A group of MIT researchers recently unveiled a new way of applying to the powers of Artificial Intelligence. This is able to predict the face of a person simply by analyzing a sample of his voice.
The results that can be obtained with this Artificial Intelligence have not yet reached perfection. Even so, it is already able to creep up on reality, as you can see from the example below.
We must take into account that all these results were obtained simply with a sample of the voice of these people. At no time was any physical trait taken so that it could reach the final result.
This AI was trained with Youtube videos
This milestone was reached using thousands of videos present on Youtube, simply with people talking. During this process, this Artificial Intelligence was able to create correlations between voices and human faces. In this way, this technology is capable of reproducing various human attributes such as gender, ethnicity or even age.
The researchers point out that this new AI is not able at any moment to reach the identity of a person. That is, all the results obtained are only based on the videos that served as teaching for this Artificial Intelligence.
It is further pointed out that if one uses, for example, a language that has not been taught, the results obtained may not be so similar when what we see above.
Privacy issues are present in the minds of researchers
Privacy issues are one of the concerns of the creators of this interesting technology. Although it is not able to unambiguously identify a person, the researchers state that this AI should only be used for research purposes.
They thus want to ensure that the technology they have developed can not be used for the purposes of spying on populations. For this reason, they emphasize that the results obtained have a direct relation with the samples used in their development period.