Speech2Face, the AI that studies your voice to reconstruct your face
The progress of artificial intelligence never ceases to surprise us. And if we have the impression that we are facing the most extraordinary innovation, other achievements remind us that this is only the beginning. The latest discovery and creation by MIT researchers will certainly leave you speechless: the Speech2Face, a tool that can reconstruct a person’s facial features based solely on samples of their voice.
Artificial intelligence finds application in all fields and some of them can be quite worrying. This is the case with this one.
The field of security and surveillance has much to gain from the rise of artificial intelligence, that’s for sure. Researchers at the Massachusetts Institute of Technology are actively working on the subject…
Speech2Face, an AI that reconstructs a face
Their latest achievement is an AI that can reconstruct a face simply by listening to a voice. The algorithm, called Speech2Face, attempts to reconstruct someone’s face with their voice as the only source of information. This gives you an idea of who you’re talking to, wherever they are.
To do this, the researchers used a generative antagonistic network driven by videos from YouTube and other online platforms to match biometrics to sounds.
An AI that raises ethical questions
This raises the question of the use of such technology, in terms of privacy or discrimination, for example. “Although this is a purely academic survey, we believe it is important to explicitly discuss a set of ethical considerations, given the potential sensitivity of facial information,” said the project managers on the GitHub page dedicated to the project.
One thing is certain, hiding your identity may be very, very difficult in the years to come.