Hollywood-quality deepfakes: the end of visual trust
The latest real-time deepfake demo, where a human operator controls a digital avatar with near-perfect facial movements and gestures, has set off every alarm. The level of detail in lighting, contrast and motion physics surpasses anything seen so far, and although some experts doubt it is truly real-time, the consensus is that the technology is advancing faster than society can absorb.
The qualitative leap in deepfakes
The video circulating shows an individual controlling, with a delay of under a second, the expressions and gestures of an AI-generated female character. The lip sync, blinking and emotional delivery are practically indistinguishable from a real recording. Part of the analysis community argues this is an offline rendering process, not real time, and that the apparent immediacy is the result of post-editing. Others, however, argue that today's computing power, especially with the arrival of quantum processors such as Google's Willow chip (capable of performing in 5 minutes calculations that would take a supercomputer thousands of years), makes real-time transmission plausible with the right hardware.
The technical debate centers on latency: skeptics point out that in the video the operator's image appears delayed relative to the avatar, which would suggest the operator imitated the avatar, not the other way around. But even if the demo is not genuinely real-time, the pace of advances in open-source models, many developed in China, suggests the gap will close in months, not years.
Industries in the crosshairs: Hollywood, scams and porn
The immediate impact is expected on three fronts. First, the film industry: if a single creator can generate photorealistic scenes with one actor and a powerful PC, traditional shoots with crews of hundreds of professionals (camera, wardrobe, art direction) are called into question. There are already examples of entire rock albums generated by AI, and audiovisual production will trinc the same path. Second, scams: training an AI on a person's videos and photos allows their identity to be impersonated in video calls, opening the door to mass fraud, blackmail and disinformation. Third, adult content: the generation of hyperrealistic avatars with customizable movements and expressions will transform the porn market, where on-demand scene creation is already being speculated about.
The debate intensifies when considering the combination of deepfakes + quantum computing. Recent quantum algorithms have proven up to 13,000 times faster than classical ones on specific tasks, which would allow real-time rendering at resolutions and realism unimaginable today. That would turn any video call into a potential deception and destroy trust in digital communications.
The social response: toward a digital notary?
Faced with the prospect that any video could be fake, one school of thought advocates creating a profession of "digital notary", a public official who certifies the authenticity of digital documents and communications, something already proposed in legal circles. Another, more skeptical view points out that technology moves faster than any regulation, and that the solution will lie in educating the public and developing decentralized verification systems, such as blockchain applied to the chain of custody of videos.
The question nobody answers is: when any audiovisual content can be generated by AI, what value will the image retain as evidence? Society faces a paradigm shift where visual truth will cease to be a fact and become a matter of faith.
Related discussions in the forum