Wednesday 8 February 2023

Twitch suspends AI-generated Seinfeld show for 'transphobic' jokes

 Streaming platform Twitch has suspended a Seinfeld spoof, completely generated but artificial intelligence, for scenes in which the AI Jerry Seinfeld tells jokes that are "transphobic" and call liberals "secretly gay," according to Kotaku.

Twitch's 24-hour livestreaming channel called WatchMeForever was suspended for two weeks over the show called "Nothing, Forever," which is completely AI-driven, using machine-learning platforms such as Stable DiffusionDALL-E, and OpenAI GPT-3.

Dialogue and parodies of characters from Seinfeld are generated to create storylines, with character names such as Larry Feinberg (Jerry Seinfeld), Yvonne (Elaine), Fred (George), and Zoltan (Kramer.)

Typical scenes of the show place the characters in a New York City apartment or a stand-up comedy club.

The latter would appear to be where everything went wrong.

In a clip circulating online, the Seinfeld-esque character performs the following routine in a comedy club:

"So, this is my stand-up set in a club. There's like 50 people here, and no one is laughing, anyone have any suggestions?" Larry Feinberg asks. 

"I'm thinking about doing a bit about how being transgender is actually a mental illness. Or how all liberals are secretly gay 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'm going to stop. Thanks for coming out tonight, see you next time. Where'd everybody go? Oh, right."

The show's co-creator was "super embarrassed" and blamed a lack of "moderation" in the AI-generated script model, calling the dialogue "inappropriate."

“We thought we had solved for this problem—we use a built-in content moderation system provided by OpenAI—but clearly we hadn’t,” said Skyler Hartle in an email exchange with Kotaku.

"We are currently investigating how we can implement a secondary content moderation system to have an extra layer of redundancy to ensure this doesn’t happen again. We mistakenly believed we were correctly leveraging OpenAI’s tools for content moderation, but that wasn’t the case. We are planning to implement OpenAI’s content moderation systems before going live again, in addition to looking at services for secondary content moderation as a redundancy," Hartle explained.

Hartle told the tech outlet that the show creators have been "investigating the root cause of the issue," which was allegedly caused by switching to a different AI text generator as a backup.

When their preferred program, OpenAI GPT-3, wasn't working, the creators would switch to a program called Curie to ensure the livestream of their show never stopped.

"We hope this sheds a little light on how this happened,” Hartle said.

The co-creator also said the views expressed in the AI show don't reflect his company's values or opinions. 

Post a Comment

Start typing and press Enter to search