This post goes into a question I have pondered since the first legal challenges against OpenAI. As crazy as it might sound, "Could they be forced to start over?" If there are all these web-like connections between training content, can you make an AI forget specific content? Can you put legitimate guardrails on it after the fact? I highlighted earlier that the New York Times adjusted its policies so AIs cannot scrape their content. According to the post, "The result, experts speculate, could be devastating to OpenAI, including the destruction of ChatGPT's dataset and fines up to $150,000 per infringing piece of content." Given the potentially thousands or tens of thousands of infringing references that could indeed be a demise of OpenAI. Normally I'd say I suspect the NYT would simply follow in the footsteps of the Associated Press and craft a licensing deal with OpenAI, but if they fear that such a position would create a "competitor" by "creating text that answers questions based on the original reporting and writing of the paper's staff," then maybe not. And if OpenAI could be forced to start their large language module over again, wouldn't all of them, since they all behaved the same way? Be sure to read more at ars technica: Report: Potential NYT lawsuit could force OpenAI to wipe ChatGPT and start over
Published August 18, 2023.
