TECH4GOOD
Like many of us, you would start your day by opening your social media apps to begin your daily dose of doom-scrolling. As usual, you get bombarded with posts and feeds in different forms. You watch a video of your favorite celebrity endorsing an investment product you believe is unusual for her to do. The video looks real, her voice very familiar, and the emotions etched on her face raw and undeniable. But you suddenly noticed that her lips were uttering words different from what you heard.
Welcome to the disturbing world of deepfakes, where reality is a meticulously crafted illusion generated by algorithms and thrusting us into a future where truth is a pixelated mirage. You can forget about photoshopping. The future of image manipulation is here, and it is far more deceptive because it blurs the lines between real and fabricated. That is the unsettling truth about deepfakes, and what is even more worrisome is the fact that this technology is just emerging.
Deepfakes, the unholy offspring of artificial intelligence and human ingenuity, are able to stitch together convincing lies from mere scraps of truth. Imagine politicians delivering fabricated speeches and your friends saying things they never did. Deepfakes have garnered widespread attention for their potential use in creating child sexual abuse material, celebrity pornographic videos, revenge porn, fake news, hoaxes, bullying, spread misinformation, and financial fraud.
Deepfakes are synthetic media that have been digitally manipulated to create fake images, videos, and audio. This often involves digitally altering someone's appearance or voice to make it seem like they said or did something they never did. Think of it as super advanced photoshopping, but for video and audio as well.
The use of this technology is rampant in social media platforms, and its emergence has caused alarm to governments and companies alike. In fact, social media platforms have announced that they will remove content modified or fabricated by artificial intelligence. Some countries have introduced regulations that would totally ban the use of the technology or require people uploading deepfakes to disclose that such have been altered, and, further, failure to disclose would be criminalized. Legislative proposals have started to materialize in the Philippines as well.
The regulation and banning of deepfakes is a complex and evolving landscape because it is a nuance technology, with different countries taking varied approaches. Globally, no comprehensive ban on deepfakes makes preventing its spread across borders challenging. Regulations must also balance free speech concerns by being careful not to restrict freedom of expression unduly.
Even if policies are in place, enforcement can be challenging because even defining what constitutes injurious deepfakes has not yet been established and can be subjective. As the technology becomes mature, people will get better at using it. Who can best decide which media is deepfake and which is not?
Crude ways of creating deepfakes have been around for sometime, using early techniques like photoshopping that can actually switch images. Deepfakes today look almost realistic because they are done using a form of artificial intelligence called deep learning. It is able to analyze a massive amount of data, such as photos and videos of a specific person. It allows it to learn and understand the person’s unique facial features, voice patterns, and gestures. Once it has learned enough, it can swap someone’s face or voice into another person’s body in a video or audio recording, convincingly making it challenging to tell if it is a deepfake.
Deepfakes can be used for various purposes, both good and bad. In entertainment, they are often used to create humorous videos using celebrities’ faces. They can also be used to develop satirical or critical videos of politicians and public figures. The worst cases include their use for scams, frauds, the spread of misinformation, and black propaganda.
Fortunately, tools are now available to help detect whether a video is a deepfake. These tools also use machine learning algorithms to identify inconsistencies in video or audio content. Personalities involved in viral video content cannot just quickly tell everyone that those materials are deepfakes.
It is becoming evident that there is today the need for effective regulation and international cooperation to prepare for technological advances. We also need to see local innovations in the area of detection tools, which are essential to help combat misinformation and build trust in cyberspace. Awareness programs to help people remain vigilant and critical of the information they see online, especially in the lead-up to elections. Being aware of the potential for deepfakes and taking steps to verify information can help them protect themselves from misinformation and manipulation.
It is vital for everyone to be discerning when viewing or sharing media online. But let us not use deepfakes to demonize AI further. AI is good. ([email protected])
(The author is an executive member of the National Innovation Council, lead convenor of the Alliance for Technology Innovators for the Nation (ATIN), vice president of the Analytics and AI Association of the Philippines, and vice president, UP System Information Technology Foundation.)