Samsung is developing AI that will teach your pictures talk

Fake news is a phenomenon that is difficult to distinguish from the actual news, not only because of the content of the articles and their viral exposure on social networks, but also because now it’s easier to create a really realistic looking fake videos and photos. In addition, fake news is not a good practice because they affect people’s behavior, the results of the elections and help the crooks to commit their bad acts. Imagine what would happen if artificial intelligence will be able to create fake, but realistic video with people of only their photos. Samsung recently developed a new AI that can do it.

Within the artificial intelligence lab located in Russia, AI Samsung taught to fabricate fake videos with people having only a single picture or even a picture. And it even looks quite realistic. One of the posted videos shows how the Mona Lisa by Leonardo da Vinci moves and talks, but the picture was the only image of the Mona Lisa, which was used to create the video because her other images simply do not exist. Far to create fake videos to provide a realistic result, it took a lot of pictures, but thanks to the efforts of the South Korean giant now requires just one photo.


Samsung new creation is called “neural realistic talking heads” (realistic neural talking heads), and the word “neural” is used here not just to be witty. A neural network is a type of machine learning that can simulate the human brain, and they have a lot in common with the new algorithm Samsung. To create a video clip AI views a lot of videos to learn how to move human faces, and then applies all this knowledge to create realistic videos just from one picture.

There are many areas that could benefit from this technology, but I hope it will not be used to create a fake viral video or blackmail.

Share your opinion in the comments under this material in our Telegram chat.

Leave a Reply

Your email address will not be published. Required fields are marked *