Dev

‘AI divide’ across the US leaves economists concerned


Artificial intelligence-related technologies show promise but are clustering in AI hubs across the world, a group of economists has reported.

The econ boffins looked at government data from a 2018 survey covering 474,000 firms, as a representative sample of some four million US businesses. They found that fewer than 6 percent of businesses used any of those AI technologies at the time. But many large firms with more than 5,000 employees were doing so – when weighed by employment, average adoption came to more than 18 percent.

AI adoption appears to have risen significantly since then, which is entirely unsurprising given the industry hype and the slow pace of legal pushback. A recent McKinsey study reveals that 79 percent of survey respondents say they’ve had some exposure to generative AI inside or outside of work, and that 22 percent say they use the technology regularly.

The focus of the paper is primarily to document who is adopting AI technologies and what technologies are being adopted; its aim is not to look at the revenue consequences of doing so. But the paper does allow that there’s an association between AI adoption and revenue growth. And that linkage has invited scrutiny of the distribution of AI’s projected economic benefits.

“[W]hile the spread of AI use across the country remains in its initial phase, the potential for an ‘AI divide’ across regions and cities is attracting concern,” the authors state in their paper. “Investigating where early AI use was located among startups, we find considerable concentration.”

In the paper distributed via the National Bureau of Economic Research, authors Kristina McElheran (University of Toronto Scarborough), J. Frank Li (University of British Columbia), Erik Brynjolfsson (Stanford Digital Economy Laboratory), Zachary Kroff (US Census Bureau Center for Economic Studies), Emin Dinlersoz (US Census Bureau), Lucia Foster (US Census Bureau), and Nikolas Zolas (US Census Bureau) have analyzed a snapshot of the AI economy in the US.

AI here refers to five specific technologies: autonomous vehicles, machine learning, machine vision, natural language processing, and voice recognition.

Nashville, who knew?

AI usage at the time of the survey centered in California’s Silicon Valley and the San Francisco Bay Area, but was also noted in Nashville, San Antonio, Las Vegas, New Orleans, San Diego, and Tampa, as well as Riverside, Louisville, Columbus, Austin, and Atlanta.

Kristina McElheran, associate professor of strategic management at Canada’s University of Toronto Scarborough, told The Register that it’s common to see the clustering of economic activity.

“This can be great – especially for advancing a new technology – because ideas and knowledge can flow more-easily between firms and workers when they are close to each other,” McElheran said.

The concentration of experts and skill labor can help technologies develop at a faster pace, she explained, and over time, knowledge and skills tend to spread.

“The downside of these ‘agglomeration effects’ comes if particular locales get regularly and systematically left behind. Or, if an area becomes very specialized in a set of activities, things can become challenging pretty fast if there is a shock to that system,” said McElheran.

“I grew up in Michigan, and saw first-hand what happened to an auto-intensive economy when domestic auto manufacturing went into sharp decline. That can be hard to come back from. That said, Detroit is enjoying quite a renaissance right now – proof that it’s hard to say for certain what the long-run future looks like.”

McElheran said she and her colleagues were not trying to make policy or investment recommendations. Rather, their goal was to establish a baseline from which to plot the trajectory of AI adoption, a precondition for developing policies and directing funds.

“The only thing I feel pretty sure of, and it’s based on a long line of research into technology adoption, is that technological innovation – in this case ‘what the AI can do’ – is only one piece of the puzzle,” she said.

“Another essential piece is working through all of the hiccups, missteps, and experiments needed to understand how to put these new tools to work in practice,” McElheran elaborated.

“This innovation in processes, job descriptions, value chain linkages, and skills is where things get really hard – and really interesting. Finding pockets of difficult problems – and then solving them – is how firms and societies achieve great things. Promoting this sort of innovation – which includes learning about risks and how to mitigate them – is a bet that will almost surely pay off.”

The future’s amazing?

McElheran observed that alongside AI adoption, she and her colleagues found a high level of process innovation.

“This makes sense in terms of the history of technological change,” she said. “It also reminds me that we have yet to understand how many of these applications will play out. It’s easy to look at what we have, think about automating all the humans out of it, and panicking. Instead, these findings make me optimistic that there will be really innovative possibilities down the road, and some of them will be amazing.”

She also pointed to another finding about the importance of business leadership in AI adoption.

“We looked in-depth at a rather large sample of young firms (over 75,000) and discovered that founders motivated to bring new ideas into the world and give back to their communities were highly correlated with AI use,” she said.

“A new insight this study offers is that high growth, innovative startups are exerting pressure on the rate and direction of change, and while we can’t promise anything, this sounds disruptive and dynamic in ways that could, ultimately, be very positive.” ®



READ SOURCE