Experiments in Deepfakes @ MIT Media Lab Fluid Interfaces Group

I was fortunate enough to be accepted in to the first cohort of the MIT Media Labs ‘Experiments in Deepfakes’ run by Prof Pattie Maes from the Fluid Interfaces Group. The course ran the gamut of positive and negative impacts of AI on society, different use cases, the history of AI and some of the core mathematics behind it.

Throughout the course we were tasked with conducting our own experiments with different algorithms.

One of my favourite experiments is one that is very topical (once) again.

Here, Putin is used as the ‘driver’ video to then animate five still images of Trump. By including the driver video the viewer immediately understands who is pulling the strings. It’s a deliberately benign use case, where the viewer understands the game being played and there is no question of this being real footage of Donald Trump.