www.noahpinion.blog/p/why-i-didnt-sign-the-we-must-act

This sentence overstates the statement’s vagueness. The text of We Must Act Now does not merely say "we must act" in the abstract. It explicitly says that economists, policymakers, and technology leaders must act "to understand the economics of transformative AI" and "to build the incentives, guardrails, and institutions needed to steer AI in a direction that complements humans and benefits society."

Those are still broad goals, and a reader may reasonably think they are underspecified. But it is factually incorrect to say the statement gives no indication of what action is being urged, or that it says only "we must act."

This sentence inaccurately describes what Acemoglu’s robot papers found.

Two of his best-known empirical robot papers do not support the simple claim that "companies that buy robots tend to hire fewer workers":

1. "Robots and Jobs: Evidence from US Labor Markets" is not a firm-level paper about what robot-buying companies do. It studies US local labor markets / commuting zones and estimates negative effects of robot exposure on employment and wages across commuting zones.

2. Acemoglu’s actual firm-level robot paper, "Competing with Robots: Firm-Level Evidence from France," says the opposite of Noah Smith’s summary on the firm itself: the abstract states that robot adopters "expand their overall employment as well". The paper’s negative employment result comes from effects on competitors and industry reallocation, not from adopters simply hiring fewer workers.

So the post compresses Acemoglu’s findings into a claim that his paper showed robot-buying firms reduce headcount, which is not what these papers say.