A lot has already been said about the various potential dangers of AI, even if the worst of it may still be quite a bit into the future. AI may be powerful and game-changing, though something like a rogue machine takeover – with the current technology, at least – looks to be quite unlikely.
One application of AI, however, has advanced by a worrying amount in recent years. Deepfakes are essentially AI-generated fake animations or videos that could be edited to do anything, and by that we do mean anything. You can make a fake video of a celebrity doing something specific you want, or a nuclear-armed world leader issuing a threat to another, or literally anything else really. The possibilities are endless, which is why it’s so dangerous. The worst part? At this point, the tech is already so good that it’s impossible to tell if it’s real without some elaborate forensics.
Deepfakes have taken cybercrime cells of countries across the world by storm, owing to its various applications in crimes like revenge porn, identity theft, political coups etc. The concerns are well-founded, too, as deepfakes – and their integration with other AI applications – have come a long way in a matter of barely five years.
10. It’s A Whole New Type of Crime
We don’t doubt that deepfakes would definitely be used for some good in the future, as most groundbreaking technologies are. What’s concerning, however, is the sheer number of its applications in crime.
Take its use in revenge porn as an example. According to the current laws in most countries around the world, the crime of uploading a photo or video of someone else without permission has a clear line of accountability, which – if found – could be traced back to the culprit. That’s not the case with deepfakes, as you don’t need anything more than a photograph to make them. Forget traceability; under our current laws, these crimes are not even properly categorized, which poses some tough questions about how to deal with it.
To take the above-mentioned example, we’re not sure what type of crime that would be. Is it voyeurism? There’s arguably no actual invasion of a real person’s privacy without their consent, as the fake image or video isn’t real for all legal purposes. Under the current system, the only problem here is taking content from someone’s social media account that’s already available in the public domain, even if that doesn’t even come close to accounting for the seriousness of the crime.
9. You Can Now Deepfake A Whole New Person
Even for the most skilled artist out there, it’s impossible to draw a face that doesn’t exist. All of our art is mere imitation of the world we see around us, and creating something truly alien or non-existent is not possible for almost every one of us. That’s where computers come in, as they don’t have the same barriers on their imagination.
Art generated by AI has gone viral quite a few times in the recent past, and stands out because of just how different it is from anything we know of. While AI art was just restricted to illustrations and undecipherable visuals till now, deepfakes have taken it one step further. We can now generate entirely new faces with the help of AI and deepfakes, and – worryingly – it’s impossible to tell that they don’t exist at all. The truly scary part, however, is because of machine learning, deepfakes can do all that without any human supervision at all.
8. They’re Already Triggering A New, Global Arms Race
New technologies are often viewed with suspicion, which is usually what triggers arms races. It’s the same with AI – and deepfakes in particular – as almost every major nation is currently working on ways to perfect the technology. There’s one key difference between deepfakes and any other military technology of the past, though, we don’t have a clear view of what we’re building towards.
At best, we know what we’re developing, which is a technology that can fake anything to a level that we can’t detect. Its number of applications, however, are mind-bogglingly vast and still quite unclear.
What is clear, however, is that a global deepfakes arms race is already underway, and it’s not in the distant future, either. Deepfakes have already started influencing politics in the present world, and have been used in quite a few political campaigns and coups across the world in the past few years.
7. The Western World Isn’t Really At Risk
Enough has been said about the role of deepfakes in influencing the outcome of elections, which also happens to one of its most dangerous applications. Imagine a technology so advanced that anyone could be made to do anything on photographic record. You can make a politician deliver any fake speech you want, or make a fake state leader declare a military emergency, all by tweaking some things on a computer. The possibilities are endless, though the most devastating consequences of this tech won’t be felt by the developed world. While it’s true that the use of deepfakes among technologically-savvy, Western populations to influence elections remains an ever-present threat, these countries still have many checks to not let it get out of control.
The global south, on the other hand, is much more impoverished and susceptible to manufactured campaigns aimed at sowing unrest with the help of deepfakes. Ethnic tensions still run high in many countries across Asia, Africa and the Middle East, and it’s almost a certainty that deepfakes would be used to further entrench those lines.
6. They Discredit Actual, Admissible Evidence
While deepfakes could be used for a lot of nefarious purposes, its impact on the collective human psyche would be deeper than just a higher rate of crime. As a society, we rely on photos and videos to be accurate representations of reality, as everything from political candidacy to crime-scene investigations assume that things in photos and videos are how they are in reality. Even with our best fakes, it’s possible – and often quite easy – to tell what’s real and what’s not.
