i'm also wondering whether there are confounding factors like more domain experts and talents joining the AI research field because there are more interesting problems to be solved (or other 2nd order effects of AI models being more powerful).