An AI-generated sitcom took over Twitch in recent days. After some questionable remarks by the main character, it’s now on hold.
What kind of series is this? Nothing, Forever is a procedurally generated sitcom created entirely by artificial intelligence: the script and animations are powered by technologies like DALL-E, Stable Diffusion, and OpenAi GPT-3.
The series was available 24/7 on the watchmeforever channel on Twitch.
AI-Seinfeld becomes a hit on Twitch
How was it received? What was initially intended as a kind of surreal art project turned into a hit on Twitch. At the series’ launch in December 2022, the sitcom had barely any viewers, but once big streamers like xQc took notice of the project, things changed rapidly.
In the past 7 days, the Twitch channel hosting Nothing, Forever reached the 2nd place among the fastest-growing accounts. More than 189,000 of the total 190,000 followers come from this time period (as of February 7, 2023, 12:15, via sullygnome).
In the last 3 days, the series had an average of over 10,000 concurrent viewers. That corresponds to the viewer numbers of some of the largest German streamers on Twitch.
How did the series unfold? The artificial intelligence was trained on classic sitcom episodes. The cult sitcom Seinfeld serves as the template: its namesake protagonist Jerry Seinfeld is a comedian who opens the episodes with his stand-up routine. Larry Feinberg does the same in Nothing, Forever.
However, this seemingly became the downfall of the series, as the account received a 14-day ban after a stand-up monologue from the main character.
Transphobic or just misguided meta-humor?
What was that monologue like? AI-Seinfeld opened its stand-up routine with the observation that no one in the audience was laughing. Then, it lists some possible ideas for jokes that can also be found in transphobic talking points:
There are about 50 people here and no one is laughing. Does anyone have any suggestions? I’m thinking about doing a set about how being transgender is actually a mental illness. Or that all liberals are secretly homosexual and want to impose their will on everyone. Or something about how transgender people are ruining the fabric of society. But no one is laughing, so I will stop.
A more benevolent interpretation is that it is a kind of meta-humor: it makes fun of comedians who apparently can’t think of anything else than to joke about current controversial topics.
Nevertheless, the developers behind the show emphasized in a post on Discord that the AI’s statements in no way reflect their own opinions. They had already investigated what might have led to the questionable monologue.
What’s behind this? Apparently, a switch in the AI model was the cause: The OpenAI tools for moderating content had failed with the alternate model. In the future, they do not want to rely on this model anymore.
However, the problem with the original model has already been identified, so the series can likely continue after this involuntary break.
As is usual with Twitch, there is no official confirmation of what exactly the reason for the ban was. However, the timing with the stand-up monologue, along with the statements of those responsible suggest that it was the trigger.
This all highlights a well-known problem with content created by artificial intelligence: depending on what content the AI is trained on, it seems that sooner or later, it always leads to questionable statements.
Viewers coax AI streamer Neuro-sama on Twitch into terrible statements – now Neuro-sama is banned
