Series created by artificial intelligence is closed due to scandal after two months of continuous broadcast

Vice journalists wrote about GPT-3's complex creative path in show business. The Nothing, Forever series, created entirely by AI, existed on Twitch for about two months, gathered a large audience, and was banned for violating moral standards.

«Nothing, Forever» is a peculiar interpretation of the popular sitcom of the 90s of the last century «Seinfeld». Its ideologues are Skyler Hart and Brian Gabersberger. The author is Davinci, the latest and most powerful version of GPT-3.

The show lasted 24 hours a day and consisted of strange dialogues of pixel characters in one apartment. Interestingly, at the beginning of the «sitcom», on December 14, 2022, only 16 people watched the show. But then the series went viral: thousands of viewers fell under the «supernatural charm of the low-resolution 3D rendering of the four main characters.» The sitcom characters moved around awkwardly, walked through furniture, and made some bad jokes. The channel where Nothing, Forever was broadcast, watchmeforever, eventually gained about 190 thousand subscribers. And the organizers even managed to open an online store of branded merchandise.

However, the show had to be closed when the AI started behaving inappropriately on February 5. Trying not to interrupt the stream, the organizers decided to switch to Davinci’s predecessor, Curie. Only a few minutes after that, one of the characters launched into an obscene tirade.

on February 6, Nothing, Forever was banned from Twitch for 14 days. The show’s organizers decided to improve Davinci’s creative skills and apologized to the audience.

«We are very sorry for what happened. The comments and inappropriate content in no way reflect the views of our employees. We hope that our new security mechanisms will help prevent such situations in the future,» they said.

In the wake of the popularity of chatbots, people are prematurely expecting «creativity» from artificial intelligence. Skyler Hartle, one of the ideologists of Nothing, Forever, reminded once again: «AI models lack the ability to generate something that has never been said before, so as a result, everything you hear has probably been said by someone somewhere at least once.»

Currently, the organizers are improving the scripting and directing skills of the language model so that it does not go beyond the limits of morality. As Hartle assured the journalists, they «will try to maintain the unpredictability and generative nature of the show, but will try to eliminate the possibility of inappropriate remarks.»