Danger that AI deep fakes could usher in an era of rampant misinformation - Dr William Stephen Jones

When we watch news updates on the TV, view videos on YouTube, browse social media clips, or even see documentary footage, we generally assume the videos and images are real.

But what if a technology existed capable of creating fake videos and images so lifelike that distinguishing them from the real thing was nigh on impossible? Welcome to the world of AI-generated deep fakes.

So, what exactly are AI-generated deep fakes? AI, or Artificial Intelligence, is the science and engineering of creating intelligent machines, especially intelligent computer programs and algorithms; this represents the aspirational goal. Deep fakes are a product of this technology, where media is either digitally manipulated or entirely created anew. This manipulation often involves replacing one person's likeness with another's.

Hide Ad
Hide Ad

The current use of deep fakes range from arguably harmless parodies to more serious and malicious cases where they depict individuals in false situations. Most recently, this has been an issue for Taylor Swift who has been the target of fake explicit images which have been viewed millions of times on social media.

Taylor Swift at the Super Bowl LVIII. PIC: PA Wire.Taylor Swift at the Super Bowl LVIII. PIC: PA Wire.
Taylor Swift at the Super Bowl LVIII. PIC: PA Wire.

In contrast, a classic example of parody is the creation and viral spread of deep fake videos featuring celebrities, such as Tom Cruise, often a target of these technologies. Here, AI might be employed to falsely depict him engaging in comedic acts or delivering funny lines that are out of character. Relatedly, deepfake technology has found its way into film and television, used to de-age actors, or resurrect deceased celebrities on screen, such as the recent announcement of the Elvis Presley AI concert later this year, in London. However, the same technology can also be used in more directly concerning and destructive ways, such as manipulating speeches of politicians, creating fake endorsements, creating fake pornography, cyber bullying, identity theft, fabricating legal evidence, deceptive advertising, psychological warfare and creating and spreading misinformation and distress in various other ways, some of which are yet unknown.

The creation of deep fakes is currently limited to AI programmers who possess reasonably advanced computer hardware. However, as the march of technology and AI continues, it will soon become easier for non-specialists to develop deep fakes, possibly through the convenience of a mobile app.

The drawbacks of widespread deep fakes are clear and troubling. Beyond the disturbing applications listed above, these technologies threaten to erode the trust we have in the legitimacy and authenticity of potentially all videos and photos, since they could just as easily be fake as real, and perhaps even more likely fake, since the possibility for real visuals is limited to their original creators, whereas almost anyone could create a fake. This scenario could have the most far-reaching consequences to society, ushering in an era of rampant misinformation, a misinformation apocalypse, where discerning truth and authenticity becomes increasingly challenging, if not impossible.

Hide Ad
Hide Ad

There is the possibility that AI technologies will be developed to detect deepfakes in lockstep with the deep fake technologies themselves. Nevertheless, even if a robust validation technique is established, once doubt is introduced, it will be difficult to regain public trust in digital media, since the possibility of further undetectable deepfake innovation is ever present.

So, these are the negatives – potential Armageddon level of trust erosion, and a plethora of egregious specific uses – but what about positives, for there certainly are some of these. Firstly, as alluded to above, the deep fake technology opens the doors to new forms of entertainment, allowing actors to transcend age or physical limitations and even bring past legends to new films. How about giving viewers the option to see themselves in movies or shows. Similarly for gaming, this technology could make experiences more immersive by allowing players to insert their own images into games. And perhaps could bring historical figures to life in documentaries, making learning more engaging.

Deepfake technology also has practical applications. In my own area of research, AI applied to healthcare, this technology could be used for medical training and simulation, creating realistic patient cases and surgical procedures, which could be paired with augmented virtual reality for a more immersive training experience.

Deep fakes are a double-edged technology, with an extremely sharp negative edge. Once this technology becomes fully developed, available, and widespread it will very likely undermine trust in visual media. Perhaps the only silver-lining here is that it could reinvigorate the idea of community engagement, live events, face to face meetings, etc., an aspect of life that has been eroded with the advent of social media and exacerbated by lockdown isolation.

Dr William Stephen Jones is director of research and lecturer in Data Science, Artificial Intelligence and Modelling (DAIM) at the University of Hull.

Comment Guidelines

National World encourages reader discussion on our stories. User feedback, insights and back-and-forth exchanges add a rich layer of context to reporting. Please review our Community Guidelines before commenting.