Exposing the deception of DeepFakes

This week we bend reality to expose the deceptions of deepfake videos. We talk about what they are, why they are so dangerous, and what you can do to detect and resist their insidious influence. In a political environment rife with distrust, disinformation, and conspiracy theories, deepfakes are being weaponized and proliferated as the latest form of state-sponsored information warfare. Join us for an episode scarier than your favorite horror movie, because this AI bogeyman is real! Sponsors: DigitalOcean https://do.co/changelog – Check out DigitalOcean’s dedicated vCPU Droplets with dedicated vCPU threads. Get started for free with a $50 credit. Learn more at do.co/changelog https://do.co/changelog. DataEngPodcast https://www.dataengineeringpodcast.com?utm_source=changelog&utm_medium=podcast&utm_campaign=practicalai – A podcast about data engineering and modern data infrastructure. Brain Science https://changelog.com/brainscience – For the curious! Brain Science is our new podcast exploring the inner-workings of the human brain to understand behavior change, habit formation, mental health, and being human. It’s Brain Science applied — not just how does the brain work, but how do we apply what we know about the brain to transform our lives. Fastly https://www.fastly.com/?utm_source=changelog&utm_medium=podcast&utm_campaign=changelog-sponsorship – Our bandwidth partner. Fastly powers fast, secure, and scalable digital experiences. Move beyond your content delivery network to their powerful edge cloud platform. Learn more at fastly.com https://www.fastly.com/?utm_source=changelog&utm_medium=podcast&utm_campaign=changelog-sponsorship. Featuring: Chris Benson – Website https://chrisbenson.com, GitHub https://github.com/chrisbenson, LinkedIn https://www.linkedin.com/in/chrisbenson, X https://x.com/chrisbensonDaniel Whitenack – Website https://www.datadan.io/, GitHub https://github.com/dwhitena, X https://x.com/dwhitenaShow Notes: The Dark Knight’s Tale (DeepFake) https://youtu.be/TgcvQA6-qBgOfficial PyTorch Implementation of StarGAN - CVPR 2018 https://github.com/yunjey/StarGANFew-Shot Adversarial Learning of Realistic Neural Talking Head Models (YouTube) https://www.youtube.com/watch?v=p1b5aiTrGzYFew-Shot Adversarial Learning of Realistic Neural Talking Head Models (PDF) https://arxiv.org/pdf/1905.08233.pdfMelNet: A Generative Model for Audio in the Frequency Domain https://arxiv.org/abs/1906.01083Text-based Editing of Talking-head Video https://arxiv.org/pdf/1906.01524.pdfBetter Language Models and Their Implications (GPT-2) https://openai.com/blog/better-language-modelsFake-porn videos are being weaponized to harass and humiliate women: ‘Everybody is a potential target’ https://www.washingtonpost.com/technology/2018/12/30/fake-porn-videos-are-being-weaponized-harass-humiliate-women-everybody-is-potential-targetThe National Security Challenges of Artificial Intelligence, Manipulated Media, and “Deepfakes” – 13 June 2019 https://intelligence.house.gov/uploadedfiles/clint_watts_-_house_select_committee_on_intelligence_-_ai__deep_fakes_-_13_june_2019.pdfStudent with access to TPU credits reproduced GPT2-1.5B and plan to release model https://www.reddit.com/r/MachineLearning/comments/by4n8p/p_student_with_access_to_tpu_credits_reproducedPapers With Code: fake+detection https://paperswithcode.com/search?q=fake+detectionIn Ictu Oculi: Exposing AI Generated Fake Face Videos by Detecting Eye Blinking https://paperswithcode.com/paper/in-ictu-oculi-exposing-ai-generated-fake-faceA Retrospective Analysis of the Fake News Challenge Stance Detection Task https://arxiv.org/pdf/1806.05180v1.pdfDeepfake debunking tool may protect presidential candidates. For now. Sometimes https://www.cnet.com/news/deepfake-debunking-tool-may-protect-2020-presidential-candidates-trump-warren-obama-hillary-clintonThe Era of Fake Videos is Here https://www.deeptracelabs.comS.3805 - Malicious Deep Fake Prohibition Act of 2018 https://www.congress.gov/bill/115th-congress/senate-bill/3805/textPractical AI episode #47 - GANs, RL, and transfer learning oh my! https://changelog.com/practicalai/47Synthetic Realities: Deep Learning for Detecting AudioVisual Fakes https://icml.cc/Conferences/2019/ScheduleMultitrack?event=3512Machine Learning for fake news detection: theory and practice https://github.com/amld/workshop-fake-news-detectionUpcoming Events: Register for upcoming webinars here https://practicalai.fm/webinars!

Speaker names marked ? are inferred from speaking time and the show’s known hosts — they are not voice-verified.

• 0:00Speaker 00

  Being with your change log is provided by Fastly. Learn more at fastly.com. We move fast and fix things here at Change Law because of Rollbar. Check them out at rollbar.com. And we're hosted on Linode Cloud servers. Head to linode.com/changelog.

• 0:16Speaker 00

  This episode is brought to you by Digital Ocean, the simplest cloud platform out there. And we're excited to share they now offer dedicated virtual droplets. And unlike standard droplets, which use shared virtual CPU threads, their two performance plans, general purpose and CPU optimized. They have dedicated virtual CPU threads. This translates to higher performance and increased consistency during CPU intensive processes. So if you have build boxes, C I C D, video encoding, machine learning, ad serving, game servers, databases, batch processing, data mining, application servers, or active front end web servers that need to be full duty CPU all day, every day, then check out Digital Ocean's dedicated virtual CPU droplets. Pricing is very competitive, starting at 40 bucks a month.

• 1:01Speaker 00

  Learn more and get started for free with a $100 credit at do.co/changelog. Again, do.co/changelog.

• 1:21Speaker 00

  Welcome to Practical AI, a weekly podcast about making artificial intelligence practical, productive, and accessible to everyone. This is where conversations around AI, machine learning, and data science happen. Join the community and slay with us around various topics of the show at changelog.com/community. Follow us on Twitter. We're at Practical AI FM. And now into the show.

• 1:47Chris Benson?

