What Is Machine Learning Anyway?

By Tim Joyce

Tim Joyce, CIO at WDS, A Xerox Company
“Brands will be able to map customer lifecycles and predict future interactions with every customer, as well as potential issues that are likely to occur.” – Tim Joyce, CIO at WDS, A Xerox Company

In a world where concepts, like the Internet of things, are fast becoming a reality, machine learning has emerged as a key player. But what is meant by “machine learning”?

An arm of computer science and artificial intelligence, machine learning focuses on the development of programs that can learn from data, adapt to changes, and improve performance with experience. In short, machine learning is making one of computer science’s initial dreams a reality: Technology that emulates the human mind, computers that can see, hear, and understand.

An integral part of cognitive technology, machine learning systems are able to adapt and progress without the need for explicit programming. Thus, systems that learn from real-time data are more attuned to your customers’ needs,  and more dynamic.

Scientists at Xerox Research Centre Europe (XRCE) have been developing new machine learning capabilities since the early 1990s. Here, Guillaume Bouchard, a senior scientist at XRCE, describes what machine learning is, and how we anticipate it transforming the delivery of customer care.

ARVE Error: id and provider shortcodes attributes are mandatory for old shortcodes. It is recommended to switch to new shortcodes that need only url

The Opportunity of Data Captured from Multiple Touch-Points

The different touch-points from which tomorrow’s, and increasingly, today’s data is captured, means that a combined omni-channel collection of data will empower machine learning systems to learn continuously from every exchange — even from the use of the product or service itself.

This collective data will highlight patterns across customer-bases, and give brands deeper insights into each customer, based on the data captured from them and other customers with similar attributes. From this, brands will be able to map customer lifecycles and predict future interactions with every customer as well as potential issues that are likely to occur.

These technological innovations mean that in the future, customer care systems will adopt a different manner and tone of voice with different customers, based on previous interactions with them. This will pave the way for the next big step in brand-customer relationships.

Machines That Think Like Humans?

By 2020, customers will manage 85% of their relationship with brands without interacting with a human.
Source: Gartner Customer 360 Summit 2011

Looking ahead, with advances in artificial intelligence technology, machines will successfully emulate human cognition. This will take the potential to serve customers to different frontiers. Imagine healthcare virtual agents with the ability to diagnose ailments based on camera-time with the patient, or a virtual banker that is able to walk first-time homebuyers through the complexities of mortgages, and answer their specific questions with answers that are directly and explicitly relevant to their circumstances.

It’s exciting to consider the future technologies such as machine learning, but what’s more exciting is this journey has in fact already started. Today, WDS, A Xerox Company, is integrating machine learning research from the Xerox Research Centre Europe to deliver a truly cognitive virtual agent platform. Learn more here.

This article was first published on the WDS blog.

Subscribe to Simplify Work and receive email updates when we publish a new article.

Related Posts

Receive Updates


  1. Findia Group December 15, 2014 - Reply

    Thanks for sharing!

  2. Randy Cain December 16, 2014 - Reply

    The real value in machine learning will come when the machine can provide us with the “reason” for a particular outcome. Human cognition includes the ability to answer the “why” question, in many ways and to many levels. Learning something and making a decision but without being accountable for explaining and justifying that decision leaves out the important but often neglected “oversight” function. Without the ability and requirement to explain their decisions with something more than “this is the way others are doing it”, machines and applications that make decisions will never be able to “understand” the knowledge that they are using and, indeed, creating. Furthermore, without that understanding, the synthesis of innovation will not occur and, most importantly, the wisdom of when to take action OR NOT will always escape us. It’s true that the industry and academia have been working towards effective machine learning for many years. But, it is still truly in its infancy. Let us hope that we don’t send a child to do the job of an adult.

    • Tim Joyce January 14, 2015 - Reply


      I think this is an excellent point and a key area of concern with some fashionable machine learning technologies. For example Deep Neural Networks are essentially black boxes that do not bear introspection. When Google Plus (which uses DNNs) looks at your pictures, it can find all the ones of cats, but cannot tell you why any given picture is of a cat. Interestingly, this is similar to how humans think. When we look at a cat, we compare it with a mental image of a cat, and say “it’s a cat”. It requires a second effort of cognition to then explain why it is a cat – it has whiskers, paws, a tail and looks “cute”. As you point out, this is much harder.

      Probabilistic methods such as DNNs my never be able to deliver this insight. More formal logic approaches may prove more effective. YMMV.


  3. […] L’article original écrit par Tim Joyce a été initialement publié sur le blog Simplify Work […]

Post A Comment

Your email address will not be published. Required fields are marked *

To see how we protect your personal data, view our Privacy Policy.