A $243,000 Deepfake Heist, a $10M Deepfake Competition, and More Deepfake News

It’s unrealistic to think that a technology can create impact immediately after its release. Most would say it’s ludicrous. Look at a technology like mesh networking. Established 2012, every future-thinker could see its societal relevance. But it took an event seven years later – the riots in Hong Kong – to bring the tech into mass prominence.

On the rarest of rare occasions, an invention comes along that fits a purpose, finds customers, and has instant cultural virality. The technology enters society like a meteor entering Earth’s atmosphere. Everyone is stunned.

The most recent meteor to hit Earth was a technology known as deepfakes.

The implications of deepfakes were instantly evident to everyone. That’s why Inevitable/Human (and thousands of others) have been following deepfake technology closely over the past year and a half. Here’s the latest on this technology.

The Latest Creations

The anticipation for a deepfake disaster has been building for quite some time. And it finally came:

In what may be the world’s first AI-powered heist, synthetic audio was used to imitate a chief executive’s voice and trick his subordinate into transferring over $240,000 into a secret account. 

The company’s managing director was called late one afternoon and his superior’s voice demanded the subordinate wire money to a Hungarian account to save on “late-payment fines”. The software was able to imitate the voice, and not only the voice: the tonality, the punctuation, the German accent.

Christopher Furlong, Vice

This is a poignant reminder that audio manipulation isn’t to be ignored. The first time I wrote about deepfakes, in October of 2018, I said this:

In the next 4-6 years, this technology [deepfakes] will reach the usability (and believability) of Snapchat’s FaceSwap. You won’t have to be well-versed in machine learning to create one of these fake videos.

FakeApp Threat

My prediction was right, but the timeline was way off.

Zao, a free deepfake face-swapping app is able to place your likeness into scenes from hundreds of movies and TV shows after uploading just a single photograph, has gone viral in China. The app was released on Friday [and by Sunday it had become the most downloaded free entertainment app in China’s Apple Store].

Jon Porter, The Verge

Another frightening example, called DeepNudes (safe to open) came earlier this summer. DeepNudes used neural networks to generate a realistically nude image from an image of a clothed woman. In a sense, it was an X-Ray app. Really sickening stuff that obviously angered many.

Both show the “user-friendliness” I was worried would come. 

Zao was taken down just three days after its release. And DeepNudes is no longer available either. The swift action in halting these deepfakes is great, but we cannot rely on reaction time as our defense.

The Latest Detections

Being the platform everyone feels deepfakes will spread on, Facebook has had to answer to this looming threat.

Facebook just announced that it’s launching a public race to develop technology for detecting deepfakes. The challenge, called the Deepfake Detection Challenge, will have a leaderboard and prizes.

As part of the challenge the company says it will release a dataset of faces and videos. The dataset will be built up by commissioning paid, consenting actors, and the company promises not to use any Facebook user data.

Samantha Cole, Vice

The challenge is starting in December 2019 and will run through 2020. They’ve teamed up with Microsoft and 7 academic institutions to run the challenge. And I must say that I’m interested to see how this one plays out. Especially considering there really isn’t a technological solution to deepfakes.

Not to mention, who doesn’t love a $10 million cash prize?

There are also a few companies that already exist in the periphery of this manipulated media movement.

Synthesia is serving businesses and organizations the ability to use video manipulation technology to scale their video production needs. Their David Beckham video manipulation campaign showcased just how deepfakes could be a great marketing tool, when we don’t have to worry about fake media.

Gigapixel AI uses machine learning to enhance low-resolution images, by understanding how pixels interact with their surrounding pixels.

Modulate is deepfakes for speech. Their intentions are not to provide more opportunities like the $243,000 heist above. Rather, Modulate wants to license its tech to social media and gaming companies, allowing you to have a cool audio avatar on these platforms but nowhere else outside them.

Meo can take 2D video of someone’s face and convert that into a 3D avatar. The avatar could be anything from an emoji to bear – conveying your emotional states and micro-expressions, like a subtle smile or scowl.

Regardless of the massive scare surrounding deepfakes, there will still be a lot of good to come from the technology. These are just a few of the early companies to capitalize. I’m predicting there will be hundreds more deepfake marketing agencies like Synthesia. And deepfake tools galore!