A skeptical look at AI investment

Alison: Generative AI is being heralded as one of the most transformative innovations in human history, and AI optimism has become one of the market's biggest drivers. Companies are estimated to devote to over $1,000,000,000,000 to AI related spending in the coming years. So will the benefits and returns of the technology justify the cost?

Jim: We're a couple years into this, and there's not a single thing that this is being used for that's cost effective at this point.

Alison: I'm Alison Nathan, and this is Goldman Sachs Exchanges. Every month, I speak with investors, policy makers, and academics about the most pressing market moving issues for our top of mind report from Goldman Sachs Research. This month, I looked at generative AI. I think we're all familiar with the bull case by now. My colleague, Joseph Briggs, from our global economics research team estimates that generative AI could ultimately automate a quarter of all work tasks and boost US productivity by 9% and US GDP growth by 6.1% cumulatively over the next decade.

And investors have certainly bought into the theme with the big tech firms at the center of it all, accounting for over 60% of the S and P 500 indexes year to date return. But given the enormous cost to develop and run the technology with no so called killer application for it yet found, questions about whether the technology will ever sufficiently deliver on this investment have grown. I spoke to 2 people who are skeptical. Darren Asimovu, an institute professor at MIT, is the author of several books, including Power and Progress, Our 1000 Year Struggle Over Technology and Prosperity. In a re

I asked him to explain.

Daron: Part of the reason why I wrote the paper is because I did see a lot of enthusiasm, some quantitative, some qualitative from commentators, some experts, others observers of the tech industry about the transformative effects that AI is going to have very quickly on the economy. And I think economic theory actually puts a lot of discipline on how some of these effects can work once we leave out those things like amazing new products coming online, something much better than silicon coming for example in 5 years. That's of course, if that happens, all right, that's big. But once you leave t

And my prior, even before looking at the data, was that the number of tasks that are going to be impacted by Gen AI is not going to be so huge in the short run because a lot of the things that humans do are very multifaceted. So almost all of the things that we do in transport, manufacturing, mining, utilities has a very central component of interacting with the real world. And AI ultimately can help with that as well, but I can't imagine that being a big thing within the next few years. So therefore, my intuition was that from the beginning, it's gonna be pure mental tasks that are going to b

So the way that I go about doing that is I rely on one of the most comprehensive studies by Elendou, Mishkin and Rock that codes what the current AI technology, generative AI technology combined with other AI technologies and computer vision could ultimately do. What they did seemed fairly fairly solid. So I decided to start from what they did. So if you take their numbers, it suggests that something like 20 percent in terms of value added share, in terms of the economic importance of the tasks that we do in the production process could be ultimately transformed or heavily impacted by AI. But

And in that, there is another paper by Neil Thompson and Martin Fleming and co authors, which looks at a subset of these technologies, t