Machine learning is one of the current buzzwords frequently found in articles about the future of business. Other terms with which machine learning is often associated include artificial intelligence (AI) and cognitive computing. The question is: Can machine learning really help your business? For many businesses, the answer is yes. Three professors from the University of Toronto’s Rotman School of Management, Ajay Agrawal, Joshua Gans (@joshgans), and Avi Goldfarb (@avicgoldfarb), make a keen observation about the value of technology. “Technological revolutions,” they write, “tend to involve some important activity becoming cheap.” They note, for example, that the dot.com revolution reduced the cost of search and communication. They posit the business world could be sitting on the cusp of new revolution involving what they refer to as “machine intelligence.” As I understand their use of the term, machine intelligence refers to machine learning and advanced analytics embedded within a larger cognitive computing system. They explain, “Today we are seeing … hype about machine intelligence. … Technological revolutions tend to involve some important activity becoming cheap, like the cost of communication or finding information. Machine intelligence is, in its essence, a prediction technology, so the economic shift will center around a drop in the cost of prediction. The first effect of machine intelligence will be to lower the cost of goods and services that rely on prediction. This matters because prediction is an input to a host of activities including transportation, agriculture, healthcare, energy manufacturing, and retail. When the cost of any input falls so precipitously, there are two other well-established economic implications. First, we will start using prediction to perform tasks where we previously didn’t. Second, the value of other things that complement prediction will rise.”
Machine Learning and Cognitive Computing
The simultaneous rise of big data and artificial intelligence is no coincidence. Artificial intelligence (especially machine learning) requires large amounts of data from which to learn. Big data, on the other hand, requires artificial intelligence to find hidden nuggets of insight in a timely manner. Anand Srinivasan (@anand_sriniv), founder of hubbion.com, explains, “In the last few years, machine learning techniques have proven to be incredibly effective for predictive and deep insights; when used with data analytics. Many companies keep big data as their biggest asset because it reflects their aggregate experience. After all, every partner, customer, defect, transaction, and complaint gives the company an experience to learn from. While in the recent years, many companies have focused more on how to store and manage all this data, it’s not just about the quantity of data or how it’s being stored. By combining data analytics with machine learning, companies can predict the future with their existing data and not just use it for historical analysis.” That’s revolutionary.
As the analysts cited above note, predicting the future is a complex activity which involves a dizzying array of data inputs. Cognitive computing systems can deal with many more variables than previous analytic platforms and they integrate both structured and unstructured data into the analysis. As Srinivasan observes, “The real value of machine learning comes from its ability to create predictive models which can guide an organization’s future actions and discover never seen before patterns. … Compared to traditional methods, machine learning can easily outperform all other forms of predictive analytics on speed, scale, and accuracy.”
A Brief Machine Learning Primer
Fiona McNeill (@fiona_r_mcn), a Global Product Marketer at SAS, and Hui Li, a senior staff scientist at SAS, explain, “Unlike standard algorithms that are designed to perform a particular task, machine learning methods are designed to learn how to perform a task — learning as they are exposed to data. … These models not only learn both from the initial building and validation, they also continue to learn dynamically — from ongoing feedback as the model is applied over time. … Learning methods include supervised learning, semi-supervised learning and unsupervised learning, and in one or more of the above can adjust learning through reinforcement techniques.” They go to note that machine learning is used in many solutions. “Machine learning,” they explain, “lies at the heart of many advanced intelligence solutions, from AI to deep learning neural networks to natural language processing (NLP) and cognitive computing.” Why is machine learning gaining traction in the business world? According to McNeill and Li, “We owe these breakthroughs to advances in inexpensive commodity hardware that can be chained together to form massively parallel computational environments. Machine learning software can now execute across hardware clusters, running learning processing in tandem — whether in-memory, in-database or both. These environments can hold all the big data necessary to feed greedy methods like deep learning. By centralizing input data, these systems give algorithms unprecedented maneuverability to cycle through neural layer iterations, test reinforcement rewards and fuse different types of data — while delivering answers at human-like speed.” In fact, humans can’t match the analytic speed of cognitive computing platforms.
Machine Learning and Your Business
“From Google search, to Tesla’s driverless technology and on to Netflix’s movie recommendations,” writes Elliott Yama (@ElliottYama), chief data analyst at Apttus, “we all interact with machine learning on a day-to-day basis. While the super companies such as Google and Amazon are leading the charge on pushing the boundaries of such technology, most businesses today can also leverage machine learning to drive efficiency.” He offers four best practices on how to implement machine learning in your business. They are:
1. Clearly Understand and Define Your Particular Business Problems. “While machine learning is on the cutting edge of software technology, it is important to understand the business needs that you hope the technology can solve, before you wander down the purchasing path. Clearly defining these problems across the company is a requisite to successful implementation. For example, if you are attempting to leverage machine learning to analyze and optimize your company’s churn rate, the first step you must take is to define the different factors to go into that churn rate. Once the business problem has been well articulated and understood throughout the company, your business will be able to properly track and collect the targeted data.”
2. Understand what Machine Learning Can’t Do. “Machine learning does not come out of a box to solve any of your company’s problems, nor does it improve processes in every type of company. Often, many small businesses are better paired with some simple rules, rather than cognizant, and expensive, machine learning solutions. Going in, companies must know what is possible — the technology relies on large data sets (if this is something you company lacks, focus on building up you data as a preliminary to investment), and it will not be 100 percent accurate (Is the level of accuracy acceptable in your business?). While machine learning might be the hot topic around the business world, it is not for everybody.”
3. Ensure Data Cleanliness. “Ensuring the health of the data sets that the machine learning algorithms will be drawing from is vital toward providing insights that are accurate and trustworthy. Effective machine learning models are 100 percent contingent on a foundation of well-prepared data.”
4. Monitor and Update. “Like every new piece of technology meticulously implemented throughout an organization, measuring the effectiveness and iterating improvements for machine learning technology is vital and never ending. While one must allow the technology to run through its paces (the outputs of machine learning are never immediate), gauging the overall effectiveness will give indications to the accuracy of inputs. In some cases it may be necessary to go back to square one — redefining your company’s individual business problems.”
Srinivasan concludes, “Machine learning when combined with data analytics can impact the world in truly meaningful ways. Since it runs on a machine scale and is data driven, it can easily extract value from disparate data, with way less dependence on human input. Also, unlike the traditional methods, machine learning actually thrives on big datasets, which means the more data is fed into a machine learning system, the more it will learn, and the better results it will deliver.” Agrawal and his colleagues add, “As machine intelligence improves, the value of human prediction skills will decrease because machine prediction will provide a cheaper and better substitute for human prediction, just as machines did for arithmetic. However, this does not spell doom for human jobs, as many experts suggest. That’s because the value of human judgment skills will increase. Using the language of economics, judgment is a complement to prediction and therefore when the cost of prediction falls demand for judgment rises. We’ll want more human judgment.”
 Ajay Agrawal, Joshua Gans, and Avi Goldfarb, “The Simple Economics of Machine Intelligence,” Harvard Business Review, 17 November 2016.
 Anand Srinivasan, “Why Machine Learning is the Future of Data Analytics,” Datafloq, 20 September 2016.
 Fiona McNeill and Hui Li, “How Intelligent Machines Learn to Make Sense of the World,” Datanami, 31 January 2017.
 Elliott Yama, “4 Best Practices For Implementing Machine Learning In Your Organization,” Information Management, October 2016.