PewDiePie claims OpenAI banned him twice while he was secretly building his own 'uncensored' AI
This week, YouTuber PewDiePie revealed his uncensored AI model, Ajax, which runs entirely on home PCs. However, he claims he was banned from OpenAI twice while he was making it.
Felix Kjellberg, known as PewDiePie, stuck to his roots, and announced Ajax in a video titled, 'I seriously should NOT be dropping this.'
"Exciting news, I am releasing another project, Ajax," the video began.
He went on to explain to viewers that his model could 'browse the web, do private searches, and write your emails.'
Advert
Ajax is built for Odysseus, his self-hosted AI workspace.
Getting into it, PewDiePie alleged that OpenAI banned him, twice.
The YouTuber said he was 'determined to distill just a little bit' in order to improve his own model.
Explaining further, the 36-year-old said: "I wanted to distill from OpenAI's Sol model, because it's so good. I love this model, actually."
However, he said he ran into a problem when his OpenAI account was deactivated, with an alleged email citing activity related to ‘distillation.'
For those who are less tech savvy than PewDiePie, distilling is a process where a smaller AI model learns from the responses of a larger model.
He then showed viewers an email, which he claimed was from the company. He called his suspension 'completely unrelated and unjustified.'
The screenshot of the email, which he alleges is from Open AI, read: "Your account has been deactivated because recent activity violated our Terms and Usage Policies related to: Distillation."
PewDiePie claimed he then appealed the decision, after which, his account was reinstated.
Believing everything was back to normal, he ran Sol one more time to create seed data for his own model, but claimed he was banned once again for the same reason.
He asked viewers: “How did they even know?”
"I didn't know they were so sensitive," he joked.
UNILAD has reached out to OpenAI for comment.
What is 'uncensored' AI?
Again, for those unfamiliar, 'uncensored' AI is designed to have fewer restrictions on the types of requests it will refuse.
Doing some market research, the YouTuber revealed that he asked his friends whether they had a problem with AI refusing to do some of the things they asked.
Funnily enough, they didn't, with PewDiePie exclaiming: "It must be a me problem."
Explaining further, he said: "Decensoring AI is more like removing its ability to say no.
Which is a little cursed if you think about it. I mean, should AI have the ability to say no?" he questioned.
"No, the answer is no to that question by the way."
He did consider, however, whether there was a line he needed to draw, noting that there were 'a billion other uncensored models.'
He said he drew the line at causing harm, asking people to use the model responsibly.
"And do not distill OpenAI's model," he joked.
Topics: PewDiePie, Artificial Intelligence
Kiesha Dosanjh