AI Challenges Human Writers in CAH’s Latest Black Friday Stunt
Cards Against Humanity – termed a party game for horrible people – has always been a fan of wacky Black Friday stunts that protest several issues the company sees as problematic. In the latest iteration of their Black Friday shenanigans, the company has pitted human writers against an AI-trained system to make expansion cards for the game. At the end of the period, if the AI had more sales, the company claimed it’ll fire their writers. If the writers had more sales, they would earn a five-thousand-dollar bonus. The commentary on how AI may replace human workers isn’t lost on the myriads of onlookers watching the stunt.
An AI Trained to be Horrible?
Cards Against Humanity is a game where players try to match answers with questions. The answers are usually some combination of humorous, tactless, morbid, or tasteless. To produce cards that fit what the company wants from its system, they pressed an open-source GPT2 system developed by OpenAI into service. The system’s training started by reading internet comments to create a sense of punctuation. It then read through all of the cards that already exist, including those that were never released publicly. Once the system was able to produce grammatically correct cards consistently, the creators stopped training.
Stiff Competition
The company live-streamed the event through the entirety of Black Friday, with the humans winning the competition by an extremely narrow margin. Cards Against Humanity’s site states that the humans won the day by about $1725 ahead of the AI competition. The gag is precisely what regular buyers and consumers of the company’s games would expect from them, but the competition raises some real-world, not-so-funny issues. With AI progressing at such a rate, it’s only a matter of time before they start replacing people in their jobs. For now, most people consider those creative roles, such as writing and art, may still be beyond the ability of AI to comprehend and mimic. The competition’s results show that humans may be wrong about their complete dominance in these fields.
