There is a new and excellent post by titotal, here is one excerpt:
The AI 2027 have picked one very narrow slice of the possibility space, and have built up their model based on that. There’s nothing wrong with doing that, as long as you’re very clear that’s what you’re doing. But if you…