Fiction

Deeper than Fake

The Challenger — Part II

“Okay. I’m seriously impressed!” For the first time in weeks, Amy had a smile on her face.

Overnight, Zeke had made a four-minute video of the Candidate delivering his weekly video update. At least, Zeke’s mysterious Challenger program had made the video while everyone slept.

The campaign team planned to spend the day with the Candidate, cajoling him to stay focussed and on-message as his energy flagged. Now, it seemed, Challenger had saved them a day’s work, and possibly earned their flagging candidate — and the rest of the team — a day’s rest.

“I had a look at deepfakes last night. It looks to me like most of it is revenge porn, and trolling celebrities,” Amy offered.

“That’s true,” Zeke jumped in. “At least it’s true of the deepfakes we know about. Obama, Nancy Pelosi, Mark Zuckerberg, and a lot of jilted boys putting the faces of women who scorned them onto porn scenes. But as always, we don’t know what we don’t know. I’m sure there are videos out there that haven’t come to light yet. Unfortunately, for every person who says cries ‘deepfake’ honestly, there are several who are just using it to weasel out of saying something they regret.”

“Like Trump claiming he eradicated AIDS?” Mitch offered.

“No. That was a deepfake. And not a very good one. A French charity created it as a clever cut-through, calling it ‘the first fake news that could be true’,” explained Zeke. Trump evidently hasn’t ever regretted anything he said enough to want to use this particular out.”

“What about Ali Bongo?” Mitch had clearly spent a bit of time researching deepfakes too. “Some Gabonese army officers tried a coup in early 2019, claiming that President Ali Bongo was dead. He hadn’t been seen in months, and then he appeared in a bizarre televised address. Social media went nuts, saying it was a deepfake, and that led some opportunists to try stage a coup.”

“The Ali Bongo incident is an example of how deepfakes are changing the landscape,” Zeke admitted. “Ali Bongo has been sick, and it’s likely he suffered a stroke while on a state visit. But he’s been seen since, and the general consensus is that the video was legit, but maybe he had been drugged and botoxed a bit too much before recording it.”

“What happened with Ali Bongo nearly threw West Africa’s most stable country into civil war!” Mitch clearly hadn’t had a peaceful night’s rest. “Imagine what would happen here.”

Amy seemed a bit more open to the possibilities. “Zeke, what are the odds that somebody is going to figure out that this video is a Deepfake?”

“One hundred percent,” Zeke replied without a pause. “One day the technology for detecting Deepfakes and for scrutinizing video integrity will definitely be good enough. That won’t happen for another ten to twenty years, though. Challenger is so advanced that, barring a leak from this team, there is a zero percent chance of anybody figuring this out before the inauguration next January.”

“What makes you so sure?” The worried look had crept back across Amy’s face.

“For one thing, nobody knows what Challenger can do, or how it does it.” Zeke explained, his slow speech not disguising his paternal pride in Challenger. “Most deepfakes are made by a single generator network that makes new content, pitted against a single discriminator network that looks to detect if the new content is any worse than the real content it learns on. Challenger currently comprises eight generator networks. They compete to produce the best content, and then they share parts of their code to improve their efficiency in making content.”

“On the other side, another eight discriminator networks work to distinguish fakes from the real deal. Four of them work like regular discriminator networks, some working only on one aspect of the video and others looking at the integration between the person, background, movement, and speech. The other four apply the best that deepfake detection technology currently has to offer, identifying red flags and sending what they learn back to the generators.”

“Okay. There’s no doubt that Challenger is good,” Amy clearly seemed to have reached a decision. “I think we should do this. Our candidate clearly needs a day off right now. The time and rest this buys him are going to be more valuable than the video itself.”

“You’re the campaign manager,” Mitch reluctantly fell into line. “Just know that I’m not happy. I think the voters deserve to know who and what they are voting for. I think this is the thin end of a massive democracy-wrecking wedge. If the other side were doing it, and I knew, I would burn them to the fucking ground.”

“Okay, Mitch. I’m going to save you the embarrassment if this thing blows up later on. This is my call. As far as I’m concerned, making a better video of the speech he would have given — or we would have wanted him to give — is not wildly different from shooting the most flattering real video we can.”

“Anyway,” Amy offered, “it’s not like the people won’t have plenty of chance to see our candidate over the coming weeks. AI can’t help us when we are out on the campaign trail, interacting with real people in real-time.”

Zeke didn’t dare look up from his computer screen. Now would not be the right time to tell Amy or Mitch about the next steps he had in mind.

If you missed Part 1, then check it out here.