The internet was recently set abuzz by Mohamed Salah‘s video on Palestine, leaving viewers with an uncanny feeling.
If you found yourself unnerved, you’re not alone; numerous voices on X (formerly known as Twitter) echo your sentiments, questioning the authenticity of the video, which some speculate might be AI-generated.
A discerning eye noticed peculiarities – Salah’s motionless shoulders, limited upper-face microexpressions contrasting with his unchanged lower face, and a haunting lifelessness in his eyes. Deepening the intrigue, a Deepfake detector indicated a staggering 60% probability that the video might indeed be a product of advanced AI and deepfake technologies. The video in question was suggested to have been edited using “Avatarify,” although many speculate that a more advanced software might have been at play.
The emergence of companies like Synthesia has amplified this discussion. Synthesia offers custom avatars, replicating facial expressions, accents, and speech patterns, all for an annual fee of $1000 per avatar. The process, as it turns out, is shockingly simple.
We decided to test the waters and tried a free demo video on Synthesia, which was effortlessly created in less than one minute. This rapid demonstration underscores how accessible and user-friendly these AI platforms have become, amplifying the concern regarding the potential misuse of such technology.
According to the company’s website, “It currently takes about 10-15 minutes of video recording in front of a green screen. After you’ve submitted the footage to our team and paid for the custom avatar, you’ll receive your custom avatar on your Synthesia account within 10 days.”
As for the safety of usage of the avatars, Synthesia stated: “Creating a new custom avatar requires explicit consent from the person to be made the avatar and our team manually checks and processes all custom avatar video footage.
Moreover, existing images or video footage can’t be used for creating custom avatars – our custom avatar creation requires about 10-15 minutes of dedicated video recording.
This means that it is impossible to impersonate a celebrity or a politician.”
While Synthesia is the most known service to provide this feature, there are plenty of other companies, and we’re sure Salah’s team can even afford to have their own AI platform just for him.
Now as to why everyone is creeped out by the video, it’s because of a phenomenon called “Uncanny Valley”.
The underlying discomfort experienced by viewers finds its roots in the “Uncanny Valley” phenomenon – a concept prevalent in the realm of robotics and artificial intelligence. When humanoid robots or animated characters closely mimic human features but miss the mark in subtle ways, an unsettling sensation ensues.
This disquiet arises from the stark contrast between near-lifelike realism and discernible imperfections, creating a dip in the graph of human comfort. AI-generated videos, often straddling this fine line between reality and the surreal, frequently evoke this eerie unease.
Now the question remains: is Salah’s video genuine, or is it a testament to the power of AI ingenuity? And if it is indeed AI-generated, does this revelation diminish its impact or significance in your eyes?