What’s behind the growing skepticism about AI? Daron Acemoglu on excesses of enthusiasm, returns on investment, and the distinctive nature of human reasoning.
When OpenAI released its ChatGPT-3 chatbot in November 2022, it sent a shock wave around the world. Here was a machine that looked a lot like the artificial intelligence many had been imagining for generations—not least through science fiction like 2001: A Space Odyssey, Blade Runner, or The Matrix.
ChatGPT also sent a hype wave across the tech industry—leading even to predictions that AI would soon be on its way to replacing most human work. An enormous investment boom followed. Over the next couple of years, tech firms are on track to spend more than US$1 trillion on AI. This year, they’ve already spent hundreds of billions. Microsoft alone is set to invest more than $55 billion—more than the gross domestic product of Tunisia or Latvia.
But as money followed the exhilaration, skepticism is now following the money. Some investors say the industry’s high stock prices are a bubble, soon to burst. Others wonder where AI's “killer app” is—or whether one’s ever coming. Meanwhile, all this new computing power is straining electricity grids, as AI firms build enormous new data centers—which, worldwide, use as much electricity as Italy every year—and driving carbon-dioxide emissions.
Now, the stock market appears to be siding with the skeptics. Shares in AI and tech firms dove throughout late August and early September. Executives at top investment firms like Goldman Sachs openly doubt AI’s long-term economic prospects. And analysts are forecasting that OpenAI will lose around $5 billion in 2024. What’s happened?
Daron Acemoglu is a professor of economics at MIT and a co-author of the 2023 book Power and Progress: Our 1,000-Year Struggle Over Technology and Prosperity. Acemoglu says AI has a basic economic problem: Its chatbots and other apps don’t generate enough revenue to cover the massive investments they’re built with—and they aren’t likely to any time soon.
More fundamentally, the AI industry is struggling against all the expectations that have driven so much investment in it: It’s supposed to be revolutionary, but it still can’t build models that do even passably well what humans are distinctively good at—solving complex problems and doing complex tasks. But because of the way the industry thinks about itself, that’s not slowing it down …
Michael Bluhm: The hundreds of billions tech firms are investing in AI this year—where’s it all going?
Dylan Gillis
This article is for members only
Join to read on and have access to The Signal‘s full library.
