Anthropic researcher Evan Hubinger said on X that he believes there is more than a 10% chance AI could kill all humans within the next decade. The warning came as senior figures at AI labs were said to be terrified by recent models' capabilities, and a group of AI CEOs, including Anthropic's Dario Amodei, publicly backed calls for a slowdown in AI development.
MoneyWeek argues the timing is suspicious. Anthropic is preparing to launch what the article describes as the largest initial public offering in history in a matter of weeks, with Amodei seeking a valuation of up to $2 trillion. The article notes that Amodei is not a neutral commentator and that much of the tech media treats AI executives' pronouncements as objective fact.
The article cites Cal Newport, a computer science professor at Georgetown University, who describes the industry's approach as 'doom trolling'. Fear sells, the argument goes, and outrageous claims about large language models generate free media coverage while making AI products appear powerful and desirable. MoneyWeek notes that in 2019 OpenAI said it would hold back its GPT-2 model on safety grounds, despite the model struggling with primary-school-level reasoning tasks.
The article suggests the AI industry's calls for regulation carry the whiff of 'regulatory capture', since red tape is more burdensome for upstarts than for established players. It also raises the possibility that calls for a slowdown are an attempt to put a brave face on the rising cost of the AI arms race, noting OpenAI is on course to spend $45 billion this year on training and inference while new models show diminishing returns.
For ordinary investors, MoneyWeek advises steering clear of this year's big flashy IPOs, diversifying widely, and keeping your head. It notes there is no knowing how long the current AI fever will last or how wide the damage will be when it breaks.