AI is Causing Fake Video and Fake News Problems – But This App May Help
We all know the old phrase “seeing is believing.” We’re hard-wired to believe the things that we can see with our own eyes. But with advances in artificial intelligence (AI) technology, the line is starting to get blurry. AI can write comprehensible text, create images, make deep fakes, and fabricate fake videos that look real. If seeing is no longer believing, how are we supposed to know what’s real?
See AI, Fake News, and Reviews with Peter Sisson for a complete transcript of the Easy Prey podcast episode.
Peter Sisson is Silicon Valley-based serial entrepreneur. He has an MBA from Tuck and a Master’s in Artificial Intelligence from Stanford. Everything he learned about AI at Stanford in the late 1980s, however, is no longer applicable. He has started four companies before his current one, all of which have been acquired in some form. He calls himself a product-focused entrepreneur, as he likes to solve problems with software.
Misinformation and polarization has exploded since Peter started in Silicon Valley over twenty years ago. His latest company, CertifyIt, offers an app to help combat fake news, fake reviews, and fake products.
The Rise of Fake News
When Peter talks about “fake news,” he’s not talking about news with spin or an agenda, or news that people disagree with. He is talking about literal fake content, where what’s being reported on the news never actually happened. This is the kind of problem we’re heading towards. Technology is so good that you can create photographic evidence of you doing anything. Legitimate news sources encountering fake video evidence and reporting it as genuine is a real risk.
These technologies are going to be so good that you basically can provide photographic evidence of you doing anything that you need to prove.Peter Sisson
Our brains are wired to believe what we see significantly more than what we hear. Seeing is believing, as they say. When AI can create an fake video or photo that shows you anything, it’s going to be hard to know what’s real. It may even become difficult to tell which articles are generated by AI writers and which ones are actually written by humans. It will be hard to sort through it and identify what’s genuine. This isn’t necessarily a bad thing, but it definitely could be.
The State of Content-Generating AI Technology
Producing written content through AI is going great. You can go to AI tool like GPT-3 or ChatGPT and tell it to write a story or article about anything, and it will generate that content. It will be comprehensible, and it will probably even be factually accurate.
What Peter is most concerned about is certified video content. Many of us have seen AI-generated images on social media. We can fake photos pretty realistically at this point. Faking videos is the logical next step – after all, a video is just a sequence of photos.
When we start seeing things that have been fabricated by computers and believe them to be real, that’s when reality stops being real and people start not knowing what to believe anymore.Peter Sisson
The technology is already working towards that goal. But not being able to tell if a video or article or anything else is genuine or is a fake video or article written by a computer has implications for society. Peter started CertifyIt as a way to guarantee what you see is real.
AI and Deepfake Videos
The technology to make fake video content has come a long way. But right now, AI is not great at generating videos. The deep fakes we have seen have been a public figure’s likeness edited over an original video. The best AI can do right now is static images. And if you look at those images, there are still clues that let you know it was AI-generated. You can check the leaves on trees, for example, or look at the details in humans or animals in the photo.
In five years, though, everything in tech will get better. Resolution will get higher and believability is going to get even better. In five years, we may not be able to easily tell an AI-generated fake video or image from a real one.
That’s not Peter’s concern right now, though. His concern is that with technology we have today, you can reanimate someone and make them say anything you want. The tools to use AI to create fake content are sophisticated and used to be very specialized, but they are now open to anybody. Peter can go to OpenAI, sign up, and get a number of AI-created images for free and even more for a small fee. He could get access to GPT-3 or ChatGPT to generate articles, research papers, stories, and more. And these AI tools will create images and text that are very convincing because they’re trained on the entire internet.
The Limitations of AI-Generated Content
The biggest limitation of AI-generated content is its training. AI learns from the content it’s trained on, and in a sense, it’s always the average of its input and training. An AI can’t invent things because it can only build from what’s already been done. However, a lot of great ideas are built out of prior ideas. There is talk that AI can’t truly be original, but it can definitely seem very original.
Another challenge of AI-generated content is erroneous data. An AI has access to a lot of facts and can organize them into in article in just a few seconds. But since it doesn’t actually “know” the material, it has no way of knowing if some of the facts it pulled into the article were incorrect. It judges purely based on the frequency of the idea online, so it can easily take a widely-believed error as fact.
One thing AI ethicists are struggling with is the question of what to train it on. When AI image generation first started, the images tended to skew sexual because there’s so much sexual content on the internet. The programmers had to adjust the training parameters to remove it. But humans are sexual beings too, so is it a good idea to remove that in training and bias AI against sexual content?
It’s a quasi-human entity that we don’t quite know how to handle yet.Peter Sisson
We’re still not fully sure how to deal with AI yet. It is having, or appears to have, intelligent conversations with us. It writes amazing articles and creates believable photos. That’s pretty powerful. There’s lots of ethicists involved in the debate, and lots of people are trying to figure it out.
When Photographic Evidence Isn’t Evidence
There’s a whole different conversation to be had about what it means when images no longer count as definitive evidence. If a photograph or video might be a fake video or image generated by an AI, what takes its place?
That’s what Peter is trying to solve with CertifyIt. It’s trying to fill the space originally filled by cameras for providing proof. It guarantees a video submitted is a genuine video, not a fake video or one that’s been tampered with. It can be used to help prevent fake products, fake reviews, and more.
