Update March 23, 2024:
The New York Times has responded to Microsoft's motion to dismiss its lawsuit against OpenAI over the development and use of ChatGPT.
It argues that an LLM is not like a VCR because a language model is trained with copyrighted works and can then produce content that competes with the original works.
"If VCRs had been built with movies to make movies that compete with movies, or if Sony oversaw the VCR's infringing users, [the Betamax case] would have gone the other way."
The Times argues that it doesn't stand in the way of technological progress. Instead, Microsoft could have trained the LLM using public domain works, but it chose to use copyrighted content for free.
"Microsoft - the world's richest company - claims it has the right to take what it wants for free."
OpenAI recently stated that training the best AI models is impossible without copyrighted data.
Via: JDSupra
Original article from March 5, 2024:
Microsoft compares New York Times to '80s movie studios trying to ban VCRs
The legal battle between the New York Times and OpenAI over copyright infringement could be groundbreaking for generative AI. And it promises to be entertaining.
A motion to dismiss filed by Microsoft gives us a foretaste: the major OpenAI investor accuses the New York Times of "doomsday futurology," the Financial Times reports. Microsoft presumably means overly pessimistic or alarmist predictions about the future.
The company argues that copyright law is not an obstacle for ChatGPT because the content used to train the tool would not displace the market for that content.
"Content used to train LLMs does not supplant the market for the works, it teaches the models language. It is the essence of what copyright law considers transformative—and fair— use."
Microsoft draws an analogy, comparing the New York Times to the Hollywood studios that tried to prevent the introduction of the VCR, also a "groundbreaking new technology," in the 1980s.
The studios believed that the VCR would violate their copyrights and eliminate the market for their content. What changed was access to the content.
"The entertainment industry flourished when the VCR opened new markets and revenue streams. But at the time, empty warnings nearly won the day."
The NYT's lawyer argues that VCR manufacturers never claimed that they had to commit massive copyright infringement to make their products. OpenAI recently admitted that current AI models cannot be trained without copyrighted data.
Microsoft is OpenAI's largest supporter, pledging up to $13 billion in investment. Microsoft also provides OpenAI with computing power and has exclusive distribution rights for OpenAI's AI models. Microsoft will receive up to 49 percent of the profits.
Similar to Microsoft, OpenAI argues that "ChatGPT is not in any way a substitute for a subscription to The New York Times" and that it is not possible to reproduce NYT articles with normal use of ChatGPT.
Furthermore, while NYT articles were used to train ChatGPT, they did not contribute significantly to the training of the models, OpenAI claims.