or something,

ripping out the cash from the cash drawer, and he goes, Your Honor, that wasn't me. That is a deep fake. I was at home watching Netflix at the time. And then they're going to have to go through a multi-the, the government's going to have to spend $80,000 authenticating the video frame by frame with some sort of expertise. Hopefully AI can do that, but it's looking like that. The arms race of being able to detect one AI that's created using another AI. Actually, do you know anything about that?

Like, how are we with detection here? So the first thing to say is that when AI started to be able to synthetically generate this media, it's been hugely exciting to the AI research community, right? It was a massive breakthrough that they got to this place where AI, could not just categorize. So this is kind of what they do, for example, to use driverless cars, so categorize things, but actually create something, generate something. So there's been a tremendous amount of interest to work on the generation side. It pretty quickly became clear that this could potentially be a problem. And then there have been research efforts on the

detection side as well. What's been covering us for the past three years in terms of detection has largely been digital kind of forensics. So researchers looking, at fake videos being like, oh, okay, we can tell that this is fake because the eyes are not blinking correctly or behind the ears, the background is blurred. So kind of looking for tell tell signs. However, that is only ever going to be a plaster, right? Because the eye is getting constantly better. So you can't rely on using these kind of forensics methods to identify deep fakes, not only because they won't exist pretty soon. And also because when they become ubiquitous, it's going to be impossible to use a human detector, right?

Nobody is, you're not going to have enough manpower to look through all the videos and kind of frame by frame look at which are real and which are not. So then you have to think about building an actual AI detector. Can you get AI to find something in the DNA of a piece of media to show that it's not authentic, that it's synthetically generated? The problem is that this is always going to be an adversarial game because the better the detector gets, the better the generator can become. So as soon as you build a really world-class detector, the people who are building the generator

can kind of look at what the detector is doing and find a way to beat it. And the second problem is that there are so many different ways of generating deep fakes that you're never going to have one size fits all, right? You're never going to have one model, which is going to be the ultimate detector, which you can put on all social media platforms to show you that a piece of content is synthetic or a deep fake. it's just not going to happen. So you're going to have to build lots and lots of different models

to find all these different deepfakes in the wild. And the jury is still out as to whether or not you get to a point where the generation sides of the AI becomes so good that you can never find a detector to tell that this is a piece of synthetic media. So can the AI become so good that even AI can't detect that it is synthetic? That's kind of inevitable almost, right? And even in the interim, it'll come down to the person. They'll say like, oh, is that the real Jordan or the fake one?

And then somebody who's got like a million hours of my videos is going to go, nope, my AI detector says that the real Jordan's left nostril is slightly smaller than the right nostril. And it's imperceptible by the human eye, but the computer can tell. And look, the AI version has two perfectly symmetrical nostrils. This is fake, you know? And then, but even then, you know, if you're really working, like, if I really want to make a deep fake of you or me, the AI will go, you know what, he's got two sort of different nostrils. her dimple on the right is deeper than the dimple on her left by like this imperceptible amount.

They're just going to fake that. The blip of detection, like any window we have for detecting is just going to get erased within, well, with AI, within minutes, days, hours a year. I mean, I don't really know. The technology, when you explained in the book how quickly this tech moves, I was blown away. Can you tell me about the Irishman? That was something where I just went, okay, this is like coming at us like a train. Yeah, I mean, a really powerful way to demonstrate just.

how incredible this technology is, is when you consider the Irishman, right? Martin Scorsese's latest blockbuster film, the storyline spans seven decades. And the big challenge he wanted to take was to kind of bring the old ensemble back together, Joe Pesci, Robert De Niro. But he wanted to dh his actors, right, to make them young again so he can tell this story over these seven decades. He started his project in 2015. And in order to do it, they knew it would be a big technical challenge.

challenge and a lot of special effects work. He had to basically film with this three-rate camera. He said it was awful. It was very, very cumbersome, annoying. And then they had a lot of post-production work. And if you went and saw the movie, by the way, the budget was in the millions of dollars. And the movie came out in 2019. And if you went and saw it in the cinema, it didn't bridge uncanny valley in the sense that you could kind of tell that they weren't, it didn't look quite right, right, the de-aging effect. Kind of good, but not entirely convincing. Fast forward to 2020, just last year.