  Hey, welcome to the Practical AI Podcast. This is going to be another fully connected episode where Daniel and I keep you fully connected with everything that's happening in the AI community. We're going to take some time to discuss the latest AI news and we're going to dig into some learning resources to help you level up on your machine learning game. My name is Chris Benson. I'm one of the co-hosts. I am Chief Strategist for Artificial Intelligence, High Performance Computing, and AI Ethics at Lockheed Martin. And with me is my co-host Daniel Whitenack, who is a data scientist with S I L International. How's it going today, Daniel?

• 2:18Daniel Whitenack?

  It's going great. How about with you, Chris?

• 2:21Chris Benson?

  Going very well. As we are recording this, I just got back from the LiveWorks Tech Conference in Boston. Had a good time there, gave a talk, and it went well, so I'm a happy camper.

• 2:34Daniel Whitenack?

  Awesome. Yeah, well, things are looking up here as well. Over the past month or so, it seems like my internet at home kind of has gradually been degrading, and I haven't been able to figure out why. I've updated my computer and done all the restarts of everything, checked all the things. But had the technician out today. It turns out that squirrels were eating the cables coming into my house, so producing obvious degradation. So I'm happy that that's actually figured out, and I have internet, good internet again.

• 3:11Chris Benson?

  There you go. Humane removal of squirrels to be considered here.

• 3:15Daniel Whitenack?

  Or reconfiguring the cable positioning to get a little bit harder.

• 3:21Chris Benson?

  Or that. Either one will work. Yeah. Okay. So this, we actually have a really, really timely topic right now, because we are having conversations about this constantly. So today, we wanted to talk about deep fakes, and we're kind of going to cover an overview with it. And so much is happening in the news right now regarding deep fakes. We'll tell you what, we'll tell everyone what they are and such as that. But everywhere I go, whether I was in Boston, or we had, this past week, we had our monthly Atlanta Deep Learning meetup. The topic came up there. It's just kind of coming up everywhere, and it's coming up in the news on a daily basis. So we're going to delve into this topic and see what we find.

• 4:04Daniel Whitenack?

  Yeah, and actually, even just this morning, the Policy Director from OpenAI was testifying before the House Intelligence Committee here in the US. The topic of that was the National Security Challenge of Artificial Intelligence, Manipulated Media, and Deep Fakes. So this is reaching the highest levels of government, and certainly something that people ask about a lot, and something that we, I think it's time that we spend time talking about it on the podcast. So I'm glad you brought up the idea.

• 4:39Chris Benson?

  Yeah, I actually, I saw your tweet about that was going on and tuned in. I was, I was too late to catch the beginning of the show of the testimony, but I saw at least the full second half and it was fascinating and it was interesting to see how startled the, the members of the Intelligence Committee were receiving this information. I think they already had some insight into it, but it was one of the reps who I, whose name I don't have at the top of my head noted that it was, it was very scary stuff. And so the potential of how it can be used nefariously and we'll certainly get into that today. So we're going to kind of talk about what deep fakes are and then get into kind of how

• 5:22Chris Benson?

  they can be used, how they have been used, what can be done to prevent bad actors, that such, such thing. And I'll certainly, I took notes and will refer to some of the congressional testimony in this episode.

• 5:34Daniel Whitenack?

  Awesome. I appreciate that. Yeah. And I guess this episode will be kind of a downer. So sorry, folks. Sorry in advance. We'll try to at least keep it interesting, even if we are talking about sort of quote unquote dangerous things and you know, try to, you know, bring some of our thoughts into it, but try to give you a kind of good overview of, of the topic. So maybe one, one good way to start is just by defining what a deep fake is. So in my understanding, a deep fake, the deep part really refers to like deep learning models. Now we talk a lot about deep learning on this podcast. So if you want to know more

• 6:17Daniel Whitenack?

  about that, there's certainly a lot of links throughout the podcast about deep learning. But now I guess the question is for deep fakes, if we're talking about deep learning models faking something, so generating fakes of something.

• 6:30Chris Benson?

  Media.

• 6:31Daniel Whitenack?

  Yeah. What are we, what are we talking about? So what are deep fakes faking, Chris?

• 6:36Chris Benson?

  So deep fakes are where you, and I'll, I'll get to a specific example, but that's where you are using deep learning technology, and we'll talk about the specifics in a moment, to change, to either create or change. Videos that that may be out there. It could also be audio. It can be any kind of media that people will watch to take in information, whatever that might be, and change those so that what you are seeing and hearing is not actually what really happened with the original unchanged video. And so it opens the door for all types of manipulation.

• 7:13Daniel Whitenack?

  Yep.

• 7:13Chris Benson?

  And so that's a that's a start. It's a broad, very broad definition.

• 7:16Daniel Whitenack?

  Yeah, so the some of our regular listeners might remember that I think our last fully connected episode, we kind of did an overview of various advanced sort of methodologies in the AI world. One of those was GANs. And in that episode, we talked about how GANs or generative adversarial networks are. One of the things they can do is generate art or generate images or generate videos or change styling or these other sorts of things. So this is one sort of methodology that's applied, a deep learning methodology that's applied to generate certain things. Like you were saying, Chris, I think the thing that people probably think of right away when they think of deep fakes, if they've seen some of these things is the video things.

• 8:01Daniel Whitenack?

  So there's there's been some like funny ones or satirical ones as well. So I was just watching one before the episode where kind of Joker's face was, you know, applied to these different videos. There've been ones with like President Obama's face where he's he's dancing and and other things. So it's kind of like, you know, the Joker or President Obama didn't actually act in those videos, but their face is in those videos doing certain things when when they didn't. So I think that that's probably what comes to mind. There's, of course, like you said, this isn't only video.

• 8:43Daniel Whitenack?

  So we talked even in the last fully connected episode about generating kind of fake people, so pictures of fake people. But in this case, probably the deep fake part of it would be faking someone's face with an emotion or an expression or a scene. That they were never in or faking someone's voice saying something that they never said, or maybe it's both of those things together, faking someone's voice and in a video saying something they they never said. So replicating someone's voice and mouth movements. But then it also goes beyond kind of the video and imagery into text as well. Of course, there was a lot of focus on OpenAI's GPT-2 model recently.

