AI’s Steady Compounding

My general sense of things is that folks who were optimistic about AI progress at the start of the year remain optimistic that generative AI is a powerful general-purpose technology that will evolve sooner rather than later into something worthy of the term “artificial general intelligence”—and then perhaps into something even more. (Some leading voices have even accelerated their timelines .) Normies seeing headlines about the government putting frontier models on a leash might well think transformative AI is already here. Three things: First, whatever might be happening with benchmark performance and frontier model release, prediction markets aren’t suggesting the Singularity is nigh, or at least not more nigh than it was back in January. Second, the generative AI economy—meaning services (software, model access, and rented AI compute) not chips—has generated $110 billion in revenue over the past 12 months and is now running at an annualized pace above $175 billion, according to the new Exponential View analysis, “The State of the AI Economy.” Those are important economic data points suggesting how business sees the technology’s potential. Third, it remains early days as far as businesses putting GenAI to productive use and getting important bottom-line results. The Exponential View report itself offers evidence of value creation that is more suggestive than conclusive. This slide gets at the idea: A similar story comes from my recent podcast chat with Erik Brynjolfsson —senior fellow at the Stanford Institute for Human-Centered AI and director of the Stanford Digital Economy Lab . Brynjolfsson: I think we all can see that the technology is just mind blowing. What it does is almost magical. . . . [Yet] if you talk to businesses, there’s a lot of frustration out there. These projects are not instantly delivering and some companies are having some success, some aren’t. But if you look at them more broadly, I think that the returns have been somewhat disappointing. To me, that’s totally natural. . . . These incredibly powerful technical capabilities don’t translate to business productivity and improve business performance until you make the complementary changes. . . . That’s what companies are doing. I have a startup called Workhelix that helps them do that more rapidly. As we help them use the technology more effectively, they’re beginning to get some of the productivity gains. I think within the next couple of years, it’s going to be visible in the aggregate numbers. I think that’s right, but even now when you dig down you can see the technology is helpful, as documented in a recent Wall Street Journal story “ See How AI Giants Are Using AI in Their Own Offices ” by Katherine Bindley. Among the examples cited: Google says an invoice agent lets finance review five times more invoices and is on track to save $200 million annually. OpenAI uses Codex to resolve billing questions, build customer dashboards, and create sales demos, among other uses. Anthropic uses Claude to automate marketing-operations tasks that once took 15 minutes to an hour. But humans are definitely still in the loop reviewing the AI-generated work. No, none of these are miracle materials, miracle cures, or productivity superbooms. Still, of such incremental advances an economic revolution is built. And those other things might still be on their way. The post AI’s Steady Compounding appeared first on American Enterprise Institute - AEI .