And a YouTuber basically took free AI software. He had a budget of zero. And he, in a week, taking kind of the footage from the Irishman, had a go at de-aging the protagonist. And you can see it on YouTube. His result is arguably far, well, it's not arguably, it is far, far better than anything. Scorsese was able to achieve over five years with, lots of help, a huge budget, and teams of special effects artists.

So that's the power of AI. The budget for, I look this up, the budget for the Irishman, and this includes actor salaries and catering and everything, but still, 159 million US dollars. Let's assume that at least one million of those dollars was for the de-aging. It was probably a lot more than that. And this YouTuber did it with the gaming PC that he uses to play Fortnite on or whatever, right? Exactly.

Yeah. Exactly. And the barrier to entry, sometimes, When you see the public debate on deep fakes, you know, you can imagine that, oh, my God, I could just go and do this on my smartphone now. I could make a video of you using a racial slur. But the barrier to entry is still higher than that. You still need to have some level of expertise to make a really good deep fake. You need to have a computer with a lot of processing power.

However, the trend, the direction of travel is clear. These barriers that are basically technical are not going to be there for. much longer. Like I said, by the end of the decade, according to some experts, 90% of the video content online is going to be synthetic. There's a concept called, and brief aside, we may have covered this. I just don't know. There's a concept called processing fluency. What is this and what does this mean in terms of humans believing what they hear and see? The reason why deep fakes are just so visceral and powerful, right, is because their visual media. Processing fluency is basically the word that psychologist, a concept that psychologists use to refer to the fact that we as humans,

when we see something that looks and sounds right, we want to believe it's true. We're hardwired to do that. It's far more difficult for us to read something, you know, words, a piece of text on a piece of paper that makes outrageous claims and think that's true. Far, far, far, different when it comes to visual media. And the kind of Rubicon, if you will, when it comes to the manipulation of visual media thus far, because technology has consistently made it easier for visual media to be manipulated, right? In the 20th century, you had Stalin with his crafts, people kind of working 20 hours on each photograph to unpersoned people. Okay, by the time the Soviet Union fell in 1990, you had the release of Photoshop, the first kind of mass market image manipulation

software. To this day, 30 years later, people still look at images, be they on Instagram, the front of a magazine cover, and they don't realize that those images are edited, right? People think that they're real. But video was still pretty much in the domain of Hollywood. To really create a video piece of visual media that was manipulated to a highly sophisticated degree, you would still have the need for special effects artists, big, big budget. However, deep fakes and AI is going to blow all of that out of the water. So if you think about how powerful visual media is to us, first of all, just because we're wired to want to believe it, but also in the changing context of our information ecosystem, right?

Again, over the past 30 years, smartphone, internet, social media, so much of the content that we interact with is visual. It has to be. It's the only thing that kind of grabs us. This is potential leading to this perfect storm where you're tearing away the barriers, to manipulating even the most sophisticated type of media, the thing that was hardest to do, video. And you're doing it in this environment

where everybody in the world is not only consuming video, but producing video as well. I think it's 80% of internet traffic next year is going to be driven by video uploads and downloads. And there's another risk factor here, which is that perhaps in the Western world, where we've had a bit more time to kind of deal with the internet,

understand some of the ramifications and the darker sides of the disinformation. We've seen it, for instance, around COVID. There is a degree of digital literacy. However, when you start talking about the other half of the world that is yet to join that is soon going to be joining this information ecosystem, mostly in India and Africa, it is potentially even more damaging because the degree of digital literacy is obviously far lower. They're unprepared. Right, yeah, they are unprepared. And speaking of India, and this is a good transition here, there was a case of, I guess, deep fake porn used against a journalist in India. And it was very effective because, well, one, that type of thing would be very effective, generally, in a

conservative society, I would imagine. But also India has low media literacy. I don't know exactly what that means. Why don't you define it? And can you place India on maybe a spectrum with the U.S. and Canada? Because I would suspect the United States is higher, media literacy, but I don't know by how much because I, you know, I talked to my parents and their friends and I'm like, that, what you believe, that thing? You read it on like West Texas Patriot blog? It's not a news source. It's written by a guy in a basement. You made it up. Yeah, you're right to bring up the case of deep fake porn. Because I don't think we discussed it yet, but the kind of first application of deep fakes, the most malicious and widespread use, is a non-consensual pornography.