• 9:27Daniel Whitenack?

  This this spring, which was capable of producing some like really realistic sort of news kind of articles or or text on certain subjects and that sort of thing. So generating text in a certain style or based on a certain subject or something like that is also something that that should be considered as we're kind of talking through this.

• 9:49Chris Benson?

  Absolutely. And and there have been it's it's been in the media so much recently because there have been several notable things that that I imagine most or all of our listeners are already somewhat familiar with. There was recently a video of the Speaker of the House of Representatives in the US Congress, Nancy Pelosi, where she in the video she was speaking and they made her appear kind of, I think the the most common references were drunk and slurred speech and such as that. And, you know, that was that one thing where it changed the characterization of her having that conversation. And that's a big part of it. It's not just changing the words. You can change how people appear to you. And, you know, and and a lot of people believe that video.

• 10:34Chris Benson?

  They they were like, oh, there's this video of Nancy Pelosi, you know, drunk on stage or or whatever, you know, talking. And so that was one. And there was a bit of an uproar and Facebook refused to take it down. So earlier this week, there was a video posted on Facebook and other places of Mark Zuckerberg, the CEO of Facebook, saying things that he never said, where they took a, I think it was a 2017 talk of his and changed what he was saying. And so I think it was a, you know, I think that was intended somewhat as a look. I told you so, Facebook, you should have taken down the the political. See video as and so you know, how does it feel? So those are certainly big. And I know in my day job, working at Lockheed Martins,

• 11:17Chris Benson?

  you know, focusing on national security issues. This is certainly something that we talk about because there are all sorts of uses here. And we'll get into some of that, those potential use cases, you know, in the world as we go here.

• 11:30Daniel Whitenack?

  Now, some of you are probably thinking, well, this kind of has already been around for a while. So Hollywood and movie studios have been doing sort of CGI and video tricks and movie tricks for quite some time that might kind of put someone in a scene that they were, you know, standing on a mountain where they weren't really. Or maybe it's like superimposing a face. I'm thinking of the, the Star Wars example.

• 11:58Chris Benson?

  Yep.

• 12:00Daniel Whitenack?

  With Princess Leia. So there, there's certainly something that's not new. I think what is new is kind of traditionally or always in the past, these sorts of techniques were pretty much restricted to experts. So it required a lot of effort or quite a lot of time and money to kind of pull off these things convincingly. Now with these deep learning networks that have kind of these, these large encoding layers and decoding layers, you can have a training data set where you have a bunch of images with, you know, one person's face and output, you know, with the, the other face or, or from a,

• 12:42Daniel Whitenack?

  you know, with a certain pose or, or representation of a body and then output with someone in that pose. Or whatever it is, if you have the right training data, now it's just a matter of applying models that are existing to, to that training data. And it really doesn't require a lot of expertise beyond maybe some hours of compute time on AWS or something like that. And in some cases, you can even get pre-trained models. So like in terms of kind of generating these new looking faces, you know, corresponding to a certain facial expression or something like that. So in this case, maybe the input is. Facial expression or something and the output is someone you know in that facial expression.

• 13:25Daniel Whitenack?

  You know, that may only take a few, a few images to fine tune on if you've already got a very good pre-trained model. So it's not even like you have to have a bunch of images of someone to be able to fake them in a certain, in a certain position or with a certain facial expression or in a certain scene. So I think the big difference now or what's been the development recently is kind of the ease with which you're able to do these sorts of things.

• 13:52Chris Benson?

  That's true, and there's, there's a whole host of software programs that have come out that are basically kind of dumbing down the process so that you don't have to have a lot of deep learning training. In some cases, you can just install the applications on your system and they come with, you know, there's some for Mac and Windows and things like that. Some of the really well-known ones that people talk about or there was kind of the original, what was called the fake app program. An open source version of that followed called Deep Face Lab, which I've seen referenced quite a lot. Although that has recently closed down because the, the primary maintainer has moved on.

• 14:33Chris Benson?

  Though that person is, is encouraging other people to, to use the, the source and, and move on and do things with it if they choose. There's Face Swap, there's My Face, I'm sorry, My Fake App, which is on Bitbucket. And so there's, there's, and I think you're going to see more and more of these coming about. And in addition to that, obviously you can use the standard tools, you can use Keras and TensorFlow to do the same things with the data as you just alluded to. So it's, I think that the key takeaway there is that, you know, when, when it was a Hollywood thing, it was a highly skilled thing that required, you know, special software, things that you would find in a movie studio, but not, not everywhere. And, and that's changed. It's now something that, that any computer savvy person can handle.

• 15:16Chris Benson?

  And so I think that's, that's what's changed as we hit 2019 is that it's now been democratized.

• 15:30Speaker 03

  The Data Engineering Podcast is a weekly deep dive on modern data management with the engineers and entrepreneurs who are shaping the industry. Go behind the scenes on the tools, techniques, and difficulties of data engineering so you can learn and keep up with the knowledge to make you and your business successful.

• 15:47Speaker 04

  Can you give a bit of an outline about the motivation for choosing Jupiter notebooks in particular as the core interface for your data teams?

• 15:55Speaker 05

  Yeah, and actually, when I first joined Netflix, it was a sort of toss at me. And I was definitely like, are we crazy? And the answer was like, we might be a little crazy.

• 16:03Speaker 03

  Go to dataengineeringpodcast.com to listen, subscribe, and share it with your friends and colleagues.

• 16:22Daniel Whitenack?

  All right, Chris, so let's maybe get to the more, I don't know, important or depressing or however you want to put it, part of the, part of this story is, you know, in some ways, it's kind of cool that this technology exists in the sense that, you know, technology wise and kind of intellectually, this is kind of interesting that these techniques can do this with, you know, with such little data or with such realism. And when you just think about it from that point of view, you know, technology wise, it's pretty interesting. But what makes this sort of technology? Why is there so much hype around it? Why are

• 17:02Daniel Whitenack?

  these sorts of methods that produce these deep fakes? Why are they so dangerous? What's your perspective on that?

