OpenAI Implores Judge Not to Expose Communications by Its Top Researchers
Don't Speak
In one of the many copyright lawsuits it's facing, OpenAI is asking a judge to narrow the scope of discovery to limit insider communications from being aired in public.
This latest forte in OpenAI's defense against the Authors Guild, which is alleging on behalf of writers including George R.R. Martin and John Grisham that the company used its members' copyrighted works without permission to train its AI models, shows that the firm really doesn't want its top researchers' documents making the rounds.
The Guild recently requested that OpenAI submit extensive documents — including text messages and social media DMs — not only for the 24 initial "custodians" or insiders who are thought to have relevant pre-trial discovery information, but for eight additional figures as well, including OpenAI cofounder Ilya Sutskever and researcher Jan Lieke, both of whom left OpenAI this year.
Leaky Ship
While this counter-suit doesn't seem to be specifically about inside baseball at OpenAI, recent news involving Sutskever highlights why the company may not want its former and current executives' communications leaked.
Earlier this week, the sidelined cofounder told Reuters that the abilities of OpenAI's advanced large language models (LLMs) seem to be plateauing as the company seeks more and more training data and computing resources.
Sutskever was, notably, the leading proponent of the briefly successful attempt to oust CEO Sam Altman last Thanksgiving, and spent half a year in limbo before ultimately leaving the firm and starting his own venture pursuing safe artificial general intelligence (AGI).
Should his texts and DMs come out, even if just regarding copyrighted work, there could be some very embarrassing details in there — which could be at least part of why OpenAI is trying to make sure they never see the light of day.
More on OpenAI: OpenAI Buys Porn Domain for Huge Sum