So when these tools started leaching out of the AI research community, the first place that Deepfakes basically buried their head on the internet was on Reddit at the end of 2017. And this is where I first came across them. And the name Deepfakes has stuck because the guy that was creating them was calling himself Deepfakes. He was a Redditor called Deepfakes, which was a portmanteau of deep learning and fakes. And he basically figured out how to use these free AI tools to. create fake porn videos of famous female celebrities. And he posted them on Reddit and then told other reditors how he had done it. And fake pornography has a long history. So this is nothing new, the idea. But the creations were just unlike anything anyone had seen before. This is not just a

photoshopped image of a celebrity's face stuck onto a porn star's body. This is a real live video where the celebrity is moving her face, she's got different expressions. The training data he used to create those would pretty easily available on Google image searchless videos and images. And as soon as he revealed how he had done it, Reddit went wild. Other people started creating their own deepfakes. There was a furorri and all the deep fake porn was basically banned off of Reddit. But since then, an entire deep fake porn ecosystem has developed online.

It's very, very easy to find deep fake porn of every. female celebrity imaginable, also political personalities, everyone from Ivanka Trump to Michelle Obama, but it is undeniably a gendered phenomenon, right? This is primarily targeted against women. And the alarming thing is, it's not just celebrities, normal women are being targeted now as well. Because as I mentioned, the training data, the AI is getting better and better. So less training data is needed. I can make a nude image of your sister, your wife. your mom from a single photo, for example, from Facebook. And it's not only women that are being targeted. Now, minors are being targeted as well. There was a case last year where basically they found

a bot being hosted on telegram, which is basically being used to generate fake images, nude images of normal women and young girls. And there were over 100,000 of these images just being publicly passed around in these telegram groups. The key point is it's not a tawdry women's issue, right? Because the AI can basically be used to hijack anyone's biometrics. So it's pretty clear that it's going to leach out into other forms of malicious use. We're already starting to see that with fraud, with political disinformation as a tool of intimidation. So this instance that you mentioned about this journalist in India, it was a case of somebody who was a very outspoken investigative journalist who was critical of the ruling Hindu nationalist party. So she was kind of a thorned

in the side of the government. And there was a very contentious case which was speaking out publicly upon involving, you know, a child, somebody who had been accused of child rape. And she said things that were obviously not in line with what some powerful people in the government wanted to hear. And she basically became targeted with a defaic porn video. And to any woman, this would be an utterly humiliating, devastating experience, perhaps even more so if, you know, you're in India where the status of women in society, I mean, I'm half South Asian, and being a woman in South Asia is probably one of the worst places in the world where we can be a woman. So not only was this fake video released, but all her personal information was released

as well. Her telephone number, her address, she was doxed. And she just became subject to this absolute campaign of harassment. People were emailing her, calling her, asking her for a rates of sex. She's a brave woman, right? She's this kind of intrepid investigative journalist, but that experience really scared her. And she even spoke about it later saying, it changed me. You know, I wish I had kind of the courage to just do what I did before, but that kind of hit too deep. So I think going back to the digital literacy question, yes, it was very effective in India, but you can see how something like that would be just as effective in Canada or in the United States. actually now we're starting to see this deep fake creation marketplace emerge on the internet

where business executives are being targeted, either by deep fake porn or perhaps you can imagine how in this environment if a video or an audio clip of you emerged using a racial slur, how that could potentially devastate your career and your business. So porn is the beginning. There are going to be many, many other use cases. And in terms of digital literacy, yes, in a place like India, it could potentially do more damage, but, you know, just as much damage in Canada and the United States as well. Sure, yeah. Look, we feel bad for celebrities and public figures when this happens, but how are we

