I enjoyed this paper taking down the current practices of "Human Centered AI" and a lot of AI hype generally. What Guest does here is reframe "AI" as any cognitive tool, and center the question of what relationship that tool has with the people it necessarily depends on. I appreciate the attempt to put new AI into context as "just another cognitive tool," to emphasize that AI always involves tapping into human cognition (in ways that are often obscured), and to point out it's not what we put into the AI that matters so much as how we use it in our lives. Basically, if you care about Human Centered AI, it doesn't matter if you train an LLM to manifest "human values" if you're just going to deploy it in ways that deskill the populace and eliminate jobs people depend on.