'Our team needs sleep': Sam Altman on ChatGPT's Ghibli image art frenzy

OpenAI CEO Sam Altman has acknowledged the surge in demand for ChatGPT-4o's new image generation feature, admitting that the company's infrastructure was under significant strain

OpenAI CEO Sam Altman has has said that ChatGPT users need to "chill out a bit," citing "insane" demand for its new image generator. In a post on X, Altman said, "Can yall please chill on generating images this is insane our team needs sleep (sic)."

 

Users, however, were not convinced with the appeal, with some even suggesting to fire his team. "No thanks. In addition to building AGI, this team is on trajectory to build the biggest website in the world from a cold start 2.33 years ago," Altman replied to the post.

 

On March 25, Altman announced the launch of ChatGPT’s latest AI-powered image generator, integrated with OpenAI’s flagship model GPT-4o. This feature enables iterative refinement, ensuring consistency across multiple interactions. 

 

 

Studio Ghibli mania

 

The introduction of this feature led to a surge in user engagement, with users experimenting by transforming photos into various artistic styles, such as Studio Ghibli. The rapid adoption of ChatGPT-4o's image generation feature resulted in unprecedented demand, causing strain on OpenAI's infrastructure.

 

“It’s super fun seeing people love images in ChatGPT, but our GPUs are melting,” Altman said. OpenAI also implemented restrictions on generating images in the styles of certain living artists and specific studios, including Studio Ghibli, due to copyright infringement.

 

He added that OpenAI would introduce temporary rate limits while working on optimising efficiency, assuring users that the restrictions would not last long. The free tier of ChatGPT will soon have a cap of three image generations per day.

 

How does OpenAI's new image generation tool work?

 

Unlike earlier AI-generated art that depended on external tools such as OpenAI’s DALL-E, ChatGPT-4o’s new feature is a built-in image generator. The feature allows users to create visuals seamlessly within the model itself using simple text prompts. Users can specify attributes such as aspect ratio, colour schemes using hex codes, or opt for transparent backgrounds with ease.

 

According to OpenAI, ChatGPT-4o can also analyse and learn from user-uploaded images, seamlessly integrating their details into its context to inform image generation.

 

“We trained our models on the joint distribution of online images and text, learning not just how images relate to language, but how they relate to each other,” OpenAI said in a blog post.