The promise sounds automatic
The promise is appealing. Give people better tools, let them produce more, and everyone becomes better off. I want the first part to happen. Wasted effort is not a noble tradition that needs defending. But the movement from more output to a better life contains decisions that the language of efficiency often makes disappear.
Labour productivity measures output per hour. It does not, by itself, tell us how the resulting income will be divided or what a particular worker will be able to negotiate. A larger pie still leaves a question about who is holding the knife. I think discussions of automation become evasive when they treat that question as an unfortunate distraction from the exciting technical part.
Measure the gap carefully
It is also easy to make the opposite mistake and use a chart as a shortcut to a complete political explanation. The BLS discussion of productivity and compensation points out that comparisons depend partly on different price measures and partly on changes in labour’s share of output. Wages and total compensation are not interchangeable measures either.
Those distinctions matter. If I care about who benefits from growth, I should want the measurement to be careful. A dramatic picture built from incompatible quantities does not become a better argument because I sympathize with its conclusion. Getting the accounting right helps identify which question still needs a political answer.
The question I care about is what gives a worker the ability to claim part of an improvement. Can they change jobs without an intolerable risk? Do they know what comparable work pays? Can they negotiate collectively? Who owns the tools and controls the process? These are institutional questions, and a faster piece of software cannot answer them on its own.
Efficiency with a destination
Imagine a tool that cuts the time needed for a task in half. The organization could increase output, reduce working hours, raise pay, lower prices, or distribute the gains through some combination of those choices. The technical achievement creates possibilities. It does not issue a command selecting the fairest one.
That is why I want promises about AI productivity to include a sentence about distribution. Not a vague assurance that benefits will eventually spread, a concrete account of the arrangements through which people will receive them. Training, ownership, competition, bargaining, and public policy deserve to be in the same conversation as model capability.
I am in favor of making useful work easier. I am less impressed by a future in which everyone becomes more productive and still has to plead for enough time to live. The point of efficiency should include a question about the life it enables. Otherwise we have improved the machine while leaving its destination suspiciously undefined.