Silicon Valley's best-known technology optimist just told the Federal Reserve that the AI revolution may never reach the parts of the economy Americans complain about most.
Venture capitalist Marc Andreessen used his closing slot at the San Francisco Fed's AI conference yesterday to deliver a warning that cuts against his own reputation: both the utopians and the doomers are going to be disappointed, most of all by AI's real-world economic impact.
Andreessen co-leads the central bank's Productivity and Jobs Task Force, a panel Fed Chairman Kevin Warsh created to assess what super intelligence will do to growth, prices and jobs. The group is advisory and carries no vote on interest rates, with recommendations due by year-end and the Fed's next rate decision set for December 9.
His argument leans on a decades-old idea, Baumol's cost disease, the reason flat-screen TVs get cheaper every year while hospital bills, tuition and rent climb. Andreessen splits the economy in two. Fast sectors like consumer electronics and media adopt new technology quickly, and their prices fall. Slow sectors like health care, education, housing, and law and government are, in his words, not legally allowed to have rapid change. A machine, he noted, cannot get an MD or be admitted to the bar.
That divide matters because the cheap fast sectors shrink as a share of the economy while their prices crash, leaving the slow, expensive sectors to take over. Health care is already about a fifth of US GDP, and Andreessen sees no reason it cannot climb toward half or more. He framed the result as a mixed blessing: the jobs may be safe, but productivity and economic growth may simply stagnate.
He is normally the man cheering the opposite. Andreessen's 2023 Techno-Optimist Manifesto declared, We believe in accelerationism, and named stagnation as the enemy. Yet at the Fed, the country's most famous accelerationist became the one suggesting stagnation might win, and that the decisions about how to deploy AI will matter more than what the technology is capable of.