As businesses and other organizations worldwide manage ongoing pandemic effects and consider a slowly materializing post-pandemic future, a key question emerges: Just how much did COVID impact the AI industry?
“AI is still in the early stages of a long secular growth trend,” says Erik Brynjolfsson, Jerry Yang and Akiko Yamazaki Professor and senior fellow at Stanford Institute for Human-Centered AI (HAI) and director of the Stanford Digital Economy Lab. “But the pandemic is accelerating that trend, especially in areas like biotech, including drug design and discovery.”
Brynjolfsson is a steering committee member of the recently released 2021 AI Index, an annual study of AI impact and progress developed by an interdisciplinary team at Stanford HAI in partnership with organizations from industry, academia, and government. The report examines AI and the global economy through the lenses of investment, hiring, and corporate activity, using data from Burning Glass, LinkedIn, and McKinsey’s Global Survey on AI, among others.
Overall in 2020, AI hiring, investment, and adoption increased across the board, following long-term trends in the industry that will likely outlast the pandemic’s effects.
Investment in AI
Investment in AI-focused private companies remained strong during the pandemic, especially in sectors linked closely to COVID. Total global investment in AI grew 40 percent from 2019 to 2020, compared with only about a 12 percent jump from 2018 to 2019.
Moreover, as for company investments in AI capabilities “while some firms decreased their investments, slightly more firms increased their investments,” Brynjolfsson says. About half of McKinsey survey respondents said the pandemic had no effect on their investment in this area; 27 percent said COVID increased their investment.
“The biggest increases were in health care and pharma, where four times as many people increased their investments as decreased them,” Brynjolfsson says. Some of those applications led to pandemic breakthroughs: “One of the reasons we had a vaccine in record time was these technologies allowed us to analyze protein structures much more rapidly. We'll be seeing a lot more of that.”
Indeed, the same AI-driven techniques behind COVID vaccines can be used for other vaccines as well as interventions in heart disease and cancer. In fact, private companies channeled $13.8 billion into AI drug discovery in 2020, more than any other investment area, and 4.5 times higher than in 2019.
But sectors beyond health care showed increased AI investment, including education, retail, and automotive. “Pandemic or not, these are big investments for the long term,” Brynjolfsson says.
The AI industry also witnessed strong hiring growth during the pandemic. Across 14 countries analyzed, the AI hiring rate was 2.2 times higher in 2020 than 2016, on average.
Hiring in the U.S. grew more slowly than in other countries, like Canada, Brazil, and South Africa. Indeed, the U.S. saw a decrease in total AI job postings between 2019 and 2020 — from about 325,000 jobs to about 300,000 jobs.
“The slowdown in the U.S. was a surprising finding in this year’s AI Index,” Brynjolfsson says. “My hypothesis is that the U.S. is further along in AI hiring, so it's a little bit more mature. A lot of other countries are starting from a very small base, so there's a lot more room to grow. Hiring tends to go through an S-curve when there's a new opportunity, and the U.S. is further along that curve than most and has a much larger base than any other country.”
Hiring demand going forward, Brynjolfsson suggests, will reflect AI’s application as a “general-purpose” technology, like electricity: “Unlike some technologies that are very narrow, AI cuts across just about every industry and occupation — anything that’s information-worker-oriented.”
Thus, while health care, finance, and other service industries have been strong early adopters of AI, manufacturing, retail, and other sectors are expected to grow in hiring and AI-skill penetration as well. “It’s a little harder to use some of this technology in blue-collar-type areas,” Brynjolfsson says. “Physical work is harder to automate than knowledge work, but we’re seeing growth across the board.”
Case in point: Agriculture saw the largest jump in share of AI jobs from 2019 to 2020, growing a full percentage point. Brynjolfsson notes in a prior study that every occupation features some tasks suitable to machine learning.
The AI Startup Scene
More dollars are flowing into AI startups, the report found, up 9.3 percent from 2019 (comparatively, startup funding increased 5.7 percent from 2018 to 2019). The U.S. remains the largest private-investment destination, with over $23 billion in funding in 2020, followed by China with about $10 billion, though the latter country has strong public investments.
But this increased investment is being funneled into fewer startups worldwide — the number of newly funded companies decreased for the third consecutive year, from nearly 4,500 at peak in 2017 to about 700 in 2020.
“I’m hesitant to make a strong claim here,” Brynjolfsson says, “but this may reflect some consolidation or economies of scale. Initially, we may have had tens of thousands of flowers blooming and now we have thousands; as AI matures it moves out of the garages and consolidates with the more successful enterprises getting bigger.”
AI Adoption Now and in the Future
Beyond the pandemic, AI is expected to grow within and across industries, moving from fast adopters such as automotive manufacturing and financial services risk-management to less tech-focused industries, functions, and geographies.
More broadly, the context in which AI technology is deployed matters — a lot. “It’s not that you just buy the technology and slap it in,” Brynjolfsson says. “You need to do business process changes. That's been a hindrance to rapid adoption of AI, even though COVID has pushed companies to be more aggressive about it.”
While there are “plug-and-play” approaches to harnessing AI in certain sectors, more companies will have to consider broad contextual changes to get the most from the technology. “It’s a rethinking of how the distribution center or factory works and how you coordinate with suppliers and customers,” Brynjolfsson says. “We found that it could be up to 10 times as much investment in business process redesign and reskilling of the workforce as in the technology itself. It can be so costly that you may even get a productivity J-curve where productivity goes down before taking off.”
Indeed, Brynjolfsson’s recent research shows that “digital capital,” or investment in resources complementary to core technology like AI, is growing rapidly as a share of all capital in the global economy — now in the trillions of dollars, he says
As Brynjolfsson sums up, “2020 was a year when AI accelerated its move from the lab to commercial and industrial applications.”
Stanford HAI's mission is to advance AI research, education, policy and practice to improve the human condition. Learn more.