Speech2Face was presented at the Massachusetts Technical Institute. The prototype of the newly developed AI has the ability to determine with high confidence the appearance of a person by his voice.
While this development is distributedexclusively in academic circles and is purely scientific in nature. Speech2Face is another kind of generative-competitive neural networks, the task of which is to select the parameters of a person’s appearance based on voice data by trial and error.
The results of AI are impressive: The key features of the appearance of participants in the experiment are reproduced as accurately as possible, while it is not yet known what algorithms the American developers used when creating Speech2Face. And another important addition: the system can, on the basis of a voice, determine both gender, and age, and even the skin color of a person.
Most of all, developers fear that theirdishonest people will want to use the work in their mercantile plans even before the work is completed. Scientists will have to improve the algorithms, for which large data arrays and tests are needed to increase the representativeness. The developers understand that without the help of the global scientific community is not enough.