The benefits of AI in industries as diverse as healthcare and finance are already being realized, thanks in no small part to huge investment by the world's tech giants. But in their race for the talent required to dominate the field, are these organizations resorting to short-term tactics that could end up sinking their efforts?
The world's largest technology companies spent an estimated $20-30 billion on AI in 2016, according to McKinsey, and this number is only going up. In October, Chinese giant Alibaba alone pledged to invest over $15 billion on AI research and development. Much of this is going on talent, with a dearth of trained - or even educated - candidates driving huge wages and increasingly cutthroat recruitment tactics. According to the job search website Indeed.com, June 2015 to June 2017 saw a 500% rise in the number of job postings in the field of AI. Of these job postings, 61% in the AI industry were for machine learning engineers, 10% were for data scientists and just 3% were for software developers. Recruitment giant Indeed put the national average salary for a Machine Learning Engineer at $150,429 per year in the US, while according to UK regulatory filings, salaries at DeepMind, an AI company owned by Alphabet Inc., averaged $345,000 in 2016.
This talent has to come from somewhere, and in a field as nascent as AI, few are sufficiently qualified to do something so complex. This has led companies to turn to academia, pillaging the sector for talent by offering dramatically higher salaries. High profile examples include Geoff Hinton, who left the University of Toronto to join Google, and Andrew Ng, who left Stanford University to become chief scientist at Chinese internet search company Baidu, but these are the tip of the iceberg. In one of the more extreme examples, Uber poached a team of 40 researchers from Carnegie Mellon University. According to the National Science Foundation, 57% of new computer-science doctoral graduates in the United States leave academia for industry jobs, up from 38% a decade ago. Many are even poaching PhD students before they finish their degree, in a similar way that sports teams scout players as young as 7 in order to get the jump on their rivals. The Guardian relayed one story from Maja Pantic, professor of affective and behavioral computing at Imperial College London, whose student left for a six-figure salary at Apple despite having one more year left to complete his studies. And these examples are becoming increasingly commonplace.
This could prove to be a real problem - clearly for academia in that it's losing its best researchers and professors, but also for the companies themselves as the pipeline for talent diminishes, with teaching standards dropping and students not completing their education. Both university executives and professionals have warned of a 'missing generation' of academics who would normally teach students and be the creative force behind research projects. Oren Etzioni, chief executive officer of the Allen Institute for Artificial Intelligence in Seattle, is among those who fear a brain drain, noting that, 'It’s a question for the whole ecosystem – who is going to train the next generation?' Peter Morgan, chief AI officer at Ivy Data Science, an AI-as-a-service platform and training company based in New York City, agrees, arguing that, 'The number of graduating master’s and Ph.D.-level computer scientists may decrease, which is the opposite to what the current market is demanding. These companies are in effect eating their own lunch.'
There is also a further issue in that these candidates are being swept up by a handful of companies, meaning their skills and experience are not shared among smaller organizations. Google, for one, has admitted that it now houses 'less than 50% but certainly more than 5%" of the world's leading experts in machine learning.' This leaves startups, where some of the most innovative developments should be arising - fighting for scraps. This also has potentially devastating complications for society. Pantic explained to the Guardian that, it’s a problem 'because only a diffusion of innovation, rather than its concentration into just a few companies, can mitigate the dramatic disruptions and negative effects that AI may bring about.'
The solutions to this problem are difficult to ascertain. It is easy to understand why AI talent would leave academia in favor of private enterprise, with salaries often as much as five times higher. Newly minted PhDs can earn upwards of $300,000, while top-ranked senior academics command multimillion-dollar, multiyear contracts. As a professor, if you are watching your students leave for starting salaries fresh out of school that far exceed your own, you justifiably feel a bit under-appreciated. It is also often the case that the work is simply more exciting, with the private sector offering the chance to tackle real-world problems. They also provide access to the kind of computing power and data sets that enable talent to test their ideas in ways that academia cannot possibly hope to compete with, allowing them to do more exciting work.
In order to stem the exodus, companies first have to respect that they need the pipeline. In fairness, it's easy to say 'you need to take a long-term view', but many do know this. The competition for talent is simply such that if one company backs off, they lose ground to a rival. What private organizations can do is take a more collaborative approach alongside academia, loaning back talent and providing universities with the data and computing power they need to test their research. In return, they can perhaps offer something to candidates for whom academic research was enjoying but the money too tempting to turn down. They could also offer terms to the university under which they get preferable access to the best talent. Furthermore, as Andrew Chamberlain, chief economist at employer reviews website Glassdoor, notes, companies also benefit from how AI academics think and putting them back into that research environment could help them in R&D at their enterprise. Many have already realized this, and a number of top academics who have gone on to large tech companies do continue to work at the universities in a diminished capacity, but more needs to be done.
However, it is not solely the responsibility of private enterprise - universities need to work harder to compete and governments also need to ensure that both AI innovation and the talent pipeline are carefully managed. It is unreasonable to expect universities fighting for funds to compete with the likes of Apple on salary, but they can offer other benefits. For example, many PhD students lack the business knowledge needed, so they can offer free access to external training and associated business schools to broaden researchers’ knowledge that could stand them in good stead for their later career, once they have devoted a certain number of years to academia. The government can also do more to subsidize salaries. In China, for example, they offer a significantly better proposition than in the US. Axios reported the case of Linsen Li, a Chinese-born, 30-year-old specialist in advanced batteries who gained a postdoc in MIT's material science and engineering program. Axios wrote, 'He received his PhD in chemistry from the University of Wisconsin, in all spending the last seven years in the US His infant son, William, is an American citizen. But he's reluctantly going home: Li tells Axios that, having received no teaching offers in the US, he's accepted a $65,000-a-year teaching slot at Shanghai's Jiao Tong University, along with the equivalent of a fat $900,000 in research funding, in addition to $250,000 to buy a house. Li is returning under China's Thousand Talents Plan, which seeks to lure back under-40 Chinese students and professionals to bolster the country's research sector.' Both the US and the UK are struggling at the moments, with anti-immigration agendas seeming to dominate the political scene that appears to be putting off the best foreign talent. But if they were to make it easier for AI professionals trained in other countries to come over, companies may find it less necessary to turn to universities.
It could be that the brain drain doesn't last. Scott Maxwell wrote in Inc that 'High salaries, global competition, open-sourced algorithms/platforms, and the lure of starting one's own company or joining an elite start-up will ensure that AI innovation will be democratized, even if it takes a few years [...] As word of such salaries filter down, more undergraduates will inevitably consider AI as a career.' This does not necessarily solve the problem though, as demand is likely to increase in line with supply. He also noted one AI professor, who told MIT Technology Review in 2014 that 'tech companies have stepped up their grants since they realized they'll soon run out of recruits.' This may be the case, but that was 2014 and the problem is only getting more severe. Both academic institutions and governments cannot sit back and hope private enterprise come to their senses, they need to do all they can to ensure they reap all possible benefits of the talent they are training.