The key thing that AI boosters and doomsters have in common is that essentially they agree that this AI tech is fundamentally different in kind than what has gone before. It's not a matter of a computer being 5% faster at a particular problem, it's about the technology having crossed some sort of fundamental barrier beyond which all sorts of crazy things happen - conciousness, self-improvement, independent motivations etc. Now why would the founder of OpenAI admit to the possibility of it being dangerous? Because if you accept that it could be dangerous, you have to accept the premise that it's not just something a little better than before but that it really does have this different nature. And that different nature would - if it were true - make this AI technology enormously valuable. What would be really bad for the founder of OpenAI would be to admit that actually the technology isn't dangerous, because it can't do all those crazy things we imagine. It's the difference between self-driving and adaptive cruise control. And remember - the promised self-driving 5 years ago looks today a lot like it ended up just being adaptive cruise control.
I fundamentally disagree. Admitting danger dissuades people. Most want the thrill but not real danger. Placing the value in the danger aspect is like saying computer viruses are the most valuable thing in the world. I'm exaggerating a bit, but the way you put it sounds to me a bit much given the fact that Altman really doesn't have to convince anyone at this point that they have something very different.
