One of Leonardo Di Vinci’s most famous pieces of artwork, the Mona Lisa, has been brought to life with the help of deepfake AI.
For those of you who don’t know deepfake is a technology that has been in development for some time now and is getting to some scarily good levels. Deepfake AI examines and uses a series of images to render video footage of the person in the images. These fabricated videos have the subjects turning their heads and even speaking. The AI is able to put together a 3D representation of a person from these images by looking carefully at what the face looks like at different angles, lighting and with different expressions among other things. The video below explain in more detail how the technology works.
Now back to the Mona Lisa. It’s pretty remarkable what the AI has achieved here. The footage you see in the video below is from a single image the artwork which is obvious because there is only one Mona Lisa painting after all.
This is the latest iteration of deepfake and was created in Samsung’s AI research lab in Moscow.
Now, you’re probably wondering how this was possible from a single image as with other deepfake renderings a series of image was used. Well, Samsung’s algorithms were trained with 7,000 images of celebrities on a public database collected from YouTube. The AI system then mapped the facial features and movements it collected onto the image to bring it to life.
The AI also made videos of other famous historical figures like Salvador Dali, Fyodor Dostoyevsky, Albert Einstein and, my personal favourite, Marilyn Monroe.
Samsung describes their work as “realistic neural talking heads” so you now know what to search for on YouTube.
You may be wondering what the use of “realistic neural talking heads” could be. Well the guy that created the fake video of Barack Obama using a similar technology, yes he did that, Dr Supasorn Suwajanakorn, back in 2017, says that one potential use of the technology could be for grieving families who have recently lost a loved one.
Although he seems optimistic deepfake, and technologies similar to it, have had much criticism since it’s development the most obvious being that people can be lied to using this technology. The misuse of such technology could lead to the duping of entire populations by creating fake videos of politicians for example. It’s a scary thought and not the only thing it can be used for.
Dave Coplin, chief executive of AI consultancy The Envisioners, said: “The rise of convincing deepfakes is something that could be really problematic unless we have this conversation. Members of the public need to know how easy it is to create convincing fake videos.”
Below is the fake video of Barack Obama. Let me know your thoughts in the comments.