going to feel when it's so easy that people are doing this and creating this for little kids in their seventh grade class? I mean, we're talking about potentially life-altering consequences for kids involved in this, whether it's a kid making it as a prank and getting in trouble and getting kicked out of school or getting arrested, or the little girl that is the subject to the video, which now has essentially this horrific child pornography floating around that everyone's laughing at, and she's seven or ten. Yeah, and that's actually what I'm more worried about, because I think when I first came to deep fakes and when I started researching and seeing what was going on, your mind goes to celebrities or how it can be used as disinformation against political figures.

But the reality is that these people are well-resourced, right? They can lawyer up. They have teams of people that can rebutt. You know, they have PR teams, crisis comms. But what happens? Like you say, if it's your 7-year-old child who's being bullied at school with a fake video or a nude image, which, by the way, is very, very easy to do right now. And as I already mentioned, they exist and they're being publicly shared.

And I think this is really, for me, one of the most alarming. things about the malicious use of deep fakes is that we're not just talking about celebrities and politicians. We're talking about you, me, our children. Every individual can be targeted in a very harmful way because, again, as I mentioned, the training data that's needed is becoming less and less. This is the Jordan Harbinger show with our guest, Nina Schick. We'll be right back. Thank you so much for listening to the show. I love the fact that I'm able to create this for you. my team loves that we create this for you, please do consider supporting those who support us.

And I know you're thinking, oh, but I can't remember the codes. They're hard to remember. Is it slash Jordan or slash Jordan 10? Don't worry about that. Just go to Jordan Harbinger.com slash deals. That's where you find all those codes, everything in one place, so you don't have to write it down. You can find it all at Jordan Harbinger.com slash deals.

And don't forget, we've got worksheets for today's episode. If you want some of the drills and exercises talked about during the show in one easy place, that link is in the show notes at jordan harbinger.com slash podcast. Now for the conclusion of our episode with Nina Schick. Right now, the things that are more readily available are, I guess you categorize them as cheap fix, right? We kind of touched on this in the beginning of the show where what was the classic example

from earlier last year was like a Nancy Pelosi video that I think they had slowed down and sped up at different points so it looked like she was, was it that she was drunk or I can't remember exactly the example here. Yeah, so Cheapank is basically the forbearer to a different point. deep fake. And essentially, a cheap fake is a type of visual media that's been manipulated or edited or taken out of context in some misleading way, but it's not been done with AI, right? So it can be nothing very sophisticated as all. And the interesting thing about cheap fakes is that they have already been so effective over the past 10 years. Over the past 10 years, right, I work in geopolitics

and looking at how the internet, social media and smartphones are changing politics all across the world. And one of the things that we've seen coming up again and again and again is the prevalence of cheap fakes, manipulated media to make service a political point or as a piece of disinformation. And it's been really interesting to track how during the course of 2020, these cheap fakes basically inundated the American political discourse in the context of the 2020 election. So the Pelosi video caused a furorre because basically it wasn't done very well, but they slowed down her words. to make her appear drunk like she was slurring her words. And this was, of course, amplified by the president of the United States, right?

So you're setting the tone from the top that in order to score political points, it's okay to use this kind of manipulated media, even though it's not done with AI. It's a cheap fake. It's already very, very effective. So if you consider how effective cheap fakes are now, then just think about where we're going to be in five to ten years time when it's deep fakes, which nobody can tell if they're authentic or not. Everyone's already far more distrustful of what media they can trust or they can't trust,

and the potential ramifications are vast. You start to see some of this in American discourse. So around the 2020 election, the use of cheap fakes was really prevalent, especially when the president started pushing disinformation that somehow the election was subject to widespread voter fraud. One of the ways he augmented this argument was, by tweeting a lot of authentic videos, real videos of election workers, right, shifting ballots, as though it were proof of ballots being stolen.

