The self-driving car space is getting increasingly more cutthroat.
The sheer number of lawsuits filed recently are a testament to that.
Tesla, for example, is suing its former Autopilot director Sterling Anderson. The lawsuit claims Anderson stole data for a competing venture, Aurora Innovations, that hasn’t even come out of stealth mode yet. Aurora denies the claims.
Buried in that lawsuit, though, was some interesting commentary on the competitive nature of the industry and the kind of culture it has bred:
“In their zeal to play catch-up, traditional automakers have created a get-rich-quick environment. Small teams of programmers with little more than demoware have been bought for as much as a billion dollars. Cruise Automation, a 40-person firm, was purchased by General Motors in July 2016 for nearly $US1 billion. In August 2016, Uber acquired Otto, another self-driving startup that had been founded only seven months earlier, in a deal worth more than $US680 million.”
(Otto is currently at the heart of a lawsuit filed by Waymo, Alphabet’s self-driving-car company, alleging Uber stole intellectual property.)
So why is it that tiny startups with little to no brand recognition are getting acquired for millions? AI specialists told Business Insider it has little to do with acquiring the startup’s tech and everything to do with nabbing talent.
That’s because there’s a serious lack of experts in the field of deep learning, a branch of artificial intelligence where computers learn on their own. Deep learning is key to advancing self-driving tech as it allows cars to learn safe driving at a much faster rate than traditional programming.
“The growth of demand is much faster than the rate of which we can produce people with PhDs or even master’s in this area,” Yoshua Bengio, head of the Montreal Institute for Learning Algorithms, told Business Insider. “There’s just an explosion of interest from the industry… and it’s like a fire growing on the prairie.”
Growth in demand
Machine learning has advanced rapidly in the last five years.
In March of last year, a program developed by Google DeepMind, a London-based AI lab, beat a world champion at the highly complex game of Go — a game that has more moves than there are atoms in the universe.
That’s not even the latest feat in artificial intelligence. In January, a bot developed by Carnegie Mellon University beat four of the world’s best poker players at Texas Hold ‘Em, sweeping them of $US1.8 million.
Companies are only expanding their AI teams as the field rapidly develops.
“Because machine learning and artificial intelligence is growing, companies are poaching professors all the time,” Geoff Gordon, an associate research professor of machine learning at CMU, told Business Insider.
But automakers, in particular, are making massive investments in experts because they have begun their AI efforts late compared to traditional tech companies.
Because deep learning has applications far beyond just self-driving cars, manufacturers are having to compete with each other and traditional tech companies.
Only 28 companies have more than 10 deep learning specialists on staff, accounting firm KPMG wrote in a 2016 report. What’s more, only six technology companies employ 54% of all deep learning specialists: Google, Microsoft, NVIDIA, IBM, Intel, and Samsung.
“The traditional power and talent of the auto industry was based in their product development group,” Gary Silberg, the head of KPMG’s automotive unit, told Business Insider. “So they would hire these amazing mechanical and electrical engineers at the top schools of engineering and they would be part of product development.”
“You can’t just turn on a dime and say, ‘OK, now we are going to go recruit AI geniuses and computer scientists and expect them to come to work with us,'” Silberg continued.
The bottleneck effect
Companies pursuing self-driving cars have turned to the universities with the brightest minds in the space to acquire talent.
In March 2015, Uber gutted Carnegie Mellon University’s AI and robotics center. The ride-hailing service poached over 50 experts, including Anthony Stentz, who had served as the director for the last four-and-a-half years.
A few months later, Uber announced a strategic partnership with the university to create the Uber Advanced Technologies Center for self-driving cars.
Andrew Ng, a renowned deep learning expert, left Stanford to become Baidu’s chief scientist in 2014. Baidu has a permit to test its self-driving vehicles in California and has tested its cars in China.
Universities are struggling to fill the void left by these poachings.
Companies have created a bottleneck where they are looking for talent at a rate that’s impossible to produce, said Bengio, the University of Montreal professor who is considered a pioneer in deep learning. It takes about five years for the average doctoral candidate to get his or her PhD.
The problem is amplified by the fact that professors are leaving universities to work at tech and auto companies.
“The demand for people who want to work in this area is growing but the bottleneck is that there are not enough academic labs, not enough professors to supervise all these students who want to get into this field,” Bengio said.
Gordon said CMU is definitely seeing an increase in the number of people who want to specialize in deep learning. For the upcoming academic year, CMU’s machine learning doctorate program received 800 applications. It had received 300 applications just two years prior.
Bengio said he gets 600 applications every year for students who want to study machine learning at the University of Montreal. He can only teach around 20 students.
That growing interest in machine learning is encouraging, but classroom sizes can only expand so much with a limited pool of professors.
“There are very few experts in this area until quite recently and many of them are being snapped up by industries because they are in so much of a need for that expertise,” Bengio said. “So now you have even less professors than you did just a couple of years ago.”
A $US1 billion going rate
The shortage of AI talent means the going rate for specialists is high, especially for startups, which allow companies to acquire talent in bulk.
More recently, Ford invested $US1 billion in Pittsburgh-based startup Argo AI. That investment will be spread out over 5 years.
Julie Lodge-Jarrett, Ford’s human resources director, said the Argo AI investment is indicative of how Ford has changed its strategy to recruit and retain talent as demand for artificial intelligence experts rises.
“As we’ve shifted from being an automotive company to being automotive and mobility we’ve seen an increase need for tech talent,” Lodge-Jarrett told Business Insider. “So in places like machine learning or AI or robotics… we see a greater need for ourselves at Ford and also a more competitive market for that particular talent.”
Lodge-Jarret said Ford is looking to buy more talent in bulk because it also helps them recruit in the future.
Argo AI is co-founded by Bryan Salesky, the former director of hardware for Google’s self-driving-car efforts, and Peter Rander, Uber’s engineering lead at its autonomous cars center.
Those kinds of connections make it easier for Ford to acquire the kind of talent that has traditionally worked for major tech companies.
Bengio said the Argo AI investment is indicative of how “the game has been in the last couple of years.”
“The sad thing is most of these are just for recruiting,” he said. “It’s kind of a loss for the economy because most often the projects these small companies had don’t continue once they have been integrated into the company…. It’s kind of a waste of resources and investment in some sense.”