CertifyIt could be especially helpful for news media. News media is struggling with the fact that a lot of source content is going to be consumer-generated mobile videos. But it’s a Pandora’s Box of what you get when you open the news up to consumer videos. CNN launched a program called iReport in 2006 where citizens could file video reports that might air on CNN. It worked great until people started uploading fake stories. One fake video fooled CNN and they ended up reporting Steve Jobs’ death when he hadn’t actually died. They had so many issues with fake content that they ended up shutting iReport down in 2015. With a tool like CertifyIt, they could know for sure all the video they receive are genuine and unedited.
How CertifyIt Prevents Fake Video
The CertifyIt app prevents fake video by acknowledging that the only way to know for sure a video is genuine and unmodified is to create it in the app and have CertifyIt, not the person who recorded it, control it going forward. There’s no folder or video file on your device that can be edited or combined to create a fake video. Think of it like a camera app, but when you record, it’s not going into your camera roll. It’s actually going onto the CertifyIt server and streaming from there. CertifyIt can also check the metadata. If you say you’re in one time zone but the system sees from your location that you’re in a different one, it will correct it.
Basically, it’s just a way for us to know that the video went straight from its creator to being hosted with no way to edit or change it.Peter Sisson
All the app does is record video, as well as clips you can edit into a story. It then forces you to pin your video to a map within 150 meters of where you are. When you’re done recording, it creates a link and opens a sharing window. You can text a link, post it, email it, put it on social media, or whatever you want. The app makes it easy to share videos that haven’t been and can’t be tampered with.
Preventing Fake Video Relies on Trust
In order for CertifyIt to function as intended, people have to be able to trust it. That’s why CertifyIt is set up as a public benefit corporation. A public benefit corporation is a company that’s committed to fundamental operating principles of fairness, truth, and data protection, among others. In a standard corporation, the company’s sole goal is to make money for shareholders. There could be legal ramifications if they chose upholding values over making a profit.
A public benefit corporation is a new form of legal entity. It gives all the benefits of a normal corporation, but also allows a company to be mission-driven. That mission gets put in the corporation’s bylaws, and they have to file a report every two years on their progress towards that mission. It’s kind of like a corporation with a conscience.
CertifyIt chose to incorporate as a public benefit corporation. They are now legally bound to work towards their social objectives, which are in their bylaws. CertifyIt’s objectives are to promote the spread of accurate, truthful information and fight fakes online.
The Benefits of CertifyIt
There are a lot of places where preventing fake video content and other fake content can be highly beneficial. Peter thinks news media especially would benefit. They want to be the first to break the story, but still need to make sure that story is accurate. CertifyIt removes any uncertainty about the origin or genuineness of a video. If people use it to submit to news media, they won’t receive fake video and can break news faster.
Peter also thinks travel sites like booking.com, TripAdvisor, and Travelocity could use CertifyIt as well. They have all added a level of “certified” reviews because fake reviews are such a problem. A level of review verified by CertifyIt can be helpful. By requiring a CertifyIt video before posting a review, they can verify the person was actually at the location since CertifyIt doesn’t allow video uploads without location data. Uploading a CertifyIt video could be a verification step in the review process. Or the sites could move to support video reviews.
Online shopping could use CertifyIt as well. On eBay, for example, there could be a CertifyIt button. If you’re not sure about a product, you can use the button to request the seller upload a CertifyIt video. The video could verify that they have the product in stock and it works as promised.
It could even be used for short-term rental sites like Airbnb to verify what a place looks like. Peter went to a hotel in Italy that looked nothing like its listing online. If you have the option of only going to places with CertifyIt videos available, you would know exactly what to expect.
Fighting a Problem That’s Just Starting to Happen
We live in a world where sometimes traditional media channels aren’t even involved in our news. Nobody is fact-checking, it’s just Bob filming something and uploading it to social media. Either you trust Bob and share his stuff, or you don’t. But we can also make fake video content that convincingly shows something that never happened or someone doing something they never did. It can trick our brains easily because if we see it, we tend to believe it.
We’ve got to stop believing what we see.Peter Sisson
We haven’t seen many fake videos with real negative consequences yet. We look at deep fakes and other AI-generated fake video because they’re interesting, not because they’re malicious. And right now, they’re fairly easy to spot. Eventually, people are going to be fooled and scammed by AI-generated videos, images, and articles. And when that happens, they’ll start looking for a way to be sure video is real.
CertifyIt solves that problem. People will be able say, “I saw a video, but that could be a fake video. I won’t believe it’s real unless it’s a CertifyIt video.” If you really need a video to prove something, you can’t just ask for a normal video. When you’re dealing with a situation where trust is low and you need to verify something remotely, CertifyIt is a handy tool.
- Easy Prey Podcast
- General Topics
- Home Computing
- IP Addresses
- Online Privacy
- Online Safety
Online dating can be a fun way to pass the time. It can also be a way…[Read More]
Many of us are aware that scams, fraud, and cyber threats are out there. We may even…[Read More]
We all know the old phrase “seeing is believing.” We’re hard-wired to believe the things that we…[Read More]
It’s been six years since the infamous Ashley Madison hack, the one that published the information of…[Read More]