Why it’s time for CIOs to invest in machine learning

LinkedIn

Cornell University wants to help whales avoid getting hit by ships, so it is working on an algorithm that uses audio recordings to alert ships to the whales’ whereabouts.

Dassault Systèmes is creating a 3D model of a human heart that will allow surgeons to test the performance of pacemakers before opening up patients.

winningAlgorithms uses social media feeds to keep cyclists in the Tour de France informed about the status of the race five minutes before the media broadcasts the same information.

Sure, machine learning has already had a significant impact on the worlds of science and culture, and in life, but it will be years before CIOs need to start worrying about enterprise machine learning applications … right?

No. That’s not true, according to Dan Olley, CTO of Elsevier, a global information solutions company. “If CIOs invested in machine learning three years ago, they would have wasted their money,” Olley says. “But if they wait another three years, they will never catch up.”

What exactly is machine learning?

For years, computing has been stuck in an “if/then” paradigm. “Computers are good at A=B, or A > B, but they are bad at A is similar to B,” says Olley. “Until now, only humans could handle ‘similar to’ situations, but with machine learning, we can train algorithms to perform highly complex functions from describing an image to making judgement calls.”

Say you want to sort and categorize all of your digital photos. “If every picture of a dog were identical, it would be easy for an application to recognize dog photos and tag them appropriately,” Olley says. “But dogs are not identical to one another, so the machine needs to see a series of photos labelled “dog” until it learns to recognize dogs in the abstract. But once it’s trained, the machine can sort those photos on its own.”

So, unless they are dog lovers, why should CIOs care? 

Why machine learning matters to CIOs

To Olley, machine learning fills a gap in technology that has existed for a long time: solving complex problems with pattern recognition. “With the majority of Elsevier’s revenue coming from technology-based products and services, we started using machine learning in our commercial products, but it’s equally applicable to internal IT platforms,” Olley says. “Take the classic challenge of matching customer contacts and addresses or spotting trends in your financial data. The more you train the application to ‘understand’ the data, the better your predictive analytics.”

At Elsevier, Olley puts his money where his mouth is.

“One of our divisions creates educational materials for nurses, but many of our students get frustrated with the challenging material, drop out of the course, and never take their certification exam,” he says.

Elsevier would like to increase the number of nurses who pass the test, and they use machine learning to help. “We are using algorithms that learn how students actually use the course material,” he says. “This way, we can create adaptability and personalization within the course to engage the students and drive better pass rates.”

Education is just one example of how Elsevier uses machine learning, and it finds that the technologies are applicable across product suites, from helping scientists make new discoveries to supporting healthcare professionals so they can provide the best possible care.

If CIOs can take a body of information, such as a CRM system, a call center app, or even a corporate Intranet, and from that information construct a set of data to teach a machine how to solve a problem, they will create much more powerful, adaptable systems. “You can see applications for this technology from automating your call center with speech recognition to creating more accurate sales forecasting or even spotting fraud,” Olley says. “Infrastructure tools like Splunk are using machine learning to find patterns in log files, and there are many applications in security.”

How to get started with machine learning

It all starts with research. “The first thing is to start building awareness in your organizations,” says Olley. “The algorithms required to solve most initial problems exist and are freely available. All you need to do is to start learning how to use them. We are already seeing machine learning as a service coming out of Google, Microsoft and Amazon. There are also some great free, or low cost, training resources coming out of MIT, Stanford, Coursera and Udacity.”

Next, you should pick a beta. When you have some excitement around machine learning in your team, pick a few small very manageable problems and go solve them. Likely targets involve a repeatable process that requires a human to be involved, because it is at the knowledge worker level. “Call center software is a great place to start,” says Olley. “Or you might start with a BI or financial analysis process. But either way, you need to start practicing. Machine learning requires a very new way of thinking.”

Then focus on your data. If you don’t have a handle on your data, you are miles away from effective machine learning. “With machine learning, data will become a critical asset, because it’s the training set,” Olley says. The better your training sets, the better you can use these new technologies.

Don’t aim for the perfect machine. Say you work for a mortgage company. Most of the mortgage applications you receive are a definite yes or a definite no. It’s relatively easy to write a program to automate those scenarios. However, the mortgages that land somewhere in the middle require more analysis and are sent to the underwriting team. “Your first attempt at machine learning will not remove the need for those underwriters,” Olley says. “Maybe the machine is able to make 20 percent of your underwriting decisions. The humans can make the other 80 percent, but if they are feeding their decisions back into the machine, the machine will get smarter and smarter.”

Finally, you should start workforce planning. The implications of machine learning over the next 10 years are huge. “This technology has the potential to remove a large proportion of the current knowledge worker workforce,” says Olley. “As companies, we need to understand that our investments in machine learning will reduce the number of knowledge workers that we need and increase our need for people well-versed in data and analytics. We all need to think through the broader implications.”

About Dan Olley

Olley joined the RELX Group (Formerly Reed Elsevier) in 2004 as technology director for RBI UK, where he headed up online product development, and has held successive positions ever since. A leading champion of data-driven agile development, cloud computing, and big data and analytics-based technologies, Olley led the early adoption of many such technologies, which provided technical flexibility and agility for rapid transformation. As part of the executive teams within RBI and Elsevier, Dan continues to drive organic online product growth across the portfolio. Prior to RELX Group, Dan held technology and product management leadership roles with GM Financial, Wunderman Cato Johnson, and IBM, as well as a number of software organizations in the United Kingdom and other international locales.