• 17:10Chris Benson?

  Well, I think there's, there's a number of reasons that we can, we can kind of work our way through. You know, I guess you can start off with the fact that since the data is so available, you can get videos from so many places now, and with people using their smartphones to take video and, you know, post it out on social media and stuff, it's not, you know, historically, deep fakes have, have really centered on things like celebrity. And you know, they would put a celebrity's face on, you know, a pornographic video or take a politician. I saw something on one of the software, on one of the deep fake software sites where they were superimposing Nicolas Cage's face

• 17:52Chris Benson?

  on Donald Trump and you could kind of clearly see it was, it was, it appeared to be Trump talking, but the facial expressions were clearly Nicolas Cage's. And those were, you know, kind of goofy, so long as that was as far as you were taking it and they could be a little bit fun meme like. But I think that obviously opens the same door for people who are out to, to cause harm to others and, and that could be at a lot of different levels. It could be as personal as harassing a bad actor, harassing someone they know, you know, if they, I'm just making something up. Broke up with the girlfriend and they had video of their girlfriend talking and they could, they could take that and, you know, take some other bad footage, whatever you want to do.

• 18:33Chris Benson?

  And, and put that out there to humiliate them. And I think there have been some, some instances of that. I think you had, you had mentioned that there was one that the Washington Post had reported. I know when we were talking beforehand.

• 18:44Daniel Whitenack?

  Yeah. So there, there's kind of this, I mean, one of the first ways I think this had surfaced and people have used it in this sort of harassing way is, is kind of the pornographic use, like you had said before, where, you know, maybe before there were certain people that tried this with celebrities. At least kind of leading up until very recently. Now I think it's, it's very real that, you know, if someone had this sort of video of, of someone they knew, you know, that in their circle of friends or acquaintances, they could harass them in this way by making, you know,

• 19:21Daniel Whitenack?

  explicit content with that person's face. And, you know, it looks so real that if they kind of propagate that, then kind of the harm is done. Before, you know, it may, it may come out or it may never come out that that, that was a fake. So it's, it's definitely a concern in terms of how this could affect, you know, real, real people's lives.

• 19:45Chris Benson?

  You know, it's funny. Just as when you said that about whether or not people after the fact would learn that that was addressed in the congressional hearing today. It was noted that when one of these videos goes out and goes viral, even if it's widely reported that the video was a fake after the fact, it tends not to hit as many people. And so you inevitably have changed the landscape by the initial post and then the quote fix afterwards doesn't actually completely fix it. And they noted that psychologically, that even if people know it was a fake that they

• 20:24Chris Benson?

  saw, that psychologically, they still kind of hold on to some of that bias that was introduced through the fake. So even if I, you know, even if I found out that the picture of President Trump and Nicolas Cage was in fact a deep fake, as it was, in theory, there's the potential for that to influence me in some way, depending on what the video author was shooting for. So it's an interesting side effect.

• 20:51Daniel Whitenack?

  Yeah, and part of the hype maybe that's been generated recently and part of the momentum to discuss these things, I think, has been that shift from, you know, people's thoughts of before, thinking that, well, yes, whether it was satirical, like a joke sort of thing, or whether it was actually harassment and humiliation, like in terms of the pornography type of stuff, you know, that was maybe restricted to celebrities and people are thinking, oh, well, those people put themselves out in public, there's a lot of video of them, so they're kind of, you know, asking for this sort of thing, which is kind of sad anyway, because no one should have to be subjected to that if they don't want it. But now it's like, you think about any video you see, whether it's a video of someone you

• 21:36Daniel Whitenack?

  know on Facebook, or, you know, someone that's not a high-profile celebrity, now there's this potential that even those videos are faked in some way or another to influence you, or at least there's the potential of that happening, which is kind of a shock when you think about it.

• 21:54Chris Benson?

  Yeah, it really is. I mean, it's there's such widespread application from a very personal level, as we were talking about a moment ago, you know, all the way to large societal concerns. And it is a technology that is kind of fluid enough in its application to where you can scope it however you want. And and you're seeing everything from that level all the way up to nation states using it, you know, to to influence others, which we'll talk about in just a moment. But I think one of the things that I think is certainly contributing to it being used this way and in such a successful frame is the fact that when your political environment is what we all know it to be, and and we are, we're

• 22:38Chris Benson?

  very polarized, we're very tribe oriented, and we recognize that, you know, there are messages that are supporting each of those viewpoints, that by itself, before you get to deepfakes, already kind of introduces in a lot of people's minds, the potential for conspiracy theories and and and thinking about others in ways that may not be entirely realistic anyway, regardless of which side you're on. And so when you throw in the nefarious intent that deepfakes can lend themselves to, that just exacerbates that situation. So we really have created not only here in the United States, but in places around the world, an environment where we're very susceptible to this technology being used

• 23:22Chris Benson?

  against us now. And it's certainly something that if we are, if we are to navigate safely through this, we're going to have to, we're going to have to figure out ways of coping. And we'll address some of those in the congressional hearing, address some of that. We'll talk about that in a few minutes.

• 23:35Daniel Whitenack?

  Yeah, it is interesting that even the so just the fact that these deep fake videos exist, it creates kind of an excuse, an extra kind of excuse for people that don't want to face the the truth or want to create conspiracy theories. So this has already been seen around around the world. You know, I was, I was reading in the, this is a great Washington Post article that I'll link in in our show notes, but it was talking about in Malaysia, there's a kind of viral Video clip of a man confessing to to certain things with the local cabinet minister. And that's kind of, you know, that questionable stuff is being kind of thrown into, oh, well, that's just a deep fake.

• 24:18Daniel Whitenack?

  And there's similar things in Africa, even, you know, videos of leaders who have actually contributed to, those videos have contributed and the controversy over those have has contributed to, you know, coup attempts and other things where it wasn't even, I guess we don't, I don't know if it's known in those cases, but even just the questioning of those videos, if they're a deep fake, created enough uncertainty that it actually created political and military turmoil.

• 24:48Chris Benson?