With deepfakes, that would no longer be the case. In a world where videos and photos could be made to do anything, it won’t make sense to rely on visuals for anything. In fact, photographs and videos may cease to be mediums to capture human moments altogether, as an AI would probably be able to ‘make’ better photographs than you’d ever be able to capture; it’d be a fundamental and permanent shift in how we interact with the world around us.
5. They Disproportionately Affect Women
There’s no doubt that deepfakes would cause many new and unforeseen problems in global affairs. That’s really why they’re so good already, as election campaigns are some of the biggest employers of technologies like these.
Some of its most severe consequences, however, won’t be felt by a country or global leader, but everyday women. Deepfakes are already being used to make the kind of revenge porn that was never possible before: one that doesn’t exist. As long as you have any photograph of them, you can now make porn videos of pretty much anyone you want. It could be an unsuspecting celebrity, an ex, or really just any woman off the Internet.
4. One Of The Biggest Dangers To Global Peace
Quite a few things figure on the list of the biggest dangers to world peace, like terrorism, global famine, nuclear weapons, bio-terrorism etc. While all of them are undoubtedly quite dangerous, AI is fast emerging as a viable – and possibly even the deadliest – alternative to all of them.
Many scientists and other experts have now issued warnings against continuing research in Artificial Intelligence. For good reason, too, as it has already come far enough for us to be alarmed. AI of 2020 can repair itself, go through the Internet and learn things on its own like a student, and write AI software of its own.
Combine that with deepfakes, and there’s no dearth of things that could go wrong, and all of that is already happening. Take the example of the recent elections in India, where deepfakes were used to their full potential for the first time. Or that of Gabon – a tiny country in Africa – where the fear of deepfakes has been used to initiate a coup. Deepfakes are destabilizing the world already, though it’s inherently impossible to stop them because:
3. They’re Designed To Get Better On Their Own
Most people think that the true dangers of deepfakes and other AI applications lie in what they can do. They’re already making news for meddling in elections around the world, and we’ve all seen enough cyberpunk movies to know where all this ends.
As many experts would tell you, though, deepfakes and AI aren’t just scary for what they can do, but how they do it. Machine learning is one of the most groundbreaking inventions of our times, and its this ability to use its own data to correct itself and observe the world around it to learn that makes deepfakes and AI so dangerous.
A Hollywood AI would simply go rogue, arm itself and try to kill everyone. A real AI, on the other hand, would be able to run all the outcomes in its head before taking any decision. If replacing a political leader serves its purposes better than an outright armed-takeover, that’s the path it would take, all the while getting even better at choosing its future paths.
That’s what’s truly dangerous about machine learning, and because of it, AI and deepfakes of the future would be foolproof and almost impossible to best, if they’re not already.
2. Moving Photographs
If you remember the Harry Potter universe, you’d remember that all the photographs in the world are consistently in motion. For those not familiar with the universe, it’s nothing complicated; it’s just magic. Mind you, they’re not analogous to animated photo frames or any other type of videos in the real world. Photographs in the magical world of Harry Potter have personalities of their own, and even interact with other living characters.
While something exactly like that would still take a bit more time to develop, deepfakes may have learned how to make the images move. Developers at Samsung have developed a technology using AI and deepfakes that can make any face in an image move, with a surprising – and scary – amount of control over what you can make it do.
1. We Have No Ways To Counter It
With all the alarming reports, you’d think that our cyber security infrastructure would be working overtime to counter them, as deepfakes are a serious threat to everyone. The problem, however, is that we don’t know how to proceed at all.
Currently, we simply have no way to weed out – or even accurately detect without extensive forensics – deepfake images or videos off the Internet. We have no software or algorithm to accurately tell if something was generated by an AI or good old Photoshop. It’s a classic case of technology advancing faster than it could be regulated.
Even if we did have an accurate way to detect deepfakes, it’s even more difficult to establish intent. How do you separate an edited meme making a harmless joke with a deliberately-designed online campaign by a foreign power meant to destabilize the country? You may have a clear answer to that, but there’s a good chance it won’t hold in a court of law. Even if we figured that one out, how do we hold someone accountable for a nameless, machine-generated video on the internet?
Because of our difficulty in categorizing and stopping crimes related to deepfakes, there have been few or no checks on what it can do. Deepfakes of today are dangerously advanced and able to do things we couldn’t even two years ago, while we still figure out how to stop them.