Microsoft director: AI scraping 'the largest theft of labor in human history'

A top Microsoft AI executive warned that the training of large language models could be the greatest labor theft in human history, according to newly released court documents. The comments, made by Brent Hecht, Microsoft's director of applied science, suggest that the practice of training AI on large amounts of material from the open internet could be viewed as an unprecedented scale of theft by people worldwide.

Hecht's remarks come amid a copyright lawsuit brought by news outlets, including The New York Times, against AI companies like OpenAI and Microsoft. The lawsuit argues that training AI models on copyrighted material, such as articles from The New York Times, does not fall under fair use. Microsoft and OpenAI maintain that their use of such material is permissible under fair use, similar to a student reading books.

However, internal comments from OpenAI executives, such as Greg Brockman, have suggested that AI models can reproduce news articles, potentially undermining the argument of fair use. The controversy surrounding AI training methods also extends to other AI companies, such as Meta and Twitch, which have been training their models on user data, articles, and other material.

The impact of generative AI on the news business model is another key issue in the lawsuit, with AI-generated summaries potentially discouraging readers from visiting news websites.