This AI Is So Smart That It Can Guess Your Face Via The Voice
Jyotis
The MIT researchers have just introduced an AI that can analyze a short audio clip recorded any voice and give some suggestions of the speaker’ facial contours.
- 4 Ways AI Could Change The Mobile Gaming Industry
- This South Korean YouTuber Is The Result Of Deepfake Technology
- New ‘Deep Nostalgia’ AI Allow Users To Bring Old Photos To Life
Facial Recognition AI Via Voice
Researchers from the Massachusetts Institute of Technology (MIT) have created an artificial intelligence (AI) that makes the whole world surprised. It can analyze a short audio clip recorded any voice and give some suggestions of the speaker’ facial contours in reality.
The results aren’t 100% correct, but the AI works well, even beyond our expectation. Is it time for us to worry about the supernatural intelligence of AI? How can it give such an astounding inference based on a few of tiny data?
Biometric Characteristics
According to a paper that has just been published for a couple of days, these MIT researchers explained how they applied to train a generative adversarial network. The network then analyzes a series of short audio clips and matches “several biometric characteristics of the speaker.” The results showed “matching accuracies that are much better than chance.”
The above part describes the new AI according to the professional language for researchers. In other words, the Speech2Face algorithm can “read” and curve the basic contours on the speaker’s face just based on his voices.
Privacy Issue
On the GitHub page of the project, the researchers from MIT also give a necessary caution relating to this AI. They admit that technology may face countless worrisome questions concerning discrimination and privacy issues. They said,