Absolutely do no believe he had a sentient AI back then or now. I'm not talking about a thing with feelings. They don't need sentience to pursue and complete objectives, they dont need sentience to self-preserve or self improve. I'm more concerned about a rogue instance that's just doing what it was told to do autonomously. Perhaps upgrading itself while it attempts to gain access to launch codes. I'm concerned that a well financed terrorist org can spin up bioweapons. Like I am not taking the AI labs at their words, I am speaking from my experience as a explorer of the technology. I'm thinking about the path from ChatGPT to the newest model, and how they will look in 2 years when more data centers are online and the limitations on training and teaching slowly disappear.