Miracle of Thinking

How do you guys are spending your pandemic?

I tend to either cook, read or play a mobile game (to be honest I am playing it for an emulator). The game is called “State of Survival”

Tell me what do you do during these times.

Hello guys!

Care to guess what I am referring to?

I am not sure that I could live without it.

Deepfakes

I do not know how many of you know what deepfakes are but since we are on Tumblr I assume you have some ideas. Anyway, I think I should talk a bit about the origin and you will understand better what deepfakes are.

Quick quiz: Who created/posted the first deepfake/s?

Deepfakes first appear on Reddit (duh) back in 2017 when a Redditor by the name of deepfakes (d'oh again and answer to your quiz above) made posts with videos of Scarlett Johansson (go figure) where he "copied" her face on bodies of X-rated actresses. You can imagine that meant a lot of editing (I guess on each frame) in the video. That is time-consuming and it takes a lot of determination. You can imagine that a deepfake was easy to recognize because the face would not mimic perfectly what was happening and ofc would miss completely the audio (if any).

Machine learning & deepfakes

Scary subtitle, right? Well, I think it is and let me tell you why. There are a few fakes there of Mona Lisa smiling or Salvador Dali as well as others. It seems researchers at Samsung's AI research center Moscow (sic!) developed a way to make seemingly realistic video and/or audio generated clips. Using machine learning and working with only with few static imagines (they can use even only one pic). Sounds scary, right? It makes me think in a way of a virtual Puppet Master from Ghost in the Shell. The only difference? You do not need to control those people :)

Google enters the scene.

Google made a breakthrough in controlling depth perception. What does that mean? Well, remember the Scarlett story from the beginning? The problem with deepfakes is that if you use one picture to add it to each frame you end up with the same face pretty much everywhere. Now with depth you can actually make it look real. Give depth to the face and even modify it.

Usage?

Of course, all companies are looking in these technologies for things like avatars for instance. Can you imagine a real-time avatar mimicking everything? :) Nice, right? Now imagine a deep fake Donald Trump saying "stuff" in a video to the president of China or whatever. Now that is scary, right? I mean if there was Obama saying crazy things about the president of China you would say to yourself "Clearly that is not how Obama talks/acts" but when it comes to Donald Trump going ballistic you might say "He actually said that" and accept it even faster.

You might think it is not gonna happen but there is actually a battle going on to automate the detection of deepsfakes to protect the 2020 elections in the US.

How serious is this? I will only say this: DARPA is actually actively researching to automate deepfakes detection spending tens of millions on media forensics. Weird, right? MEDIA FORENSICS. Let that sink in.