The line between offloading work to AI and offloading judgment

Yennie Jun, who works on measuring model capabilities, writes about a distinction that gets lost in most conversations about AI and productivity: the difference between offloading work and offloading judgment. Some of what people hand to AI is genuine drudgery, and giving it away costs nothing. She is more worried about a quieter habit, where people stop forming their own views and let a model decide. She describes meeting a man who records his daily conversations to feed them to an AI, because he has concluded the system thinks better than he does.

Her frame is a question: what exactly are we automating, human work or human agency? She points to research showing how far the first kind has come, including METR's finding that models can now handle tasks that used to take people hours or days from start to finish. The essay does not argue against using AI. It argues for noticing which decisions you are quietly delegating. Her reference point is Ken Liu's 2012 story about an assistant that makes every small life choice for its user until there is nothing left to choose.

Why it matters

The useful move here is personal and practical. Before you hand a task to a model, ask whether you are saving effort or outsourcing a judgment you would rather keep. The first is free. The second adds up over time.