  Yeah, and those aren't the only places. There was actually an example very similar to your African when in the testimony before Congress today about that. And so just to go ahead and get to that, there was, it was a statement that was prepared for the Permanent Select Committee on Intelligence within the US House of Representatives. And the, the person testifying was Clint Watts of the Foreign Policy Research Institution, I'm sorry, Institute. And the title of the brief was "The National Security Challenges of Artificial Intelligence, Manipulated Media, and Deep Fakes." And I read through the, the document that was submitted initially that kind of represented their viewpoint in addition to hearing some of the testimony.

• 25:32Chris Benson?

  And it wasn't actually at the top, but there was a paragraph that jumped out, and it's very short, and I'm going to read that really quick because I think it really, it really kind of gets right down to it. It was, "Deep fake proliferation represents two clear dangers. Over the long term, deliberate development of false synthetic media will target US officials, institutions, and democratic processes with an enduring goal of subverting democracy and demoralizing the American constituency. In the near and short term, circulation of deep fakes may incite physical mobilizations under false pretenses, which you can think of as troops, initiate public safety crises, and spark the outbreak of violence. The recent spate of false..."

• 26:14Chris Benson?

  conspiracies proliferating via WhatsApp in India offer a relevant example of how bogus messages and media can fuel violence. The spread of deep fake capabilities will likely only increase the frequency and intensity of such violent outbreaks. Now, that was one paragraph, and that's a scary paragraph when you think about it.

• 26:34Daniel Whitenack?

  It is.

• 26:35Chris Benson?

  You know, that was just one out of the entire thing. And he goes on, and they do make recommendations, which we'll hit in a few minutes. But, you know, if that doesn't give you pause, when talking about this, I'm not sure what would.

• 26:45Daniel Whitenack?

  Another one that I that I saw in that Washington Post article, which which I thought was really great, was a quote from, or I don't know if it's a quote, it was a paraphrase from Rachel Thomas, who's the one of the co-founders at fastly.ai, which we all we all love for for many reasons, including the educational piece and the practical packaging and everything. So, shout out to fastly.ai. But she kind of said that disinformation campaign using deep fake videos probably would catch fire because of the reward structure of the modern web. So I think what she's getting at there is basically the the shock factor of a deep fake video is really what drives

• 27:28Daniel Whitenack?

  the reach of that video. And so these videos are kind of set up to be shocking in in many cases. And so just that by itself lends to those kind of going viral, reaching bigger audiences versus kind of maybe more mundane, but but true videos. Right? So it's it is a concern. And, you know, I know that we've already talked about kind of the just the idea of these videos existing as as dangerous. But also, I mean, they are already being used by malicious malicious actors, like you were saying, in various places around the world. And I think I think, you know, the Russia piece also fit into the hearings that were this morning, right?

• 28:14Chris Benson?

  They did. So they addressed, and so I know that in politics, some people may disagree, but you know, the American FBI has stated explicitly that Russia interfered with the election of in 2016. And so taking that as a basis of fact for this, it was also noted in that same testimony, "Moving forward, I'd estimate Russia as an enduring purveyor of disinformation is and will continue to pursue the acquisition of synthetic media capabilities and employ the outputs against its adversaries around the world." And basically, I think that's representative, this is me speaking now, that's representative of the

• 28:53Chris Benson?

  fact that it's a weapon of information warfare at this point, deep fakes, or it can be used in that way. And that does, and it's not necessarily just Russia, it can be many, many, you know, nations that can, that can try to influence and sway other countries, other parts of society with that. And so these are the types of things that, it's not just you and I talking about it, it's not just the AI community or the general population, it's certainly something that the defense industry and the military at large are having to consider. It's still relatively new, and it's something that, that really all countries are going to have to contend with going forward.

• 29:33Daniel Whitenack?

  Yep. So I guess maybe one last thing here when we're moving on from the dangers, and maybe a quick point here, that I know that you posed on our LinkedIn page. So if you aren't aware, our podcast has a couple of ways for you to engage with us. We'd love for you to engage on our Slack channel, which you can join if you go to changelog.com/community. We also have a LinkedIn page. I think it was posed on the LinkedIn page, if there were any beneficial uses of deep fakes, or good use, good use cases for using this sort of technology to create fake somethings. So we would love to hear from you if you have those ideas, but I know a couple came up. What did you see there, Chris?

• 30:17Chris Benson?

  So I know that, and I hope I don't mess up his name, Konstantin Svetnov. I'm sorry, Konstantin, about mispronouncing there. He is in the Atlanta, he's part of the Atlanta Deep Learning Meetup community as well. But he is a Senior Solution Architect with Nvidia. And, you know, really savvy guy about this, but he did point out one of the things. He said, he talked about kind of what you said earlier about, you know, the technology itself is agnostic. It's not a bad technology. It's a, it's a set of tools that can be used. And we've, we've talked about some of the joking things. And obviously can be used for bad, as we've been addressing as well. But he pointed out that, you know, the forensic, you know, we can learn a lot.

• 30:58Chris Benson?

  When we do have bad actors doing that, the forensic evidence that we can then analyze and understand how people are doing that is beneficial. And so he's kind of saying, there's, there's something where you can take something good after something bad has happened and improve. And then he finishes up, he says, and also, deep fakes could create a whole new genre of TV comedy. And there's that to be said, certainly. So maybe, maybe there's some, some pretty fun things that could be done lighthearted, which I know isn't, isn't the tone we've set here. But it'll be interesting to see how people use these technologies going forward. It's here, you know, everyone's going to have access to it at large. And so I would certainly love to see that optimism express itself in people's creativity.

• 31:41Daniel Whitenack?

  Yeah, and one, one thing that, you know, is, is useful like this in terms of the kind of entertainment industry. There's obvious use cases where studios, movie studios have people's permission to create sort of these computer generated things. Like, you know, maybe someone can't dance a certain way or something, but that needs to be in a movie. And they get that person's permission to kind of make that video of them dancing or whatever, whatever the situation is. I think there are, you know, legitimate uses of that within entertainment. But also in addition to governments kind of weaponizing this sort of technology or malicious actors kind of using it against, you know, their enemies.

• 32:26Daniel Whitenack?

