How would you like to see Tom Cruise as Iron Man replacing Robert Downey Jr (RDJ)? Imagine the scene, Jarvis is taking orders from ‘Ethan Hunt’ instead of ‘Sherlock Holmes.’ If you are a fan of Mr Cruise, you might like the idea very much.
But you know the reality, you are bound to watch RDJ flying in his Iron suit. Even Doctor Strange might not be able to change this. Wait, hold on a minute, this all might be possible, in a way. Yes, you can enjoy Tom Cruise in a red iron shield blasting beams from his hands.
We actually have a better ‘wizard’ than Doc Strange to change this for you. This real-life ‘wizard’ is called deepfake technology. Actually, there are already videos on the internet that turn Tom Cruise into the richest Avenger. Just go check it out, but before that, let me present you with the curious case of deepfakes.
What is a deepfake?
In general, deepfakes are information, such as videos,that are not real, never happened in real life, yet that look real. In the age of modern technology and artificial intelligence, reality can often be deceiving, confusing, disappointing. You might see things that you would never believe or expect to happen.
Deepfakes are just another relatively new addition to this world of being-fake, more like the next-generation photoshop stuff. Technically, deepfakes refer to the fake videos, audios, photos that are mischief works of artificial intelligence. The ‘deep’ part comes from Deep Learning, a branch of AI.
There is a growing branch of software to make deepfake contents. The most prominent and renowned example of deepfake technology is deceiving videos of celebrities, politicians speaking and doing things that they would never normally say or do. In this case, using deepfake technologies, faces can be swapped with different persons, the voices are doctored in a technical way that are as similar to the real ones.
How are they made?
Deepfakes can be a serious and complicated process, but an amateur tech-geek is not enabled, empowered to produce deepfake contents. And the process is becoming more widely available, automated, and convincing. Most of the deepfakes rely on generative adversarial networks, also known as GAN.
GAN create two machine learning models that work on generating fake images and videos based on the real person’s authentic data. By encoding and decoding those data, the AI algorithms render the final result.
The dilemma of deepfakes
The aforementioned example of Tom Cruise being Tony Stark brings no harm to our society. It’s just an excellent way to get entertained by virtue of modern technology. But there is the other side of the coin too.
Just imagine, while stumbling upon social media, you get to see the newly-elect US president Joe Biden discussing imposing a new ban on US visas to all kind of Middle East citizens. Well, if you are a person of political intellect, you might not want to believe this video at first glance, think about cross-checking, but there are many people, a lot of audiences out there who would take the video for granted without thinking further.
Or, another video pops-up on your social feed in which Greta Thunberg praises a famous consumer brand that has a lot to do with world carbon emission. You might feel like ‘Et tu, Greta?’
While the contents of these videos may have nothing to do with reality, these can wreak havoc in mass-society. People would go wild; things would be unstable. One US senator considered deepfakes as the modern equivalent of nuclear weapons.
In the old days, he said some actor would need heavy military power to threaten the United States, but now, with the help of deepfakes, one could throw the country into tremendous crisis. It is debatable if deepfakes are really equivalent to nuclear weapons, but it’s a big deal when it comes to the individual threat.
For instance, making pornographic contents of celebrities, as well as ordinary people, is another threatening possibility of deepfakes. There are examples of swapping body parts such as heads of celebrities with porn-stars. In 2017, according to the Guardian, some Reddit users created porn-clips of Gal Gadot, Taylor Swift and Scarlett Johansson with the help of deepfake technologies. One only needs to briefly consider the vast potential harm to ones reputation, interpersonal relationships, the loss in potential value of ones artistic or financial worth, to realize the possible harms that can ensue.
The US election and deepfakes
Deepfakes were considered to be a devastating artifact of the US election in 2020. Now that people have seen how fake news (i.e. falsehoods presented in the guise of truth or parody, as in The Onion magazine) and cyberwars can interfere with the system in the previous election, deepfakes were thought to have a devastating effect on the 2020 US election.
But fortunately, this has not turned into serious threats. Of course, there were deepfake videos around, but they were not influential enough to be noticed. Yet, the whole election 2020 went dramatic and confrontational, thanks to, in some cases, Donald Trump and his loyal followers. This could be an understandable reason that deepfakes were an undesirable factor in out society, law, electoral system, politics etc.
How to spot one?
You can spot deepfake content if you look at it with a skeptical eye. Most of the deepfakes that arise on the internet are works of non-professional people. Deepfake videos are generally posted in low resolutions to hide the glitches inside them.
Most often, the lipsynching doesn’t sync precisely. The edge of the face can be seen breaking down from time to time. The eyes might not blink, although the latest products may not have this glitch . And if you look at the mouth movement while the character is talking, you might notice abnormal movements of the jaws.
Of course, the more professional deepfakes are hard to spot. These might need digital forensic or much more difficult technological processing. Fortunately, there are government agencies, universities and tech firms that are working on detecting deepfakes.
But as an audience, you must be careful and somewhat skeptical about everything on the internet. Think twice, don’t take for certain the authenticity of everything you see because the “seeing is believing” motto doesn’t work in our time. Watch for what you share or what your friends do. Go for the roots, the origin of the contents. The world of the internet is not as innocent as you might think it is. As attributed to Mark Twain, “it is easier to fool people than to convince them that they have been fooled.”
The future of deepfakes
AI technology is booming, as are various deepfake technologies. There seems to be no possibility that it’s going away anytime soon. Deepfakes can be used for good deeds, like making films, fun and entertaining video contents. Deepfakes even help people get back their voice lost to diseases.
Unfortunately, there are several ways to create severe damage by deepfakes. With the advancement of AI, deepfakes will be much more accurate. But these AI advances should increase the opportunity to battle against deepfakes too. AIs are already being used to detect deepfakes, and with time, deeptake Dr Jeckle will be more vital of a tool to fight back its own Mr Hyde.