Now, if I'm telling you in five or seven years' time, you can literally create a video of, like, the ballots being torn up and a recording, you could actually do this now, a recording of Joe Biden saying, huh, like, burn those ballots. Imagine what that could potentially cause in terms of unrest, civil unrest. We saw what happened with the storming of the Capitol on January the 6th. I've seen the opinion polls consistently now showing that the majority of Republican voters believe that the election was stolen from Donald Trump. And if you have this kind of discord at the heart of a society, you just can't see how

that is going to become any better in future. It's hard to overcome something like that. What people don't realize is it's not just the video imagery that can be manipulated. It's also the vocal synthesis. This is something that kind of freaked me out before, too, because as a guy with 1,500 plus hours of my own voice, you can use the same AI. The vocal synthesis is even easier. And there are podcast editing programs now where you can give it enough sample of your own voice. And then my producer can fill in a word or two, or if there's an awkward transition, or let's say I cut off in the middle of a word and he needs to fill it in, he can just type it.

And it will say it in my voice. And if it's one or two words, you just don't notice. it's imperceptible, unless I'm listening to it very carefully. I can't even sense it. So, of course, some casual listener can't do that. There's a YouTuber that I think is called vocal synthesis or VS, and basically he can just make anyone say anything. And you mentioned before, pranksters and fraudsters, I can imagine a scenario in which I'm the controller for a major company. And I get a call from the CEO and a few members of the board that say, listen, we need to take care of this. We need you to get $2 million ready and wire it right away.

We missed a payment. Here's the banking and routing info. Do you have everything? Great. And then suddenly two months later, they come back and say, why did you wire $2 million to this random guy in Bangladesh or in New Jersey? What was the deal there? Well, you called me and told me to do that.

You and three members of the board were on the phone at the same time. And I said, you answered my questions. What do you mean? And then that person goes to prison or something for. embezzlement, right? Because it's so tricky that no one's going to, they're not going to have a recording of that phone call, right? Well, do you know the crazy thing? It's already happened. You had the first reported major case of defaic fraud, is reported the Wall Street Journal in 2019. Oh, wow. Yeah, it was the top executive of a energy company. He was the head of the British part of the company,

who had a phone call with the parent company where his boss was German. Now, The challenge of getting AI to synthesize your voice when the Fakes first started merging was actually more difficult because everybody's voice is so different. The cadence, the tone, the accent. But in the three years since then, we've come a long, long way. So this executive had a phone call with his German boss had no reason to suspect that he was speaking to anyone but his German boss. the accent was right, the cadence was right, the tone was right. His boss told him to wire a quarter of a million dollars to a Hungarian energy supplier. He believed he was talking to his boss, even though it was out of the ordinary.

He did it. And the alarm was only raised when his boss called him again and asked him to do it again. And then, you know, some alarm bells were raised. This is a bit suspicious. And what it was was Frosters assisted by AI. Now, this would have been like a high value target. whoever was behind it would have obviously invested quite a lot into the technology. You're talking about a high sum of money.

But as the technology becomes accessible, you can see how this is not only going to be used for a quarter of million dollars or $2 million top business executives. One favorite thing that scammers love to do, right, is the call where your distress relative is calling. I've been in a car accident. I've hit somebody with a car. I'm in an emergency. I'm in jail.

Can you imagine how devastating that would be where you as a parent or a husband? get a call, a distress call from one of your loved ones telling you they need money now, you're of course going to wire it to them. So you can see why this is going to become a favorite tool for pranksters, especially as the technology gets better. And especially as you already mentioned, right now, you already have programs that you can use as a podcaster, right? To kind of with audio track of your voice where you can re-synthesize what you said. But when it's possible to do that with, you know, just a few seconds of a clip? found on an Instagram story, a TikTok video,

you know, anyone's voice can be emulated. How do we defend against this? I mean, you mentioned AI detection methods and things like that, but it seems like we need to be skeptical but not cynical. Because if we become cynical, then to use the cliche, right, the terrorists win, because if Americans or Westerners or people in the UK, wherever you're listening, are busy fighting each other and we're just desperately trying to figure out

