[From ‘COPYCATS’. See Chapter 1 here.]
Tiffany Belter was a teen streaming sensation in the early forties. She did makeup tutorials, filmed POV reality segments in VR, and hosted live karaoke session attended by millions. At her peak, in April 2043, she was the most popular weekly series worldwide for girls 12-14. The big beauty brands competed for her weekly review; a contract with California’s largest retailer was in the works. She was likely, one pundit predicted, to be the first female streamer to win the Presidency.
On April 12, 2043, Tiffany didn’t post her daily update. There was no warning sign on her feeds; the video just didn’t show up when it was supposed to. Millions of tweens refreshed their browsers to no avail. #WheresTiffany went to the top of the trending lists in a few hours. Eventually, she responded:
Sorry, TiffHive, I actually got into a car accident today! OMG super scary but I’ll be back tomorrow to keep you updated!
This simple statement would go on to be taught at PhD programs and discussed in the hallways of Sacramento, Tokyo, and New Cairo. It was the first known public statement by a COPYCAT.
When Tiffany started posting again, her content wasn’t as good. Fans were gentle, at first—she had been in a car crash, after all!—but the content just didn’t improve. It wasn’t terrible, but it seemed like Tiffany had gotten lazy—she was rehashing old ideas in a new light, she didn’t have that spark she used to be known for. She got the retailer contract—the dinosaurs were always slow to see when a trend was fading—but she was losing viewers every week.
The diehards started to see strange patterns. Tiffany started repeating herself more and more frequently. Sometimes she’d use the same phrase twice in one stream. Her appearance had gotten worse, and the quality of her vides got grainier and grainier until it looked like an amateur on a webcam. Fans unfollowed in droves.
That might have been the end of it. There are, after all, plenty of feeds on the web, and Tiffany’s rise and fall was hardly unprecedented. But in late May, a neighbor in her apartment complex called in a noise complaint—a loud popping sound, like a jackhammer, he reported. By the time police arrived, they could smell smoke, and they forced the door. The room was empty save for a small cube of burnt plastic sitting in front of the desktop computer. The apartment hadn’t been lived in for weeks.
Tiffany Belter was never seen again.
The V1 COPYCATS were network infiltrators, machines trained to mimic humans. They took over silently, and if they were good, they could evade detection for a very long time. That person you were texting acting strange? That feed you were streaming not as good as usual? Maybe you’re just being paranoid. Maybe not.
Whoever built the COPYCATS had learned their lessons from the failure of the basic bots of the early days of the internet. Those early models were dumb, repeating and sharing the same slogans and memes over and over. Networks got better at detection as machine learning got better—it wasn’t too hard to build a model that spotted fraud on that scale. Users figured out how to spot bots, too, neutering their effectiveness. Before long the bots stopped being used at all.
COPYCATS escalated the situation. Instead of using high numbers, they focused on small scale, targeted intervention, replacing useful influencers, and later, individuals in a position to achieve their goals. Tiffany Belter had been a failure—she was an obvious imposter. A successful COPYCAT was one that simply kept posting.
How did it work? COPYCATS were sponges, basically. A complex suite of machine learning models was trained on the electronic communication of the target. Texts, posts, videos. Eventually, the machine acquired the right tone to imitate successfully.
Who was behind it? The Chinese blamed the Americans, the Americans blamed the Chinese, and thousands of influencers in both countries vanished over the course of the next five years.