I can see where @jasonfurman is coming from, but I also agree with Claudia, in the particular case of CEA. From Furman’s probable perspective (my speculation)- a senior person’s day is usually chalked full with meetings and multi activities that require constant context switching, leaving little time for playing with data yourself and a dependence on support staff / economists to process the data. The ability to now not depend on anyone, as a senior person, is very very empowering. Now, with just the idea, a senior person can go far. On top of that, the dopamine hit from Claude is real. After years of needing research assistants, you suddenly realize you don’t need to ask people for things, or to wait for answers. Got questions? Got answers NOW, exactly answered in the way you want it. How do I know? It so happens that recently my nightly activities have been spent on building tools (for myself) on how I like to visualize economics. I am probably just a month behind Furman. But now to Claudia’s point — CEA is a very important institution in the White House. Among the many things it does, it fact checks important economic speeches of White House characters. It tells politicians they are wrong in a place where people tend to say yes. AI requires fact checking — in the high stakes world of POTUS speeches, you can’t even allow a 0.05% chance of an AI hallucination on it. Second, CEA briefs principals confidentially ahead of data releases due to market sensitivity. These releases are sent in via physical paper in manila folders. With the cyber security concerns around AI, it is inconceivable that AI can substitute for the human required to do the job. Third, to Claudia, CEA spawned a lot of people who go on to do things. I am one of them. So is @tdurie95 in my team. Without that experience, I can 100% say would not have the kind of world view that I have now. Which opens an interesting question about AI — in the productivity literature about GPT (general purpose technology), that type of innovation is supposed to spawn more innovation in a nonlinear way. But whatif AI’s productivity impact is actually front loaded? (Ppl who follows my writings would know that I insist that labor productivity impact from AI is already showing up strong in earnings, though absent in national statistics, which i believe will be revised up years later). AI allows people with extensive experience to scale their expertise. But AI also raised the entry barrier for any white collar jobs, slowing the supply pipeline of human experts. As the current generation of experts age, the size of the new generation of experts was not enough to replace the previous generation of experts. As a result the spawn rate of ideas slow. Entering into sci fi speculation— as humans were not about to replace the aging experts due to low pipeline of young human experts, AI themselves, through the training over the years, became the one spawning the ideas. That seems like a disaster waiting to happen that current social policy needs to address.