>In a Nov. 18 letter filed in federal court, attorneys for The New York Times named Balaji as someone who had “unique and relevant documents” that would support their case against OpenAI. He was among at least 12 people — many of them past or present OpenAI employees — the newspaper had named in court filings as having material helpful to their case, ahead of depositions.
Yes it's true it's been public knowledge that OpenAI has trained on copyrighted data, but details about what was included in training data (albeit dated ...), as well as internal metrics (e.g. do they know how often their models regurgitate paragraphs from a training document?) would be important.
>In a Nov. 18 letter filed in federal court, attorneys for The New York Times named Balaji as someone who had “unique and relevant documents” that would support their case against OpenAI. He was among at least 12 people — many of them past or present OpenAI employees — the newspaper had named in court filings as having material helpful to their case, ahead of depositions.
Yes it's true it's been public knowledge that OpenAI has trained on copyrighted data, but details about what was included in training data (albeit dated ...), as well as internal metrics (e.g. do they know how often their models regurgitate paragraphs from a training document?) would be important.