OpenAI and its “love” for Journalism.

OpenAI discusses its “love” for Journalism. The latest post from the AI company presents their position on training AI models on copyrighted content like news articles in response to a recent lawsuit from The New York Times.

What's going on here?

OpenAI believes training models on copyrighted content is fair use but provides their POV on the NYT lawsuit.

What does this mean?

OpenAI gives out 4 points to make their case:

• Collaboration with news organizations: OpenAI has recently partnered with AP, Alex Springer, American Journalism Project and NYU. In return, they are helping reporters and editors with time-consuming tasks.

• Training is fair use. OpenAI stands strong on this claim by linking to support comments and ruling in other countries.

• “Regurgitation” is a rare bug. Exact memorization happens when the same content is in the dataset multiple times. OpenAI is working to solve this, and claims NYT’s small footprint in the training material doesn’t matter.

• NYT is acting in bad faith. OpenAI and NYT were in talks before the court case and it claims that NYT didn’t inform OpenAI about this move. OpenAI makes its defense attack, stating that the examples shown by New York Times are cherry-picked, created by manipulated prompts with excerpts of those articles, and not an “allowed” use according to OpenAI’s terms anyway.

Why should I care?

Throughout, OpenAI pats its back with examples of their actions like linking to fetched articles, providing opt-out options, and removing the browsing feature in July.

Brilliant PR move from Open AI. The statement positions OpenAI as a “friend of journalists and NYT’s move as bad faith, anti-technology move. Will OpenAi be able to do so in court though?