  I think there are probably uses of this technology that could be beneficial in kind of the, the opposite way. In terms of bringing humanitarian or or or help to people in in the developing world as well, where the political situation is is hard. I know that, you know, getting educational material to people in certain areas is really tough to certain language minorities as well, especially if those language minorities are also religious minorities. And so, you know, sometimes, like, in my mind, I'm aware of, you know, translators who might translate like educational material or something for those people. They probably don't want to have themselves on a video. So if that video was kind of created as an avatar or something like that,

• 33:10Daniel Whitenack?

  then, you know, a lot of that could be useful as well. So I think that putting all of the weight on the bad uses is not completely fair, although I think putting a lot of the focus there is warranted, because this is a very serious topic.

• 33:24Speaker 00

  I agree with you. I'm really hoping to see some great use cases come forward where it's not just the doom and gloom thing going forward. So I'll tell you what, if anybody wants to put in my face, which is out there on a dancing video or something, my seven-year-old daughter would love that. So I'm hoping somebody will post something like that on Slack or in their LinkedIn community or something.

• 33:51Speaker 00

  Well, hello there, listeners of Practically I. How are you? This is Adam Stokoviac. If you haven't heard yet, we're launching a new show called Brain Science. It's a podcast for the curious. Are you curious? Because if so, we're exploring the inner workings of the human brain to understand things like behavior change, habit formation, mental health, and what it means to be human. It's brain science applied, not just how does the brain work, but how do we apply what we know about the brain that can transform our lives? Learn more about the show and subscribe at changelog.com/brainscience. Until then, here's a preview of episode one, where we talk about the fundamentals of being human.

• 34:31Speaker 06

  We're also all designed to be in relationship. We are fundamentally hardwired to have social groups and and this sense of attachment. And because I'm sort of a geek when it comes to research, what researchers have found is that attachment, which that's what we label, how we relate and connect with others, attachment is 100% learned.

• 34:57Speaker 06

  Which means our genetics don't actually contribute to how we learn to stay in proximity with other people. And with that, that we all develop ways to manage the threat of the loss of a relationship. But nobody gets to opt out of going, I need to be in relationship with others.

• 35:17Speaker 00

  It's almost like we need to have that echo from another human being to let us know that we

• 35:21Speaker 06

  Yeah.

• 35:22Speaker 00

  were there, or we're alive, or just some sort of feedback loop. I'm not really sure how to describe that.

• 35:27Speaker 06

  Well, it really is this sense of being with, right? Like, I can't fight battles on my friends' behalfs, or on my kids' behalfs, right? But the simple fact that I know of what's going on makes a difference. Because I would contend it's sort of like, I help them hold that weight emotionally. And so that actually leads me into the third thing. And the third thing that I would say in regards to the fundamentals of being human is that we all struggle.

• 35:57Speaker 00

  Oh, yes.

• 35:58Speaker 06

  Right?

• 35:59Speaker 00

  Big time.

• 36:00Speaker 06

  And that, you know, we don't always get to pick the way in which we struggle, but we all struggle.

• 36:07Speaker 00

  Well, if you like what you hear, you should go to changelog.com/brainscience. The show is not out yet, so don't get too excited. But you can subscribe and be notified as soon as the show launches. Once again, changelog.com/brainscience.

• 36:39Daniel Whitenack?

  All right. So we've talked about, you know, what deep fakes are. We've talked about the dangers they pose, and also maybe some benefits that they can offer. But maybe now let's kind of move to talking about how people are thinking about protecting themselves or other people or societies against deep fakes, or the disinformation that they can spread. So I know one of those approaches that I've seen in the community to protect against deep fakes is kind of a strategy that OpenAI has taken with their GPT-2 model. So we have a podcast episode about that model and the technical details of it. If you're interested,

• 37:21Daniel Whitenack?

  definitely take a listen to that episode. But the thing that they saw with GPT-2 was that it is capable of generating this sort of very realistic text and very long-form text, which obviously they saw as an opportunity to create fake news articles or fake content for social media and that sort of thing. And so they saw the danger with this. And the approach that they took to kind of prevent that was they just released the code for the model, not the pre-trained, the full pre-trained model itself. They released kind of a limited version of the model. And they didn't release the full data set that they used to train that

• 38:05Daniel Whitenack?

  model. So their hope, I think, was basically to try and slow down the malicious use of that model and kind of give researchers time to develop kind of detection methods or methods that would help in some other way fight fake news. And now that they know that GPT-2 is out there. So it's an interesting approach. I don't know if you have any thoughts on that, Chris. I don't know if it did. I guess my question would be, how much good did that approach do, given I know just recently I saw that a student who had access to certain compute resources, to TPUs, I don't, I forget if he used the cloud or what, but he had access

• 38:50Daniel Whitenack?

  to some compute resources, which are not that uncommon. I think other people could get access to those. So and he was a student and he was able to use that code and reproduce the full GPT-2 model. And that kind of was less than four months, so three to four months after OpenAI released the code and the paper and all of that. So it's not that much time between when they released the kind of partial release and when the full thing was public. And the student student released that. So the question is, I don't know that three or four months really buys us much. But that's not to say that it wasn't a good approach or OpenAI didn't try, but I just wonder if that's enough.

• 39:33Chris Benson?

  Yeah, it's a tough thing. I know when we when we did our GPT two episode shortly after its initial release there. And we we talked about this in that episode and kind of debated that approach from OpenAI in terms of the, you know, this release. And we contrasted that against, you know, kind of the norm in open source of kind of throw it out there and let the world dig in and see what your stuff is. And and one of the things that we considered at the time was was, you know, maybe this gave us a little bit of bumper, even though it wasn't purely in the spirit of open source in that way. And I think in the time, one of the things I said then was I thought probably a couple of things that it was probably good to give us a little bit of time just to absorb and and realize the new world that we're in with that kind of release.

• 40:18Chris Benson?

