There’s a wonderful “what if” scenario piece over at https://ai-2027.com. It paints a picture of the paths future AI development might take. It feels very much like science fiction, but it might be right about timelines.
While a bit outlandish, but it combines international politics, competition, and compute capacity with the risks of AI misalignment to put together two plausible futures we should pay close attention to.
I enjoyed reading it.