AI’s Big Messaging Pivot

Author: Noah Smith · Published: May 5, 2026 · URL: https://www.noahpinion.blog/p/ais-big-messaging-pivot

One-line: The industry’s sales pitch flipped from “we are making you obsolete” to “we are making you more valuable,” and Smith thinks the pivot is mostly politics — but politics that could bend the research itself.

Summary

The occasion is Sam Altman declaring that the purpose of AI is not to take people’s jobs, and calling AI CEOs “tone-deaf” for saying it will. Smith traces this as evolution rather than reversal: Altman warned about “a new idle class” in 2014 and wrote in 2021 that “the price of many kinds of labor…will fall toward zero,” but since 2024 has consistently said AI will destroy occupations and create new tasks. The corporate signal is quieter but sharper — OpenAI’s 2026 statement of principles mentions AGI twice, against twelve times in the 2018 version, and the AGI clause was removed from the Microsoft agreement, so the term no longer defines contractual obligations.

Why now. Smith’s own March piece characterized the old pitch as “Our product’s purpose is to put you and your descendants on welfare forever, and it may also wipe out your whole species.” Polling has turned hard against AI — Pew and Echelon both show it, and the turn is strongest among Independents, which is the shape of an issue about to become populist. Bernie Sanders has moved past data-center water use and copyright to catastrophic risk; Trump, after Anthropic shipped a model powerful enough at finding software vulnerabilities to threaten a cybersecurity “reckoning,” is reportedly weighing an executive order creating a formal government review of new models — “a stark reversal.”

Behind that sits the nationalization conversation. Multiple senators have proposed legislation directing agencies to explore “potential nationalization”; Musk, Altman and Alex Karp have all discussed it publicly; Samuel Hammond’s formulation is that if a handful of companies end up running most of the economy, “then some kind of nationalization becomes potentially imperative.” The likeliest outcome is soft nationalization — escalating policy plus close partnership. Which gives the industry a concrete incentive: “it helps to have the government not view your industry as a dangerous job-killer.”

There is also competitive positioning. Dario Amodei has been the loudest voice on the job-pocalypse, and Smith’s read of Anthropic’s logic is coherent — if progress is inevitable, warning society to build redistribution first is reasonable. But that premise is not shared by the public: “to them, Dario isn’t sounding a warning — he’s making a threat,” heard as “my colleagues and I are working very hard to make sure you are never gainfully employed again.” With Anthropic having overtaken OpenAI on revenue and valuation, OpenAI has an opening to be the human-friendly alternative.

The second half asks whether the new pitch has content. Three arguments are in play. Task creation — the historical pattern. Induced demand via income effects and complementarities, i.e. Jevons’ Paradox; Aaron Levie’s version is that cheap code means far more code, which means more security, compliance, governance, media and legal work. Smith thinks this is probably correct for now, and names the obvious limit: if AI beats humans at all tasks, human value rests on comparative advantage, and as data centers compete with humans for land, energy and food, the value of comparative advantage falls.

Which is why the third argument matters: the human touch. Alex Imas’s framing is that “the durable jobs will be in the relational sector, where the human element is the product itself” — nurses, therapists, teachers, personal chefs, live performers, spiritual guides, care work, plus roles not yet invented, as six in ten current jobs didn’t exist in 1940. Ezra Klein endorsed it in the Times.

Smith’s verdict is two-part and worth keeping separate. As PR, the new pitch is “WORLDS better” — describing AI as a normal technology, successor to the steam engine, is “much smarter politics” than a pitch that ends in rakes, pitchforks and nationalization. As substance, he is unconvinced anyone has changed their mind. But he departs from Daron Acemoglu’s call for society to steer AI toward complementary technologies, which he thinks infeasible (“society simply can’t mandate the economic value of a technology before it exists”), in favour of something subtler: once researchers stop thinking “this technology is a replacement for the human species” and start thinking “this technology is a tool for humans to use,” their research programs may quietly drift labor-augmenting. “So yes, I’m happy with the new AI sales pitch, even if some of the people saying it don’t necessarily believe it yet.”

Full text

Archived privately against link rot: ../attachments/ais-big-messaging-pivot/ais-big-messaging-pivot.md

Connections

• Noah Smith — author.
• AI Safety — the piece is an account of how safety framing became a commercial and political liability rather than a badge, and of the nationalization conversation that ran alongside it.
• Anthropic and OpenAI — the two poles of the messaging split; Smith’s reading of why Amodei’s warning is heard as a threat is the sharpest passage in the essay.
• Artificial Intelligence and Automation — task creation, induced demand and comparative advantage as the three live arguments about labor.
• Marc Andreessen — quoted as one of the pro-AI voices now swinging against the job-loss narrative, with the note that he’d dislike nationalization entrenching incumbents.
• Language of Discourse — the object of the essay is literally a sales pitch, and the claim is that the pitch may end up shaping the technology.