> The lawsuit against OpenAI alleges that summaries of the plaintiffs’ work generated by ChatGPT indicate the bot was trained on their copyrighted content. “The summaries get some details wrong” but still show that ChatGPT “retains knowledge of particular works in the training dataset," the lawsuit says.
Setting aside the whole issue of whether LLM constitutes a derived work of whatever it's trained on, this sounds like a very weak argument to me. An LLM trained on numerous summaries of the works would also be capable of producing such summaries itself even if the works were never part of the training set. In general, having knowledge about something is not evidence of being trained on it.
Setting aside the whole issue of whether LLM constitutes a derived work of whatever it's trained on, this sounds like a very weak argument to me. An LLM trained on numerous summaries of the works would also be capable of producing such summaries itself even if the works were never part of the training set. In general, having knowledge about something is not evidence of being trained on it.