Silicon Valley’s start-up companies have always had an advantage in recruiting over tech giants. Start-ups say, take a chance on us and we will give you ownership of the firm that can make you rich if we succeed.
Now the technology industry’s race to embrace AI could make that advantage moot, at least for the workers who have knowledge and skills in AI. (Nytimes.com)
Tech’s largest companies are putting big bets on artificial intelligence, as they bank on technologies such as face-scanning smartphones, computerized healthcare, and autonomous vehicles. As they are going down this high tech path, the big tech companies are paying big salaries to AI professionals that are extremely high, even in an industry that has routinely giving big bucks to top talent.
AI Workers With Limited Experience Can Garner Salaries of $300K or More
Some AI specialists, who usually have a Ph.D. just out of school, and sometimes people with less education and three or five years of experience can get paid from $300,000 to $500,000 per year in salary and stock options, according to a survey of nine people who are working for large tech companies. All of them asked for anonymity because they do not want to harm their career prospects.
Some of the best-known names in the artificial intelligence field get compensation in salary and shares of stock that total several million dollars over four or five years. Then they can renew or negotiate a new deal, much like professional athletes.
At the high end are the executives with years of experience managing AI projects. In a recent court filing, Google stated that one of the leaders of its self-driving car division, took home more than $120 million in incentives before he went to Uber last year.
Salaries are rising so fast that some say the technology industry should have a salary cap as the NFL does. One of the hiring managers at Microsoft says a salary cap on AI professionals would make life a lot easier.
Many Reasons for the High AI Salaries
There are many reasons salaries for AI professionals are rising. The automotive industry is competing with Silicon Valley for the same workers who can help to build and design self-driving cars. Big companies such as Facebook and Google have a lot of money to spend and problems they believe AI can fix, such as creating digital assistants for smartphones, smart home gadgets, and spotting content that is offensive.
The biggest reason is there is a shortage of AI talent. The tech giants are trying to land as many of those workers as they can. Solving difficult AI problems is not easy, like building a routine smartphone app. On the entire planet, fewer than 10,000 people have the skills to work on serious AI research, according to an analysis by Element AI, a lab in Montreal.
While what we are seeing with these increasing salaries may not be good for society, these companies are behaving rationally, according to Andrew Moore, dean of computer science at Carnegie Mellon University. The companies are anxious to grab this small number of people who can work on AI technology.
One AI Lab With 400 Employees Has $138 Million in Salaries
Costs at an artificial intelligence laboratory called DeepMind, which was bought by Google for $650 million in 2014, when it employed 50 people, show the spiraling salary problem. In 2016, it was reported by the company that its staff costs for 400 employees were $138 million, or $345,000 per employee. Those types of salaries are hard to compete with if you are a smaller company.
Current AI Research Based on Deep Neural Networks
The cutting edge of AI research is based on several mathematical techniques called deep neural networks. These networks are algorithms that can learn a task by itself by crunching data. For example, by reviewing millions of dog photos looking for patterns, a neural network can teach itself to recognize a dog. This mathematical concept goes back to the 1950s, but it was on the fringes of industry and academia until 2012 or so. (Wired.com)
By 2013, Facebook, Google, and a few other major players began to recruit the few AI researchers who are experts in these techniques. Neural networks can now help recognize the faces that are posted on Facebook. They also can identify commands that are spoken into digital assistants such as the Amazon Echo, and can immediately translate foreign languages in the Skype service owned by Microsoft.
Those same mathematical algorithms are being used to improve the performance of self-driving cars and develop hospital technology that can identify disease and illness in medical scans. Digital assistants are being created that can not just recognize spoken words but understand what they mean.
Tech Companies Are Hiring Professors and Reducing Ability to Teach AI to Students
With so few AI experts available, large tech companies are also moving to hire the best and brightest who work at colleges and universities. In the process, they are putting a limit on the number of instructors who can teach AI and machine learning.
For example, Uber hired 40 people from Carnegie Mellon a few years ago, which ran a revolutionary AI program to work on self-driving cars. Over the last four years, four of the best known AI researchers at universities have left to take high-paying jobs in the private sector. At the University of Washington, six of 20 AI professors are working for private companies on leave.
But some professors are finding ways to compromise. One professor at the University of Washington turned down a $180,000 position at Google. He chose a post at the Allen Institute that would let him continue teaching. He said there are many faculty who do this by splitting their time between academia and industry. While the salaries are much higher in the private sector, some professors choose to split their time because they really care about teaching the next generation of AI professionals.
Some Tech Companies Are Teaching AI Skills to Current Employees
To increase the number of AI-skilled engineers, Google and Facebook have been running classes to teach machine learning and deep learning to current workers. These companies say that learning basic deep learning concepts is not difficult, and does not require much beyond high school math. But some expertise requires more difficult math and intuitive talent that some say is like a dark art. Very specific knowledge is needed to create technologies such as self-driving cars, healthcare services, and robotics.
To keep pace, smaller firms look for talent in unorthodox places. Some hire astronomers and physicists who have strong math skills. Other start-ups look for workers in Eastern Europe, Asia and others where lower wages prevail.
Overall, demand for AI professionals is outweighing supply, so people with AI skills and expertise can continue to expect people to line up to pay them very well for their talents.