I agree with this assessment. I would speculate he did something in the early days to gain access to a bunch of training data under the guise of research ("...after all, we're OPENai") and used that data to fine-tune GPT-3 into chatGPT. Then once the weights were sufficiently good enough, deleted all the data and planned on using the chat interactions w/ the model itself for further refinement. Obviously, just total speculation, but the Cover Your Ass verbage of the board makes me think he did something to put the org in deep shit, legally. OpenAI suspended subscriptions last week and that's usually not something a company does, even if the service is degraded. Warn users, yes, but don't take any more money when you're hemorrhaging cash is off. I won't be surprised if it's a flagrant GDPR violation that carries very heavy fines.
OpenAI suspending subscriptions is especially suspect because they've dynamically altered GPT-4 usage limits many times before to handle increased load, and they didn't touch it at all last week before closing signups entirely.