which facts are actual facts, we can't respond to information warfare, propaganda, or other very large issues that require bipartisan cooperation. And this is, that's really dangerous. Absolutely. I think that is key, being skeptical, but not being cynical. It was very interesting because obviously I speak publicly about deep fakes and I wrote my book on deep fakes when, for instance, certain things happen, people want to interpret it with their political slant on it, right? They become almost too skeptical. So when Donald Trump went into the hospital with COVID and came out with obviously stage managed video and photographs, you know, appearing, trying to project this image of

strength that he was fine. A lot of people contacted me on Twitter asking me, are these deep fake videos, you know? So it is important again when you think about the liar's dividend, which is, I think, already the most pervasive political effect of deepfakes before deepfakes themselves become ubiquitous where people don't want to trust anything, that we approach this with skepticism without being cynics. However, the second point I'd make is that this to me is a paradigm change, right? It's a paradigm change in the way that we communicate, and it's a paradigm change in the way that we interact and think about the future of this content, which we're constantly going to be interacting with this media that we're constantly going to be interacting with. So unfortunately,

there is no silver bullet answer. It is something that has to be addressed at a society-wide level, just like a question, like climate change, because ultimately you're talking about the integrity of the information ecosystem. And everything exists within this information in the ecosystem, Unless you're, you know, one of those people who lives in the Amazon in a tribe and still hasn't had any contact with the outside world, every single human being on this planet will come into contact with this information ecosystem. So how do you shore up the integrity of the system itself? There are some technical solutions that you can use, AI detection tools. You can also talk about media provenance.

And Adobe has been leading initiative here where you basically, flip the equation on its head and you make the tools available for those people who are purveyors of the truth, whether they're journalists or activists, to be able to prove by capturing in the hardware of their devices that the piece of media that they're sharing is authentic, right? And they interestingly already created a prototype with Trupec and Qualcomm, the chip manufacturer where they have a now working prototype of a phone where you can kind of like identify the authenticity of a piece of media throughout its life. In order for that to work, though, it has to become industry standard, right? Otherwise, you basically have this question

of like, well, your authenticity approval, photo versus my, again, the trust issue comes into play. And then second, you have to think about, and this is actually a larger question for society in the exponential age, right? We are living at a time where there's going to be more disruption and flux than potentially has ever been in the history of humanity. And the reason for that is because of the exponential technological change that's coming our way. So we have to think about how do we refit and re-skin our society to be able to function in this new world that we're creating. We have to rethink policy. We have to rethink regulation. We have to rethink legal systems. All of this means networked approach. And I think the first step is overwhelming. I think the first step

that's just conceptualizing it. And that's why I wrote my book, it's just kind of understanding how does all of this fit into this information ecosystem and what is this information ecosystem that's basically come into existence in the past 30 years. In the past, in our history, whenever there's been this huge moment of change

in the history of human communication, it's always transformed the future of human society. So, for example, without the printing press, there wouldn't have been the Reformation. The world would look entirely different. And, you know, what would the world look like today if it hadn't been for the kind of the technology of the information age, you know,

there was no internet, smartphone, social media. Again, a completely different place. So we're still reeling with those kind of changes that have come about in the last 30 years. And boom, now we're about to go into the age of synthetic media. It's going to take some time for society to catch up. But it is a society-wide effort. Nina Schick, thank you so much for coming on the show.

This is terrifying. And I'm probably not going to sleep for a few weeks on this. But no, really honestly, though, it is fascinating. I think the technology is fascinating. It seems like we're going to have to become much more media literate as a society. And I guess the good news is that that seems to happen naturally, right? Like, as people use computers and see these things more, like my own parents, 10 years ago, they were using AOL email and any video or any photo.

They would been like, wow. And I remember the first time I saw something that was Photoshopped, I was like, oh my gosh, It's a birdhead on a man's body. It must be a real bird man. And now I'm like, okay, this is fake. I'm not an idiot, right? So that sort of happens organically. The problem, though, it seems like, is media literacy.

