We are living in a world where Artificial Intelligence (AI) technology is drastically progressing each day. The use of AI on different technologies and gadgets are becoming more common nowadays. From the face unlocking of gadgets to filters on snap chat, AI is making its way.
The tech giant, Samsung, introduces a new AI technology known as Samsung Deepfake AI. The deepfake AI can create a deepfake video of anyone just by using a single picture, awesome right? But, can you imagine someone making a deepfake video of you just by stealing your one portrait form your facebook? Now, it’s not so awesome right.
A Samsung artificial intelligence lab in Russia was successful in developing this deepfake AI technology. However, the criminals who would love to do that don’t have their hands on this AI, yet.
Generally, to create a fake deepfake video of someone needs large sets of data. But, Samsung has figured this out. The Samsung deepfake AI can generate the deepfake videos by using just as little data needed as a photo.
Using this deepfake video can be of very fun. A single portrait of someone can be used to say things or appear to do they actually never did. Like, while demonstration, the Mona Lisa was animated in three different videos (clips). It seriously looked fun the way I looked.
But, like everything here’s the catch. The rapid use and growth of these technologies ( AI ) also create risks in society. According to Hany Farid, a Dartmouth researcher who specialized in media forensics to root out deepfakes, the risks can be misinformation, election tampering and fraud.
” Following the trends of the past year, this and related techniques require less and fewer data and are generating more and more sophisticated and compelling content.” Although Samsung’s process can create glitches, these results are another step in the evolution of techniques leading to the creation of multimedia content that will eventually be indistinguishable from the real thing.Hany Faird, Dartmouth researcher who specialized in media forensics to root out deepfakes
The clips made through deepfake AI can be considered as “The Digital Puppets”. Though computer manipulation of videos is already there on the media, deepfake systems make it easier to create and even harder to detect. The deepfake software develops clips by using machine learning to fabricate a moving, speaking human.
Here’s a video of the creation.
As in the past, for creating deepfake videos, a large number of photos were needed. So, basically, videos of celebrities were created. Like Nicholas Cage, was deepfake into shows and events he never went. But, this Samsung deepfake AI doesn’t need large photos. It is so because it first goes through a ‘meta-learning stage’ in which it watches a lot of videos to specifically learn how human face moves. Then, it creates a video clip (reasonable) through the data obtained form learning. These clips have their details fudged though.
The rapid advancement in the AI industry means more technology like this, Samsung deepfake AI. Also, we can easily overcome the glitches by simply repeating the tasks. It is only a matter of time when it hits the market and someone can end up in revenge porn video just by the smiling portrait posted on social media.