(Bloomberg) -- Runaway confidence in artificial intelligence risks repeating the mistakes of the crypto hype bubble of only two years ago, economics professor and Nobel laureate Paul Romer warned.
“Right now, there’s way too much confidence about the future trajectory of AI,” Romer told Bloomberg TV at the UBS Asian Investment Conference in Hong Kong on Wednesday. “When people project this forward, I think they’re at risk of making a very serious mistake.”
OpenAI triggered a tide of excitement about AI with its wildly popular ChatGPT in late 2022, which has led to huge investment in computational power and cloud capabilities by tech giants from Microsoft Corp. to Alibaba Group Holding Ltd. That’s turned Nvidia Corp. into a multitrillion-dollar company, as its chips are most effective at handling AI training. But Romer warned it’s dangerous to expect the current rate of improvement in AI services to continue indefinitely.
“We’ve benefited from scaling up compute and ingesting a whole lot of data,” the former World Bank chief economist said. “Scaling up compute is pretty easy. It’s just more machines and more chips. But what’s going to happen is we’re not going to have enough data.”
The key cautionary tale in Romer’s eyes is the example of autonomous vehicles. Companies like Tesla Inc. have been promising fully self-driving cars for years, but their automation systems continue to be plagued by reliability issues and edge-case scenarios that prevent the realization of that promise.
In two years, people will likely look back on the present moment and say that “it really was a bubble, we overestimated where it was going,” Romer concluded.