It's trailing by like 10 or 20 years for many people. So there's going to be this window in which people just believe anything that they see. And I think we're in that window right now where you see a video of Nancy Pelosi looking drunk or Hillary Clinton stumbling around and you go, that's real. I saw it on Info Wars. you know, where the guy is just editing videos to make things look fake. It's going to be a while before we can clearly tell. And if we have our own bias, we have to mitigate that.

So there's like this decade long or longer window where we're chasing this. Our brains are chasing the computer technology to detect it. And like you said, there's going to be a time at which we can't do that anymore. So that is a little scary, but hopefully policy catches up in the meantime. It's going to have to. Thank you very much. Thank you. I've got some thoughts on this episode.

Before we get into that, Bob Sagitt shares how humor can be used as a coping mechanism for pain and the necessity of reinvention for career longevity and fulfillment. Here's a preview. When did you know that you were funny? Like, were you a class clown the whole time? Or was it? Last year. Last year, yeah.

Fame is bullshit. To let it go to your head, the moment you're cocky is the moment you've lost me as an audience. And a lot of people are attracted to it. You know, if I had known that secret, as a teenager, I would have had a lot of girlfriends or just been quite a stud. Because the key was not to care. I'm calm with my skin now. I don't know if it's evident during this thing. If I'm so calm, why the hell was I clicking this chapstick?

I'm demonstrating. Yeah, man. This is what you were hearing. But you were hearing it in a much slower clicking. It's like a hypnotic pattern for the listener. So I put your listeners to bed. For me to walk around scared or thinking about people that are trying to hold me back, nobody's holding me back.

If anybody's holding you back, it's you. You know, not you. Yeah. Not you, Jordan. If anyone's holding me back, it's you, Jordan. This podcast is going to, you're going to see a massive onslaught of listeners for your show. Talking Bob Sagitt's here for you is going to get eight more people. Provided we don't blow it in the last quarter here or the last 10 minutes here or whatever.

No, that's impossible. Well, we got more than that. We're in an hour and 12. and two of those minutes have to be cut. You're standing there with John Stamos, Dave Culey, and you think no one can see you. And I heard that there was a life-size doll. Yeah, let's forget this one.

This one's painful. You know too much. I'm going to have to kill you. I know I read your book. When I can get near you, I have to kill you. Yeah, we can hang out with the plague lifts. Oh, I can't wait. Maybe I'm alone here, but I don't like coronavirus.

We don't have wet markets in downtown L.A., right? No, we don't. I don't, at least not with, like, bats and pengolins and other stuff that you haven't heard of. No. What the hell happened? For more with Bob Sagget on how the big breaks can come from one of life's worst disappointments and Bob's proven remedy for dealing with the haters and we all have haters,

check out episode 372 on the Jordan Harbinger Show with Bob Sagitt. Man, what a brilliant woman she is. You know she speaks seven languages? Wow, unbelievable. My mission on this show is to create more skeptics without birthing more cynics at the same time. And I'm telling you guys, this is about to become a major, major issue in everything, from entertainment to marketing to politics to identity theft and beyond. Misinformation and disinformation, these have always existed, but nothing at the scale that we are seeing now. And nothing at the scale and level of sophistication that we will see in the near future.

AI characters could generate an entire work history and resume. For example, they could pose as a journalist having interviewed dozens of super famous people, except all the videos are fake. They're just deep fakes. Personally, I'm looking forward to never doing interviews again. I'm just going to sit around making deep fakes of me talking with Barack Obama and Donald Trump. Maybe even at the same time. Who knows? Black PR firms, so black hat PR firms, right? The bad guys. They can be creating disinformation to lower competitor share price or otherwise harm corporations

here in the United States and abroad. Our information landscape is becoming less and less stable and certainly less trustworthy. But we shouldn't just not trust any media. That's the issue we are seeing with these QAnon cultists and people who only read fake news blogs from the fringes that confirm their existing bias or conspiracy theories. We need AI detection tools because deep fakes will soon be so good,

humans won't be able to tell the difference. The insurance industry, by the way, I think they're going to be on the front end of this because they could be bankrupted with fake claims, video and photographic evidence. The technology here is really going to be something. and they're going to have to use watermarking and special device IDs signatures. Who knows?