  And as and I also said that really it would happen anyway, you know, that now that people knew what was possible, it would be recreated sooner rather than later. And this student has done just that. They came in and did exactly what I was predicting. And the reason is there's a lot of smart people in the world. And if just because one team does something and doesn't release it, doesn't mean everyone else now knows it's possible. And so you put smart people on the problem and they they know there's a solution then. And so they're going to get it. And I think as I've had a little bit of time to analyze this, I think OpenAI's approach was the responsible way to do it at this point. It wasn't too long. I mean, we saw that it's gotten out there anyway, but it gave us time to to absorb what they had achieved,

• 41:01Chris Benson?

  what was possible with that achievement and how we might think about, you know, malicious uses of it, which we started doing immediately. And then this kid came out and and released this, the student. And so I think for significant impacting technology releases, this might be the way forward, at least in my opinion.

• 41:17Daniel Whitenack?

  Yeah, given that the time period, so there may be a delay. Hopefully organizations kind of approach things in a responsible way and give people notice when something like this is coming out and own up to the implications of it and kind of expose those. But even then, like you say, the the full technology is going to be available and people are going to reproduce it. So in light of that, there's definitely people out there that are focusing on detecting fakes. So they're focusing on actual AI methods that would be able to detect or discriminate fake versus non fake. So if you just go, so I really like the website papers with code.com.

• 42:00Daniel Whitenack?

  If you haven't seen that, you should definitely check it out. But if you just search for, I just search for fake and detection. And then a bunch of papers come up, kind of recent papers on this topic. And I would say that this kind of detecting fake, deep fakes or fake news or fake images, fake videos. However, they phrase it in their papers, there's definitely some approaches out there that seem promising. But there's definitely a no one size fits all solution. So there's no, you know, this is how we're going to protect ourselves against deep fakes things. There's some solutions out there for the, on the video side, there's, there's people looking at kind of

• 42:40Daniel Whitenack?

  the way people blink in videos, actually, and apparently that's hard to replicate in a fake. There's also kind of a person's kind of facial fingerprint and also like per frame inconsistencies in videos. So there's people looking at those sorts of things to be able to tell like these, these artifacts that would give away a deep fake. On the tech side, there's people that are analyzing kind of the persuasive structure of news articles or arguments and the stances that those articles take to kind of figure out if they're fake or not. So there's a whole variety of things that people are trying and I I've certainly have only just mentioned a few. There's, there's a whole bunch out there.

• 43:22Daniel Whitenack?

  We'll link some of those in our, in our show notes, but I think probably the best thing, if you're interested in those sorts of tech. Techniques is to to search some some site like papers with code and and look at what people are doing. One of the things that that I was struck by in reading that Washington Post article was was a quote by a computer science professor at Berkeley and Dr. Fareed and he said, we are outgunned. And the reason why he said that is the number of people working on video synthesis as opposed to the detector side is 100 to one. In other words, there's 100 people working on interesting deep fake technology and only one of the one corresponding person working on detecting deep fakes.

• 44:05Daniel Whitenack?

  And that's probably also due to kind of, in my opinion, the incentives that are built into the academic system when and, you know, the the AI community. Whereas if you come out with a a deep fake technology or some video technology for generating videos, you're going to get a lot more attention than if you come out with a really great detection thing, right?

• 44:25Chris Benson?

  That's true.

• 44:26Daniel Whitenack?

  Um, so so I think the moral of that story is, you know, if you're out there and you want to work on something around this, maybe consider working on the the detection problem. We really, we really need people working on that.

• 44:37Chris Benson?

  Yeah, you know, not only having to do with deep fakes, but having to do with things like poison data and other safety issues. We actually had an episode I was wanting to call out to our listeners. Episode number 33 was called staving off disaster through AI safety research. And that was one when I was attending Applied Machine Learning Days in Switzerland. And I know you organized the the AI for good track for that conference that I met with, and I'm I'm going to try to get his name right. El Madhi El Hamdi, and I apologize if I just screwed that up because I know he listens to the podcast. And it was a fascinating conversation that we had that was recorded and and put out where he kind of talked about different approaches to that.

• 45:19Chris Benson?

  And he basically made that same point is that the number of bad actors, nefarious actors out there far, far outweighs the number of people that are trying to be pro safety in this space, he being one of those people.

• 45:33Daniel Whitenack?

  So there's the kind of release of models side of things. There's the detection of deep fakes. There's probably a third category here, which really is assuming that deep fakes are going to exist.

• 45:49Daniel Whitenack?

  That we're not going to be able to detect them all. So some are going to get through our best detection mechanisms. We should be able to criminally prosecute people that are generating these things in a malicious manner. That's what that's the stance that some people are taking. So there's been bills introduced in the United States. I'm sure there's there's other governments wrestling with this, but this is kind of probably what we're most familiar with is here in the US. Like Senator from from Nebraska, you know, other other lawmakers from Virginia and California are considering legislation around these things even at a state level. So the New York State Assembly

• 46:28Daniel Whitenack?

  has introduced bills to kind of push back against against this sort of technology. So I guess that's another approach that people are taking. I think there's still a lot of really, you know, interesting, open questions there, like, how how is this enforced? How would we enforce this while still kind of allowing, you know, legitimate entertainment companies or companies that are maybe doing legitimate work in these areas and, you know, helpful things in these areas? How do we allow those people to operate and and yet, you know, prevent this malicious malicious usage of the technology? So there's a lot of interesting, interesting questions there, you

• 47:09Daniel Whitenack?

  know, where do we draw the line of deep fake and not and malicious and not? And, you know, there's a whole range of things from, you know, jokes and satire to the really harmful, bad stuff, quote unquote. So where do we draw the line? There's there's a lot of open questions there.

• 47:25Chris Benson?

  Yeah, when I was watching the the Intelligence Committee hearing that we've been alluding to in this episode this morning, that was a big issue because a great deal of this involves, you know, First Amendment rights. And for those who are not US citizens, the First Amendment of our Bill of Rights, which is part of our Constitution, is the amendment that allows for free speech and free expression in the United States. And so that was there was quite a bit of kind of legal oriented back and forth. Much of and I'm not a lawyer, so much of that went right over my head. And so, you know, I don't, I won't speak to that directly, but they really were talking about how do you handle this without violating First Amendment protections?

• 48:08Chris Benson?

