zlacker

>>davidb+(OP)
pure speculation: ChatGPT training dataset contains massive amounts of copyrighted material, and he told the board it didn't. now there's a big looming liability.

>>crunky+Jh
Not possible. It’s well known that the training data was not public domain. OpenAI was in the process of securing private agreements for more training data. They likely trained GPT-4 on every word ever written in a book, newspaper, website or social media network.

Given the wording and the apparent stepping down of the board chair, this is very likely a conflict between the grow at all costs startup folks and the AGI true believers. Sam was likely going to fully sell out and make himself filthy rich. The AGI cultists wanted to retain control over their Frankenstein’s monster.