• Timothy Lee tries to reduce your AI-induced anxiety in his substack post Why I’m not afraid of superintelligent AI taking over the world. Basically, he posits that preexisting knowledge, not tokens/training data volume, is a limiting factor, both because of its finite, non fungible nature. It’s definitely worth reading, though I found it only mildly reassuring. Someone will eventually allow AI to test hypotheses in the real world ad lib, and then it will likely generate its own new knowledge based on empirical results.