Baidu, which is often called the "Chinese Google", has published a white paper, which focuses on the latest developments in the field of artificial intelligence (AI). The system is based on a neural network that is able to clone a human voice on the basis of the analysis of even very short piece of source material. The program is not only very good mimics of human speech, but also can provide its own peculiarities like accent.
Examples of imitation of human voices, the neural network can be found here .
Previous versions of this technology allow to create imitation of human speech based on the analysis of longer samples. In 2017, the team of engineers from Baidu Deep Voice has introduced a technology that can mimic human speech on the basis of a 30-minute source material. Competitive developments in this field, in turn, did a few faster. For example, the program VoCo from Adobe could mimic human speech based on a 20-minute demo. And the development of Lyrebird canadian startup has demonstrated an even more amazing opportunity – she had minutes original sample of the human voice to create his simulation. The new design also Baidu went even further – it is enough only a few seconds of source material.
At first glance it may seem that these technologies are of no practical use – so one indulgence. But this is a misconception. In the future these technologies will certainly find their sphere of use. Imagine a person who has lost the ability to speak and again it has attained, even by car. Or troubled child, not wanting to lie in bed until you hear your voice reading a story, while you are very far away and you just have no physical possibility to call him. This is only the small part of the possibilities that can open up this technology.
In Addition, this technology can be used, e.g., for creating personalized digital assistants that can talk with you real human and not a computer voice.
But, like any other technology, here, too, has its downside. We can deal with her abuse and use is not entirely legal purposes. The portal New Scientist reported that the current version of the program has been able to create one type of voice that when checking the voice recognition system in 95 percent of cases able to lie to her. And the people so generally appreciated the quality of the cloned samples by 3.16 points out of 4. The pace sooner or later we may encounter cases of fraud using artificial intelligence, journalists.
In the world there are already programs that use neural networks are able to change or even mimic human faces in the video. For example, now the Internet is flooded with pornorolikov in the models are replaced by celebrities. Of course, all this is still just an ordinary prank, but soon, in combination with technology that can very accurately mimic a particular voice, we may face another wave of "fake news" in which prominent figures from different fields and politicians will say the things that they would never say in reality.
Many people now can be fooled using a seemingly ordinary programs such as Photoshop. Imagine what problems we might encounter when the cause will take fall into the wrong hands, artificial intelligence, billions of times beyond the capabilities of photoshop.
As you know, everything in nature is interconnected, and the disappearance of one species will undoubtedly cause problems in others.
One of the leaders of the Stratolaunch project, Paul Allen posted a Twitter video of the testing of the prototype aircraft Stratolaunch, during which he made a few rounds on the runway.
CEO of Virgin Hyperoop One called for a week of innovation in the United Arab Emirates, where he demonstrated a full-scale model of the passenger capsules of the future transport system.