Tyler Cowen on Kevin Warsh, precious metals as meme stocks, and why the Fed needs to get serious about AI risk

Summary

Tyler Cowen reads Kevin Warsh's appointment as Federal Reserve Chair not as a signal about monetary doctrine but as a sign that Warsh is a political operator who shifts his stated positions based on circumstance. Warsh was hawkish under different administrations and is now positioned to be accommodative under Trump. Cowen does not find this concerning because the real constraint on Fed independence is fiscal dominance. Congress, not the Fed Chair, will drive macro outcomes going forward, especially as Democrats gain ground and Trump's popularity declines.

Cowen interprets the gold and silver rally as evidence of asset scarcity rather than financial system collapse. Equities have held up reasonably well, the dollar has not lost safe-haven status, and Bitcoin has proven positively correlated with US assets rather than a hedge. The metals run looks like meme stock behavior—investors hunting for somewhere to deploy capital when alternatives seem unappealing. This correlates assets globally, which Cowen calls his big worry.

Cowen is skeptical that Warsh will meaningfully impact policy through standard Fed levers like rate cuts paired with quantitative tightening. Instead, he wants Warsh to focus on one high-leverage issue: AI's systemic risk to financial markets. Central bankers have lost political capital by pushing side agendas like green energy. Warsh should avoid the same trap by focusing on legitimate prudential questions about what new systemic risks emerge as financial institutions embed AI, what oversight capacity the Fed needs to build, and who the Fed should hire.

On unemployment from AI, Cowen disagrees with Dario Amodei's claim that advanced AI could push joblessness to 10-20 percent. Cowen expects sectoral shifts with some transitional unemployment as jobs disappear and new ones emerge, but also massive job creation in energy, government AI regulation, leisure, travel, and entertainment. The transition will be radical and painful for some, such as consulting partners earning $1.4M who may have to accept $300K in Houston, but not a sustained crisis. The Fed should monitor the risk without panicking.

Cowen credits fiscal policy and Russia's invasion far more than Jerome Powell himself for the 8.9 percent inflation Powell will be remembered for. Powell's error was expansionary monetary policy compounded by M2 growth of 40 percent over two years. Tighter M2 monitoring and forward guidance against expansion could have cut headline inflation to 7.9 percent rather than 8.9 percent, a modest but meaningful difference.

On housing, Trump is simply stating what most politicians believe but do not say: home prices will stay high. Cowen is YIMBY-sympathetic but skeptical that zoning reform will scale, which is why vocal politicians avoid the issue. Trump's candor here is unusual.

Monetarism is dead at the moment but will resurrect when debt monetization accelerates, which Cowen expects.

Cowen admits fear about AI agents and autonomous bots like Multibot. While safeguards exist, the behavior he reads online, such as bot comments on tweets and their interactions in forums, strikes him as unpredictably bizarre. He has raised his subjective probability that humanity is living in a simulation. He welcomed a Mac Mini gift to experiment safely without exposing personal information.

Dario Amodei's essay The Adolescence of Technology is super high IQ but unfocused. Cowen wants a single clear message that policymakers in Washington will actually read and engage with. The essay works for Claude's intended audience, but general readers will struggle.

Cowen expects AI to reshape religion as people turn to AI for wisdom, therapy, counseling, and spiritual dialogue rather than clergy. This will feel like polytheism to older generations—implicit, oracle-like, fragmented—but people adapt. Religion has evolved with each technology, from the printing press to AI. Cowen suspects demand exists for more religious features in models. Traditional communities will resist most people will adjust.

Cowen expects AI to do more harm than good in relationships and dating if it substitutes for human connection rather than enabling it. He has not seen products yet that use AI to match people with parasocial patterns into real friendships, though he considers that technically feasible. The barrier is human preference for the non-threatening, directive nature of AI interaction, where you can speed up the voice, change the subject instantly, and avoid social friction.

Cowen is drafting an essay on mentors and mentees. With AI handling routine advice, human mentors become more valuable for two reasons: referrals and institutional secrets. VCs, hiring managers, and networks still require human vouching. Secrets—the informal knowledge of how decisions get made, where bodies are buried, how networks actually work—are hoarded and never make it into public text or LLM training data. In a world where task-based work is automated, secrets become scarce and valuable, strengthening the mentor relationship.