Sound leak? "Sound DNA" that may be copied
Have you ever thought about what you should do when your unique voice appears in front of you one day?
This is not a parody of the guests, but something that can really happen now. In 2019, the application of AI technology is becoming more and more diversified. IFLYTEK, Sogou and other technology companies have released the application of speech synthesis technology. Through AI, users can change their voices of social celebrities or other voices they want to imitate in a second.
People who especially like the game industry should be aware that many games now have their own voice systems, while some profit-making people directly use voice-changing software to change their voices. But it can only change the sound one-way, fixed a few timbre parameters, and can not reach the point of flexibility.
And what's the difference in this technology? Different from the software, the end-to-end speech style separation and modeling scheme is adopted, and multiple groups of neural networks are used to independently model the different dimensions of speech, such as timbre, emotion, style and so on, so as to guide the final synthesis. For example, smart phone robots use this advanced technology instead of the traditional fixed module compilation system. It can be used in voice synthesis, voice change, voice cos and other scenarios to convert human sound into sound from multiple angles and different directions.
Internet technology is quietly changing our lives, for artificial intelligence enterprises, the wide application of speech recognition technology is not difficult. However, the hidden dangers of ethics and security may become a problem that can not be ignored with the development of AI technology.
In fact, the phenomenon of "confusing the fake with the real" behind AI technology appears not only in the sound field, but also in a recent Samsung technology application.
According to foreign media reports, researchers at Samsung's artificial intelligence laboratory in Moscow, based on a large number of motion pictures and video materials, as well as "deep convolution neural network" training, accurately identify some facial features through AI technology, and can turn still images into motion pictures or even videos.
In the experiment, the researchers generated videos of Einstein, Marilyn Monroe and even the Mona Lisa based on still images of them, but the video quality is currently low.
In other words, with the development of AI image generation technology in the future, fake videos can be generated from just one photo.
Whether it is AI voice change or AI face change, the rapid development of high technology also brings new challenges to legal rules.
It is worth noting that in the compilation of the right of Personality of the Civil Code (draft), which is often deliberated by the National people's Congress in April, an article was formally added: no organization or individual may infringe upon the portrait rights of others by means of information technology forgery.
In other words, a good signal is that the future application of AI face changing or even speech synthesis technology may be regulated by law.
Especially in the field of artificial intelligence speech. At present, the landing application of AI technology is constantly opening the boundary between the virtual world and the real world. What we need to think about is that if these technologies become more and more powerful and popular in the future, it will become more and more difficult to distinguish between true and false. With the development of technology, how to control the hidden security problems at the same time?
