Artificial intelligence could detect DeepFakes

Artificial intelligence could detect DeepFakes

The DeepFakes, these modified faces more real than life, have been further refined in recent years. In the era of fake news, these digital manipulations of course raise important questions. If the human eye can already almost no longer distinguish the true from the false, Artificial Intelligence of the type deep learning (through deep learning) would however still be able to do the devil’s part. Thus, researchers from the Russian company ID R&D compared the performance levels of a specialized AI and ordinary individuals in the recognition of fake photos. At the end of the test, the AI ​​was able to detect the faked photos with an error rate of 0% while the poor human beings that we are were mistaken for almost one out of three faked images (30% error rate). error).

Even better, the AI ​​was on average 10 times faster than humans in analyzing each photo. In short, if some AIs are able to manipulate photos or videos, others are ALSO capable of detecting these same manipulations, proof that in terms of technology everything is not black or white depending on the point of view that the we decide to take.

Leave a Comment

Your email address will not be published.