Last week, I sat down with Intel’s Gadi Singer, vice president and general manager of artificial intelligence architecture, and Chris Rice, head of the company’s AI talent acquisition, to talk about AI workforce issues. Here’s what they had to say.
“What you see in the articles is relatively the truth,” said Rice. “One of the interesting things in AI is that it’s no longer just the technology companies that play in this space, you’ve got the finance industry, medical, retail, mobility, manufacturing—they are all starting to recruit AI engineers, whether they are developing a technology or applying a technology. Because of that, there is an increased global demand, and that is driving up the value of those engineers.”
But, interjected Singer, remember that AI is not one skill, one job description. “It is a diverse set of skills. You’ve got hardware architect, you’ve got designers, software developers, data scientists, and researchers.”
Given that the hottest area in AI is deep learning, encompassing all the neural network–related techniques, Singer continued, “people who have expertise in knowing how to develop those new techniques, these topologies, or how to implement them in the most efficient manner in software and hardware obviously have high value, he said.”
The other thing driving value, Singer said, “is that the frontier in this space is moving faster than any technology that I’ve seen. The state of the art in deep learning in 2016 is called ‘legacy’ by 2018. So, people who have the ability to continuously learn and be on or ahead of this fast-moving frontier of deep learning are obviously very valuable.”
On efforts to fill the pipeline by educating more AI engineers:
“There are a couple of issues with this,” said Rice. “Academic institutions globally have started doing pretty well in putting emphasis on this realm of skill sets. But a lot of the research is actually being conducted in industry because of that fast innovation cycle, so industry is actually hiring a lot of professors out of academia. That’s a confounding issue: industry is moving to pull more people out of academia at a faster rate than it can produce them.”
On the brighter side, Singer says he sees an increase in the number of students and in the number of classes being offered.
And helping it all, he pointed out, is the cool factor. “Consider data science,” Singer said. “Data science used to be considered something dull, some area of statistics. But today, data science is really cool. And that attracts talent into academia, into industry.”
“So,” he indicated, “even though the need is strong, it translated into a pull, both in academia and in the larger population, that will eventually increase supply.”
On retaining a company’s AI engineers:
”We are seeing, in data science, that people are changing jobs every 21 months,” Said Rice. “There is a higher turnover not because people want to bounce around between jobs, but because the problems they are working on are so diverse. They go from one place to another place to work on new and interesting problems.” The result, he says, is that companies struggle to find ways to keep them. “You have to be much more aggressive in the way you structure tenure for a person when they come into an organization,” Rice says.
This sounds simple, but it is powerful, said Singer: “Make it a fun and growing experience. So rather than talking in terms of packages and so on, the most significant factor for many of the top talents is: Do they feel that they are doing something that is at the leading edge of technology? Do they feel that they learn, so that, year over year, they grow? Do they feel that the work they are doing matters?”
Rice adds that today’s top engineers want to do more than just sitting in a cubicle punching out lines of code. Rather, “they want to work on something that is going to help the bigger society.”
On the importance of a diverse workforce in preventing algorithm bias:
“The way you train machine learning has an impact on the way it sees the world,” said Singer. “With example-based, supervised learning, the set of examples being used impacts how it is analyzed. The best way to deal with [preventing bias] is having diverse teams. When the team has diversity, and looks at the problem from multiple angles, the solution it creates is going to have a well-rounded view.”
For the longer term, he envisions the solution being to train sophisticated AI systems of the future what bias is, how to spot it, and how to avoid it. “That is not something that can be done with today’s technologies, but mid- to long-term, I see this capability evolving,” Singer said.
On why kids today should consider aiming for a career in AI:
“AI has the advantage of being both highly impactful and multidisciplinary,” said Singer. That means it can support a variety of interests: “Whether you want to go more on the human side of the interaction, or are more statistics oriented, or go more to programming or engineering; each of those [corresponds to] a branch of AI. For example, for someone who really wants to work on health-related areas—attending to elderly, for example—those have elements in AI. And because it is so diverse, it allows you to connect to it with whatever is the special thing that makes you, you.”
The bottom line according to Rice: In the future it is going to be hard to have a career that doesn’t involve AI. “I have a very young child,” he said, “so I am of the assumption at this point that any career she has is going to have artificial intelligence implications.”