  So, some good news is that this morning in the House Intelligence Committee hearing that we've been talking about over the course of this episode, they, they did in fact make some recommendations on how to contend with deep fake issues. And there were six explicit points that were called out, and I thought I'd just kind of, not cover all the verbiage on each one, but just kind of the first sentence or so of each one, which kind of gives you the, the sense of it. The first was, "Congress should implement legislation prohibiting US officials, elected representatives, and agencies from creating and distributing false and manipulated content." And as an addendum to that, they mentioned that the US government's, you know, whatever

• 48:51Chris Benson?

  kinds of statements they make should always be the truth. The official government's statements and policies should always be based in truth. The second thing was that policymakers should work jointly with social media companies to develop standards for content accountability. The third was that the US government should partner with the private sector to implement digital verification signatures designating the date, time, and physical origination of content. The fourth one was that social media companies should enhance their labeling of synthetic content across platforms and work as an industry to codify how and when manipulated or fake content should be appropriately marked. The fifth was that the US government, from a national security perspective, should maintain

• 49:34Chris Benson?

  intelligence on adversaries capable of deploying deep fake content or the proxies that they employ to conduct such disinformation. And the final one they noted was, "Public awareness of deep fakes and its signatures will be greatly assisted in tamping down the attempts to subvert US democracy and incite violence." So, those were good, and I think the, the Intelligence Committee heard that. There was some debate about the legal issues regarding First Amendment concerns, but it was good, you know, to see them wrapping up with potential ways forward to mitigate the dangers that we've been talking about on this episode.

• 50:09Daniel Whitenack?

  I see a couple of things in there. I guess one thing is there's definitely an emphasis on social media companies, both to contribute to the discussion, but also to utilize their resources to help label synthetic content. So there's definitely a responsibility placed on social media companies. It will be interesting definitely to see how social media companies receive that and the cooperation that they get. I'm not totally convinced it will be, you know, exactly. So there's, there's competing interests here always, right?

• 50:43Chris Benson?

  Of course.

• 50:44Daniel Whitenack?

  So, uh, so it'll be interesting to see how that pans out. The other thing is, um, you know, maintaining intelligence on, on adversaries capable of deploying deep fake content. I think that includes me. I don't want to give myself away, but you know, I'm capable of deploying it. I guess I'm not an adversary of the US government, at least I don't consider myself to be so. But, uh, it is interesting to me. That, that's like everyone who knows how to like clone a repo on GitHub and run that code. I don't know. I guess probably what they're getting at is, is, uh, government entities and, and other things.

• 51:21Chris Benson?

  That's what I think.

• 51:22Daniel Whitenack?

  But it seems like there's a, there's a huge range of things. I guess I don't want the government maintaining intelligence on me since I'm capable of deploying those technologies, I guess is what I'm saying.

• 51:32Chris Benson?

  Yeah, I think the intent of the recommendation was, was really, uh, from the perspective of the United States, uh, foreign adversaries is what that was looking at.

• 51:41Daniel Whitenack?

  Makes sense.

• 51:42Chris Benson?

  That was certainly the impression I came away from listening to the testimony.

• 51:45Daniel Whitenack?

  Yep. So I guess, uh, maybe one way we could wrap up this discussion is to kind of give you some good learning resources and links on this subject. Not so much on the learn how to make, uh, deep fakes, cause you probably also don't want to have, uh, uh, the US government maintaining intelligence on you, but, uh, but maybe just kind of learning about deep fakes, the state of the art and also, uh, you know, detection methods and other things. I think one, one good link, if you, if you just go back one fully connected episode to when we were talking about GANs and reinforcement, learning and transfer, certainly GANs and transfer learning come into play in, in this discussion in a, in a big way.

• 52:29Daniel Whitenack?

  There's also a good over, that good overview article that I mentioned from the Washington Post that we'll link, that's not really technical, but. Does provide a good amount of links in it as well. And then there were a couple of workshops that I found. Actually, right now ICML is going on. There's a workshop there that I'm assuming will post kind of papers and maybe some results and discussions about synthetic realities, deep learning for detecting audio visual fakes. There's also a workshop that happened at Applied Machine Learning Days this year about fake news detection and we'll link that. There's actually a GitHub repo that kind of goes through a tutorial and some ideas and

• 53:14Daniel Whitenack?

  slides about fake news detection. So I think those would be good starting places. But yeah, I appreciated the discussion today, Chris. I think it was good to talk through some of these things. It was difficult in many ways. I like to be generally positive. I hope that comes through in the podcast, but.

• 53:33Chris Benson?

  I think we both do, actually.

• 53:35Daniel Whitenack?

  It's a pretty heavy topic.

• 53:36Chris Benson?

  Yeah, I think we have a responsibility to our listeners and in general, just to be able to fairly represent things. And obviously, a lot of the things that we talk about are just exciting and fun. And I think, I hope that we come off that way to our listeners. But there's occasionally some scary stuff in this field and I think it's our responsibility to represent that as well. So I hope our listeners feel a little bit more attuned to this topic and sorry for the downer of an episode and appreciate very much everybody sticking in with us through this. It was, it was a good talk, Daniel.

• 54:08Daniel Whitenack?

  Yep, definitely. We'll talk to you next time, Chris.

• 54:09Chris Benson?

  Talk to you next time. Thanks a lot.

• 54:13Speaker 00

  All right. Thank you for tuning into this episode of Practical AI. If you enjoyed the show, do us a favor, go on iTunes, give us a rating, go in your podcast app and favorite it. If you are on Twitter or a social network, share a link with a friend, whatever you gotta do, share the show with a friend if you enjoyed it. And bandwidth for changelog is provided by Fastly. Learn more at fastly.com and we catch our airs before users do here at changelog because of Rollbar. Check them out at rollbar.com/changelog and we're hosted on Leno cloud servers. Head to leno.com/changelog, check them out, support this show. This episode is hosted by Daniel White. And Chris Benson. The music is by Breakmaster Cylinder. And you can find more shows just like this at changelog.com. When you go there, pop in your email address.

• 54:56Speaker 00

  Get our weekly email, keeping you up to date with the news and podcasts for developers in your inbox every single week. Thanks for tuning in. We'll see you next week.