In order to write, lead advertising campaigns, and power side hustles AI needs training material. ChatGPT needed about 300 billion words to get off the ground and continues to train itself based on how users interact with it.
However, human beings aren’t being credited or compensated for creating the content that AI is eating up. Authors, artists, and news organizations have already filed countless copyright lawsuits against AI giants like OpenAI and Microsoft as they find that AI bots can talk about their copyrighted work “too accurately” — indicating that the works are in the AI’s training data.
That’s why Microsoft’s AI CEO Mustafa Suleyman was asked at the Aspen Ideas Festival in late June if AI companies have essentially stolen the world’s intellectual property.
Suleyman’s answer? Almost all content on the Internet, with one possible exception, is fair game for AI training.
Related: A Microsoft-Partnered AI Startup Is Being Sued By the Biggest Record Labels in the World
“I think that with respect to content that is already on the open web, the social contract of that content since the ’90s has been that it is fair use,” Suleyman said.
Suleyman stated that “anyone” can copy or recreate the content on the open web.
“That has been freeway,” he said. “That’s been the understanding.”
However, some news sites and publishers have asked not to be scraped or crawled.
“That’s the gray area and I think that’s going to work its way through the courts,” Suleyman said.
Mustafa Suleyman. Photographer: Stefan Wermuth/Bloomberg via Getty Images
Suleyman leads Microsoft AI at a time when Microsoft has invested billions into the technology. His position on what is fair use and what isn’t fleshes out how AI companies might defend intellectual property allegations in court.
OpenAI, for example, has allegedly used more than a million hours of YouTube videos to train ChatGPT. When asked whether YouTube or social media videos were used to make OpenAI’s video generator Sora, the company’s chief technology officer Mira Murati said, “We used publicly available data and licensed data” and wouldn’t specify further.
AI also appears to be eating work generated by other AI, resulting in lower-quality output. Experts estimate that 90% of online content will be AI-generated within the next two years.
Related: The Most Downloaded News App in the U.S. May Have Published Dozens of Fake, AI-Written Stories