Talent Shortages Thwart Deep Learning Projects

Even if they aren't doing artificial intelligence yet, many enterprises are contemplating it. But what about the even more advanced form of artificial intelligence called deep learning? Are organizations planning those projects yet? How far along are they? What are some of the obstacles those projects face?

As you might guess, those projects are a little further off for many enterprises as so many are still grappling to gain a foothold with more basic artificial intelligence technologies. Yet some of the organizations that are more comfortable with these technologies -- that maybe have people on staff who subscribe to AI and data newsletters -- are a bit more advanced and are planning for deep learning projects.

That's what a new survey-based report says. It comes from a poll O'Reilly conducted with subscribers to its AI, data, and programming newsletters which yielded 3,300 responses. The results are pulled together in a report titled How Companies are Putting AI to Work Through Deep Learning.

Of the respondents, 28% said their organizations are using deep learning now. That number may seem high, but those participating in this particular survey may be from organizations that have been working with some of the more advanced technologies already.

Of those who weren't yet working with deep learning, a total of 54% of respondents -- more than half -- predicted that deep learning would play an essential or large role in future projects. Here's how that breaks down. A full 35% said that deep learning would play a large role, and 19% said it would play an essential role. Still, the largest group, 38%, said deep learning would play a small role in future projects, and just 8% said it would not play a role at all.

In spite of the large numbers that say they plan to employ these technologies, deep learning is still cutting edge. It's still a future technology that not many organizations have in place today.

"Over the last few years, companies have been building data infrastructure and platforms for analytics and machine learning," wrote O'Reilly Chief Data Scientist Ben Lorica and O'Reilly VP of Content Strategy Mike Loukides, authors of the report. "Deep learning remains a relative new technique, one that hasn't been part of the typical suite of algorithms employed by industrial data scientists. So it's no surprise that the main factor holding companies back from trying deep learning is a skills gap."

Indeed, survey respondents cited "lack of skilled people" as the number one obstacle to implementing deep learning. The talent gap was cited by 20% of respondents -- more than double any of the other reasons cited. Other obstacles included hardware and compute resources at 9%, data-related challenges at 8%, company resources and culture at 3%, and accuracy and efficiency of deep learning models at 1%.

Yet, only 11% of those who responded said they’ve hired specifically for deep learning applications. Organizations may be holding off on hiring because they aren't ready yet to begin these projects. And many organizations also seem to be favoring training their in-house talent, according to the survey. The report noted that 75% of respondents said their company is using some form of in-house or external training program, and 49% said that their company offered in-house on-the-job training.  Still 35% said their companies were using either formal training from a third party or from individual training consultants or contractors.

"We are still in the very early stages of building truly intelligent systems," the authors wrote. "While deep learning is the dominant machine learning technique associated with current AI systems, future systems will likely incorporate many other (yet undiscovered) techniques."

[Want to take a deeper dive into deep learning? Interop ITX 2018 features a two-day AI Summit, April 30-May 1 at the Mirage in Las Vegas.]