The author is a Reuters Breakingviews columnist. The opinions expressed are his own.
By Jonathan Guilford
NEW YORK, Sept 16 (Reuters Breakingviews) - As investors weigh OpenAI’s valuation, they might consider the humble paperclip. A cautionary tale about corporate profit maximizers building a robot that so excels in producing the office supply that it wipes out humanity might seem far-fetched. But a single-minded capitalist could make the economically rational decision to bear such a risk. As OpenAI races towards a fundraising that could value it at $150 billion, the implicit promise is that gains enormous enough to make that danger thinkable are on the horizon. That itself underscores the barriers to growth.
The paperclip story goes like this. One day, engineers at ACME Office Supplies unveil a hyper-sophisticated AI machine with one goal: produce as many paperclips as possible. The incomparable silicon intellect chases this task to the furthest extreme, converting every molecule on Earth into paperclips and promptly ending all life.
Profit-hungry OpenAI investors like Microsoft MSFT.O might be assumed, like ACME, to only value short-term gains, inviting the risk that they build their own Paperclip Maximizer. Sam Altman, OpenAI’s CEO, says that he is mindful of the risk. His company’s structure is meant to limit bad incentives, capping profit available to investors. Such protections are worth an asterisk now: a ceiling on profit was set in 2019 at a 100 times return for initial investors. OpenAI initially expected to lower it over time. Instead, the company's latest fundraising now hinges on changing that structure, including by removing the cap, Reuters reported.
That’s an eye-popping notion. If OpenAI is worth $150 billion - 74% above its last benchmark, according to Bloomberg - then a 100-times cap only kicks in at a $15 trillion valuation, larger than the market value of Alphabet GOOGL.O, Amazon.com AMZN.O, Apple AAPL.O, Microsoft, and Nvidia NVDA.O combined. And as crazy as that sounds, in the warped world of AI, the math could actually work out. Assume that Paperclip Maximizer replaces the value of every U.S. citizen, some 330 million people. It would suggest each American is worth just over $45,000. That’s far short of both the country’s GDP per capita and, even more so, the $11 million at which the Department of Agriculture found its peer agencies place the statistical value of a human life in 2022. It’s not a perfect corollary, but a number like that suggests the value of the AI market overall could get into the quadrillions.
Thing is, anything so obviously dangerous invites regulation. Already, competition watchdogs are scrutinizing OpenAI; the White House is holding meetings with industry leaders. Real, extinction-level risk would invite the security state to bear down. And, anyway, there’s a worse outcome for investors with all the hype: AI could be meaningless. In May, Google's much-anticipated addition of AI to search ended up politely advising users to put glue on their pizza. It could be that OpenAI is no more dangerous than a regular old paperclip. And considering that possibility makes for a much more mundane valuation exercise.
Follow @JMAGuilford on X
CONTEXT NEWS
OpenAI, the developer of artificial intelligence models that power chatbot ChatGPT, is in talks to raise $6.5 billion at a valuation of $150 billion, Bloomberg reported on Sept. 11. The terms of the fundraising are contingent on whether the company changes its corporate structure, under which it has directed investment to a "capped profit" subsidiary that limits backers' upside, Reuters reported on Sept. 13.
(Editing by Lauren Silva Laughlin and Sharon Lam)
((For previous columns by the author, Reuters customers can click on GUILFORD/
Jonathan.Guilford@thomsonreuters.com))