Why AI Superintelligence Risks Are Impossible to Price

The Economics of Frontier AI Risk and Market Accountability

Frontier AI laboratories face internal concerns that advanced models might assist rogue states or terrorist groups in building bioweapons, while extreme scenarios touch on human extinction, according to coverage from the Economist. Economist Tyler Cowen suggests that alarmists predicting a catastrophe should put their money behind their theories by making financial trades, arguing that successful trades would add credibility to prophecies of doom well in advance of a disaster. His colleague Alex Tabarrok previously described such bets as a tax on bullshit. Yet, placing those financial bets remains difficult when dealing with advanced artificial intelligence.

Cyber-Security Stocks Surge Amid AI Threats

Some milder downsides of powerful technology are already priced into equities, as concerns over AI-enabled hacking lift the value of cyber-security providers like CrowdStrike, according to financial reporting. CrowdStrike shares have more than doubled over a six-month period, pushing its price to nearly 200 times its forecast earnings for the next twelve months, demonstrating that fear benefits certain sectors. For adventurous worriers seeking untested strategies, shorting vulnerable industries like health-care, finance, telecoms, and transport, or buying credit-default swaps against regional bank debt, offers potential avenues.

Insurance Industry Struggles to Cover AI Perils

Traditional hedging products remain scarce because the insurance market for artificial intelligence risk has been slow to develop, according to industry data. Even the physical risks confronting data centres prove difficult to insure. Earlier this year, the Insurance Services Office released a template explicitly excluding damage caused by generative AI from a wide range of coverage claims, while several major firms requested regulatory permission to omit such novel threats. This wariness leaves a market gap that smaller startups attempt to fill, though those firms remain piddling in size relative to the task. Insurers struggle because standard economic risk relies on quantifiable probabilities, whereas the wild threats associated with AI remain poorly understood with a thin base of evidence, according to market analysts. Consequently, neither buyers nor sellers know how to price them. At the absolute extreme end of the spectrum, analysts note that the dangers presented by misaligned superintelligence are intrinsically unhedgeable.