Sky-High Salaries Are the Weapons in the AI Talent War
If you want to command a multiyear, seven-figure salary, you used to have only four career options: chief executive officer, banker, celebrity entertainer, or pro athlete. Now there’s a fifth—artificial intelligence expert. One reason: No one can quite agree on how many there are.
Google, Facebook, Apple, Amazon.com, Uber Technologies, and others dangle dazzling pay packages to lure top academics to work on teams developing facial recognition, digital assistants, and self-driving cars. Even newly minted Ph.D.s in machine learning and data science can make more than $300,000. Beyond the tech industry, among those betting on similar expertise tailored to their interests are banks, hedge funds, carmakers, and drug companies.
Don’t balk at such pricey hires, Kai-Fu Lee, who previously ran Google’s business in China, told an audience of CEOs at this year’s World Economic Forum in Davos, Switzerland. “Google is paying a million dollars for these superstars,” said Lee, now a venture capitalist. “You may not need someone that high, but you’ve got to break the scale for at least one person.”
Designing AI systems requires a hard-to-come-by blend of high-level mathematics and statistical understanding, a grounding in data science and computer programming, and a dose of intuition. There are widely varied estimates of exactly how shallow the talent pool is. The answer matters, because it helps companies decide whether to build their systems in-house or rely on outside vendors. It also determines how much leverage experts have in salary negotiations.
On Feb. 7, Element AI, a Montreal startup that helps businesses design and implement machine learning systems, published a report concluding that about 22,000 Ph.D.-level computer scientists around the world are capable of building AI systems. Of those, only about 3,000 are currently looking for a job. In contrast, at least 10,000 related positions are open in the U.S. alone, says Element CEO Jean-Francois Gagné.
These figures are well below another estimate put out in December by Tencent Holdings Ltd., the Chinese internet giant. It wrote that the world has perhaps 200,000 to 300,000 “AI practitioners and researchers.” Element says Tencent counted too many coders who merely contribute to projects and lack the expertise to create novel algorithms and applications from scratch. The Montreal company, however, acknowledges that its own methodology had shortcomings.
Element scoured LinkedIn for people whose profiles included doctorates earned since 2015, mentioned key phrases (natural language processing, computer vision), and listed among their skills the programming languages (Python, TensorFlow) that underlie most AI software. The company says this might exclude a lot of researchers in places where LinkedIn isn’t relevant or who have experience but not a fancy degree.
Vishal Chatrath, co-founder and CEO of Prowler.io, a Cambridge, England startup that uses AI to improve things like trucking delivery routes and investment portfolios, hasn’t had trouble recruiting developers. “Talent hires talent,” he says. The important thing, he says, is to have intriguing problems to solve and some outstanding mathematicians and technicians already on staff to stoke expert interest. An added attraction is that Chatrath and his co-founders sold their previous company, voice recognition startup VocalIQ, to Apple Inc. in 2015.
Element has an incentive to highlight scarcity. The more companies despair of hiring their own experts, the more they’ll need vendors such as Element to do the work for them. “The talent shortage is real,” says Gagné, adding that he’s been struggling to hire even with AI pioneer Yoshua Bengio among his co-founders. Bengio, a computer scientist at the University of Montreal, is one of three men credited with helping to lead the AI boom. The other two are Yann LeCun, now at Facebook Inc., and Geoffrey Hinton, now at Google.
Governments and universities need to spend more money on training, Gagné says, especially at the undergraduate and master’s levels. At the current education rate, an influx of new experts will start to moderate salaries in three to four years, he says.
Most businesses don’t want to wait that long. Intel, Facebook, and Google are creating their own internal AI training programs. Google is also one of the companies experimenting with automatic machine learning, or AutoML, meaning AI that can create its own AI. The search giant recently began offering the service to cloud customers.
Despite the possibility of automatic machine learning, the demand for expertise has attracted swarms of headhunters to once-staid academic confabs with names such as the Neural Information Processing Systems (NIPS) conference. To woo candidates, recruiters organize increasingly swanky private dinners and after-parties. Chris Rice, head of global talent acquisition for Intel’s AI product group, says there’s little choice but to recruit aggressively at such events. “With talent this scarce,” he said at a NIPS conference in December, “it can be hard to find people.”