The Controversy Surrounding OpenAI’s Training Data
The issue of training data has become a contentious topic for OpenAI. The organization has been secretive about the training data used for models like GPT4o, leaving publishers with a difficult choice. Some publishers have decided to distance themselves from OpenAI, while others have chosen to make deals with the organization.
Training data for similar language models (LLMs) typically includes social media posts, blogs, digitized books, online reviews, Wikipedia pages, and other web-based information. However, there is speculation that LLMs have consumed a significant portion of the internet to replicate human intelligence. This raises concerns about privacy and control over the data.
One major concern for publishers is that OpenAI’s ChatGPT model seems to be fueled by stories published on their sites, including paywalled pages. This has led to a copyright dilemma, with publications like the New York Times filing lawsuits against OpenAI for copyright infringement. OpenAI argues that using publicly available internet materials for training AI models falls under fair use. However, this claim is met with skepticism, as some believe it could be a way to hide copyright infringement.
The Divide Among Media Companies
Media companies have divided themselves into two factions when it comes to dealing with OpenAI. Some publications have chosen to block OpenAI from accessing their content altogether, while others have struck licensing deals with the organization.
Those who have partnered with OpenAI argue that generative AI is here to stay and that it’s better to be part of the conversation than risk becoming obsolete. These partnerships also give publications some control over how their journalism surfaces in ChatGPT responses. Vox Media, for example, recently announced a licensing partnership with OpenAI and emphasized the importance of accurate and trustworthy information reaching the public.
However, critics argue that these deals amount to settling without litigation and trading credibility for profit. They believe that publishers are undervaluing themselves and their intellectual property by partnering with OpenAI. The Atlantic, one of the publications that signed a licensing agreement, referred to striking a deal as a “devil’s bargain.”
The Terms of Licensing Agreements
While OpenAI benefits from these licensing deals by gaining access to real-time news and displaying goodwill towards media, the terms of these agreements are not publicly known. It is unclear what exactly publishers receive in return for partnering with OpenAI. Many announcements mention access to reader data and insights as part of the exchange, suggesting that ChatGPT data plays a role in the agreements.
Successful Licensing Deals and Lawsuits
Several media companies have successfully partnered with OpenAI, including the Associated Press, Axel Springer Publications (Business Insider and Politico), FT Group (Financial Times), Dotdash Meredith Publications (People, Better Homes & Gardens, etc.), News Corp (The Wall Street Journal, New York Post, etc.), Vox Media, and The Atlantic.
On the other hand, some media companies have taken legal action against OpenAI for copyright infringement. The New York Times, The Intercept, Raw Story, AlterNet, and a collection of daily newspapers filed lawsuits against OpenAI and its major investor Microsoft.
The Future of OpenAI and Media Partnerships
The issue of training data and copyright infringement is still being worked out, and the outcome remains uncertain. While OpenAI claims fair use and freedom of information protection under U.S. copyright laws, critics argue that hiding copyright infringement behind the phrase “publicly available” is concerning.
Media companies face a difficult decision when it comes to partnering with OpenAI. They must weigh the benefits of access to generative AI technology against the potential loss of control over their intellectual property and credibility. As this story continues to develop, it will be interesting to see how the relationship between OpenAI and media companies evolves.