Blockchain something, something could be a part of this, right? We just don't know. It's got to be tackled somehow. We know that disinformation is pervasive and working against us because we can't even seem to agree that Russia interfered with the 2016 election, despite unanimous conclusions of all relevant intelligence and law enforcement agencies, foreign and domestic academia, think tanks, et cetera. I'm not saying they put their thumb on the scale,

but we know they at least tried. The idea behind this is not just to confuse the issue, but foreign powers and frankly domestic powers, they want to break people's commitment to even staying or becoming informed at all in the first place. They want to make us they, I sound like a conspiracy theorist myself now, they want to make us feel hopeless so that we just don't even try anymore. I also wonder, look, my son, he's 19 months old right now, 20 months old. Will he believe anything at all that he sees because of things like this, anything he sees on a screen could have been created by somebody on a smartphone. And he may sort of natively slash intuitively know that he's not necessarily

looking at real people. But that's a generation from now. Will it take this long? We won't be able to believe our own eyes and ears at all soon. By the way, we did a whole episode on Russian election interference. That's David Scheimer. That was episode 419. It goes through the historical basis for this. A lot of evidence, a lot of proof. United States doing it as well. It's not a new thing, but if you want to deep dive on that, episode 419, René DeResta also on the show, we talked about social media and mass media manipulation. So if you want to do a deeper dive on that, she talks about the Soviet playbook and how that fools a lot of old people, not just old folks or quote-unquote dumb right-wing or

dumb left-wing people, we dive into that whole topic with Renee DeResta on episode 420 of this show as well. Big thank you to Nina Shek. Her book title is Deep Fakes and the Imphocalypse or Imphocalypse. I guess that's one of those things that's better written and read than said. Links to her stuff will be in the website in the show notes. Please use our website links if you buy the books. All those things add up. They do help support the show. Worksheets for the episode are in the show notes.

Transcripts are in the show notes. There's a video of this interview going up on our YouTube channel atjordanharbinger.com I'm at Jordan Harbinger on both Twitter and Instagram or just hit me on LinkedIn. I'm teaching you how to connect with great people and manage relationships using systems, software, and tiny habits over at our six-minute networking course. That course is free. I'm not selling you anything. You don't even need your dang credit card info or your name, nothing. Just go to Jordan Harbinger.com slash course. I will teach you how to dig the well before you get thirsty and create relationships before you need them. And most of the guests on the show, they subscribe to the course.

Come join us. You'll be in smart company where you belong. This show is created in association with Podcast One. My team is Jen Harbinger, Jace Sanderson, Robert Fogarty, Millio Campo, Ian Baird, Josh Ballard, and Gabriel Mizrahi. Remember, we rise by lifting others. The fee for this show is that you share it with friends when you find something useful or interesting. If you know somebody who's interested in the future of media or just wants to be scared about what's going to be coming for us in technology world here, or somebody who's an outside-the-box thinker, I think this is a good one for them. Hopefully you find something great in every episode. Please share the show with those you care about. In the meantime, do your best to apply what you hear on the show so you can live what you listen. And we'll see you next time. This episode is sponsored in part by Something You Should Know podcast. Finding a new great podcast shouldn't be this hard, so let me save you some time. If you like the Jordan Harbinger show, you'll probably like Something You Should Know with Mike Carruthers. It's one of those shows that makes you smarter in a practical, useful way.

Same curiosity vibe we go for here, just in a fast, focused format. Mike brings on top experts and asks the exact questions that you'd want to ask, and the topics are all over the place in the best way. Recently, they've covered things like why we care so much what other people think, the benefits of laughter, why sports fans get so invested, and what makes people like you or not, the through line is always the same. Smart ideas you can actually use in real life. Something you should know has been featured in Apple's shows we love, and it's got thousands of five-star reviews because it's consistently interesting. So if you want another show that scratches that I want to understand

how people in the world really work itch, search for something you should know wherever you get your podcasts. Look for the bright yellow light bulb and start listening. You can thank me later.