Viral TikTok videos of Tom Cruise have the internet — and digital imaging forensics experts — buzzing about how far AI image manipulation has really come. Four videos of Cruise emerged on the social video-sharing platform in late February under the username @deeptomcruise, raising eyebrows about whether Cruise himself had actually taken up contributing to TikTok.
Recent media coverage has revealed that the real Tom Cruise does not have a TikTok, but that the videos were fabricated using deepfake AI image manipulation technology.
The creator of the videos Chris Ube says that the account name should have been a dead giveaway that the videos were a hoax, though some were truly stumped when they learned that Tom Cruise did not actually post a video of himself playing golf on TikTok.
If you haven’t seen them already, you can watch the videos here.
Ube said that the videos were not intended to trick anyone or raise alarm, however notes that he did have one goal in mind when he made the posts: awareness.
After seeing the videos, many digital imaging experts have expressed concern about the pace at which AI image manipulation is developing. Hany Farid, one of such experts from UC Berkeley, said that he could not detect the usual discrepancies that separate deepfake videos from legitimate ones. He paints an imaginary scenario of a Jeff Bezos deepfake, where Bezos might be depicted discussing a plummet in Amazon stock profits, to point out how anyone with sinister motives could weaponize the technology.
Ume has suggested that in just a few years’ time, the average person may have access to this deepfake technology on simple social media platforms — which is why he hopes for better detection software and online regulation.
Deepfakes are just one of many emerging ways that call the reliability of information spread via the internet into question. As data scientists develop AI technologies at an exponential rate, Ube hopes that the regulation and monitoring of these technologies can keep up.
