Daron Acemoglu, a closely-watched voice in the debate about the future of artificial intelligence, has not bought into the excitement, or the fear, that has been generated by the rise of the Chinese AI startup DeepSeek.
The Massachusetts Institute of Technology economist, who won the Nobel Prize last year, acknowledges that DeepSeek’s latest AI product, R1, was an impressive achievement that appears to offer a cheaper and more efficient alternative to the models developed by US companies like OpenAI. But, Acemoglu said, the people who have been rushing to download R1 – just like those paying for OpenAI’s ChatGPT – are mostly "using it as a toy to explore it.”
"I don’t see a clear path from DeepSeek to business adoption that will be revolutionary for businesses,” Acemoglu said in an interview.
Acemoglu has long been known in economic circles for his work on the disruptive economic effects of new technology, like the research he did on the job losses caused by earlier waves of automation.
Since ChatGPT burst onto the scene, Acemoglu has been developing models for how AI might play out. He has come to think that AI will be powerful, potentially displacing as much as 5% of the labour force. But that forecast makes him something of an AI curmudgeon in comparison to the sort of world-changing effects – stronger economic growth, surging worker productivity – predicted by many in the tech industry and on Wall Street.
"I hope I’m wrong, I hope we get some productivity growth. That would be really cool,” he said. "But I don’t see it yet.”
The arrival of DeepSeek has scrambled many of the established narratives that have grown up around AI. Tech giants like Microsoft Corp and Meta Platforms Inc have been betting that they will need tens of billions of dollars of semiconductors and other computing gear to build and run the most powerful AI models, following the strategy that has been working so far. Despite some brief periods of doubt, most investors have supported that strategy by continuing to pour money into shares of the biggest spenders on AI.
The success of DeepSeek, though, has indicated that there may be other, more efficient paths to progress that don’t require nearly as much computing power. That prospect led to a rout in technology stocks on Jan 27, the Monday after DeepSeek’s popularity became clear. The shares of Nvidia Corp, which makes the chips that dominate the market for AI accelerators, saw a record US$589bil (RM2.61 trillion) in market value wiped out in one day.
Since then, tech stocks have pared their losses. And the biggest players in the AI world, such as Meta, Microsoft and Alphabet Inc, have indicated they are not moving away from their plans to sink ever greater sums of money into the technology.
Acemoglu said that the big tech companies have too much invested in their long-term AI strategies to back down now. But he believes that DeepSeek’s unexpected success points to the kinds of dangerous blind spots that have developed in both the investing world and Silicon Valley.
"It reveals how much group think there was in the multi-trillion dollar AI industry in the US in that they didn’t consider any of these things,” he said. "What else are they missing?” – Bloomberg
