Congress Could Kill AI Stocks Before AI Kills Us
Submitted by QTR's Fringe Finance
Yesterday I wrote that we may only get one chance to stop AI. Today, Washington appears to be moving a little closer to testing that proposition. The irony is that if lawmakers actually decide to meaningfully slow the development of frontier AI, they could also wind up disrupting one of the most important investment themes holding up the stock market.
My argument yesterday was fairly straightforward. AI capabilities are advancing faster than our ability to confidently understand or control them, and the consequences of getting something wrong increase as the systems become more powerful and autonomous. None of the strange behavior we’ve seen from AI systems proves we’re on the doorstep of catastrophe. But it does raise a question that becomes harder to ignore as capabilities improve: at what point does waiting for definitive proof of danger mean waiting until it’s too late to do anything useful about it?
The asymmetry is what bothers me. If the AI pessimists are wrong and society imposes unnecessary restrictions, we sacrifice some technological progress and economic growth. That’s a real cost. But if they’re right and sufficiently powerful AI eventually becomes difficult or impossible to control, the cost of moving too slowly could be massive. Once the technology exists and has proliferated, there may not be a practical way to put it back in the box.
My piece was spurred by this week’s drama at Anthropic. Researcher Jacob Coxon resigned and publicly warned about the race toward increasingly powerful AI, arguing in effect that the major labs are caught in a competition none of them feels able to abandon.
The disturbing part wasn’t merely that a departing researcher was worried. It was that his concerns echoed fears already held by people working directly on the problem. Anthropic alignment researcher Evan Hubinger subsequently said he assigns a greater than 10% chance to AI causing human extinction within the next decade.
Whatever probability you personally assign to that outcome, it is an extraordinary situation. People working at the frontier of this technology are simultaneously helping make the systems more capable and warning that nobody yet knows how to guarantee control over what comes next.
As I noted yesterday, the incentives make the problem worse: any individual company that slows down risks surrendering ground to competitors, while the industry as a whole has every reason to keep accelerating. It is the kind of coordination problem that eventually invites government intervention.
And, almost on cue, that intervention may be getting closer. Semafor reported this morning that Sens. Amy Klobuchar, Ted Cruz and Senate Majority Leader John Thune are working on bipartisan AI safety legislation that could be introduced soon. The details aren’t public, so it would be premature to assume how restrictive the final proposal might be. But the political direction is extremely notable....(READ THIS FULL COLUMN 100% FREE HERE).