4 January 2019


The title above is from Tim Simonite’s December 28, 2018 article he posted to the cyber and security website, WIRED.com. He begins: First algorithms figured out how to decipher images. That’s why you can unlock an iPhone with your face. More recently, machine learning has become capable of generating and altering images and video.” 

“In 2018, researchers and artists took Artificial Intelligence (AI)-made and enhanced visuals to another level,” Mr. Simonite wrote. Indeed, “software developed at the University of California Berkeley, can transfer the movements of one person, captured on video…to another,” he notes. “The process begins with two source clips — one showing the movement to be transferred; and, another showing a sample of the person to be transformed. One part of the software extracts the body positions from both clips; another learns how to create a realistic image of the subject for any given body position. It can then generate video of the subject performing more or less any set of movements. In its initial version, the system needs 20 minutes of input video before it can map new moves onto your body.”

“The end result,” Mr. Simonite wrote, “is a similar to a trick often used in Hollywood. Superheroes, aliens, and simians in Planet Of The Apes movies are animated by placing markers on actors’ faces and bodies so they can be tracked in 3-D by special cameras. The Berkeley project suggests machine learning algorithms could make those production values and much more accessible.”

And, “AI-enhanced imagery has become practical enough to carry in your pocket,” Mr. Simonite notes.

“The Night Sight feature of Google’s Pixel phones, launched in October 2018, uses a suite of algorithmic tricks to turn night into day,” Mr. Simonite explained. “One is to combine multiple photos to create each final image; comparing them allows software to identify and remove random noise, which is more of a problem in low-light shots. The cleaner composite image that comes out of that process gets enhanced further with help from machine learning. Google engineers trained software to fix the lighting and color images taken at night using a collection of dark images paired with versions corrected by photo excerpts.”

Imaginary Friends

Mr. Simonite/WIRED then displayed a series of photos showing “people, cars, and cats that don’t exist — the images were generated by software developed at chip-maker Nvidia, whose graphics chips have become crucial to machine learning projects. “

“The fake images were made using a trick first conceived in a Montreal pub in 2014 by AI researcher Ian Goodfellow, who is now at Google,” Mr. Simonite wrote. “He figured out how to get neural networks, the webs of math powering the current AI boom, to teach themselves to generate images.The versions Goodfellow invented to make images are called generative adversarial networks, or GANs. They involve a kind of duel between two neural networks with access to the same collection of images. One network is tasked with generating fake images that could blend in with the collection, while the other tries to spot the fakes. Over many rounds of competition, the faker — and the fakes — get better and better.”

AI Art

“In a scene from the experimental short film, Proxy, by Australian composer Nicholas Gardiner, footage of Donald Trump threatening North Korea with “fire and fury,” is modified so that the U.S. president has the features of his Chinese counterpart, Xi Jinping,” Mr. Simonite wrote. “Gardnier made his film using a technique initially popularized by an unknown programmer using the online handle Deepfakes. In late 2017, a Reddit account with that name began posting pornographic videos that appeared to star Hollywood names such as Gal Godot. The videos were made using GANs to swap the faces in video clips. The Deepdakes account later released its software for anyone to use, creating a whole new genre of online porn — and worries the tool and easy-to-use derivations of it might be used to create fake news that could manipulate elections.”

“Deepfakes software has proved popular with people uninterested in porn,” Mr. Simonite wrote. Gardiner and others say it provides them a powerful new tool for artistic exploration. In Proxy, Gardiner used Deepflakes package circulating online to make a commentary on geopolitics in which world leaders such as Trump, Vladimir Putin, and Kim Jong-il swap facial features.”

Really Unreal

“Generative adversarial networks usually have to be trained to create one category of images at a time, such as faces or cars,” Mr. Simonite wrote. “BigGAN was trained on a giant database of 14 million varied images scraped from the Internet, spanning thousands of categories, in an effort that required hundreds of Google’s specialized TPU machine learning processors. That broad experience of the visual world means the software can synthesize many different kinds of highly realistic looking images.”

IBM’s “DeepMind released a version of its models for others to experiment with,” Mr. Simonite wrote. “Some people, exploring the “latent space” inside — essentially testing the different imagery it can generate — share the dazzling and eerie images and video they discover on Twitter under the hashtag #BigGAN. AI artist Mario Klingemann has devised a way to generate BigGAN videos using music.”

DeepFake Imaging, Enhanced By AI, Is Going To Become A Big Problem In 2019

As cyber security guru Bruce Schneier has written, “there is an arms race between those creating fake images and videos; and, those trying to detect them.” In a blog post last year, “Detecting Fake Videos,” Mr. Schneier wrote: “These fakes, while convincing if you watch for a few seconds on a phone screen, aren’t perfect (yet). They contain tells, like creepily, ever-open eyes, from flaws in their creation process. In looking into DeepFake’s guts, Lyu realized that the images that the program learned from, didn’t include many with closed eyes (after all you wouldn’t keep a selfie where you were blinking, would you?). “This becomes a bias,” he said. “The neural network doesn’t get blinking. Programs also might miss other “psychological signals intrinsic to human beings,” according to a paper Mr. Lyu wrote, “on the phenomena such as breathing at a normal rate, or having a pulse.” “While this research focused specifically on videos created with this particular software, it is a truth universally acknowledged that even in a large set of snapshots might not adequately capture the physical human experience; and so, any software trained on those images may be found lacking,” Mr. Schneier wrote.

“Lyu’s blinking revealed a lot of fakes,” Mr. Schneier wrote. “But, a few weeks after his team put a draft of their paper online, they got anonymous emails with links to deeply faked YouTube videos, whose stars opened and closed their eyes more normally. The fake content creators had evolved.”

Mr. Schneier concludes, “I do not know who will win this arms race, if there ever will be a winner. But, the problem with fake videos goes deeper; they affect people even if they are later told they are fake; and, there will always be people that believe they are real, despite any evidence to the contrary.”

Huge Implications For The Intelligence Community — For Espionage, Clever/Sophisticated Spoofing

Obviously, there are huge implications here for the Intelligence Community, espionage and clever/sophisticated spoofing. Deception is an underappreciated talent and technique that can pay big dividends. But, the darker angels of our nature are also going to increasingly employ this emerging technology in clever and devious ways we don’t expect or understand very well. We have to expect that as AI matures, along with 3-D algorithmic imaging, and machine learning, the bad guys may have the upper hand. This genre is going to be a problem in 2019, and likely beyond. But, we can also use this technology in clever ways that could provide us with critical intelligence, or undermine our adversary. Stay tuned. RCP, fortunascorner.com

No comments: