Even temporarily imitating D players for longer than a few unsourced articles on the internet might be a start until we start getting hand wavy about B players being threatened.

I'm honestly not even sure what world we are debating. Some hypothetical one in the future when AI fills some ideal trajectory we aren't even close to being capable of predicting in some near-term measurable timeframe?

I'm as pro-exponential tech trajectories as they come but I sometimes wonder whether we get lost in scifi novel thought experiments disconnected from reality because we want that to be real rather than having some tangible connection to today (and today's political implications).

"As for AIs taking over, that will come via the corporate route." What does this even mean practically in 2030 given what we know exists in 2023?

The cool thing is we can now stop relying solely on human imagination for ideating on futures funnels, and leverage GPT-4 to help us do it, and get a much more realistic sense of what types of things an AI system might conceivably attempt given we can probe these models for what is in them. Granted that's not perfect because GPT-N+1 will have emergent capabilities not present in GPT-N and so it's not always going to be a realistic assessment, but it's an ever closer approximation and it gives much more information than relying solely on human imagination.

I've been doing this a bit lately and despite the fact that I know it's only predicting the next token, I'm still slightly unnerved as I watch a black box computer system slowly type out novel and plausible ways to kill me.