We attended a machine learning conference last week, and while the organizers presented great information about how organizations can leverage machine learning to understand and better serve their customers, it was evident that the audience was struggling to understand even what machine learning is. This article should clear up that confusion so you can think through the implications of machine learning for your business and your industry. And make no mistake, machine learning is already creating winners and losers in most industries.
Let’s start with a definition of machine learning (ML): ML is training a computer to learn from experience. If a computer can learn from experience, it can make decisions without being explicitly programmed to make a decision. ML consists of two phases: the learning phase and the extrapolating phase. In the learning phase, the “experience” is fed to the computer in the form of massive amounts of data. As the computer processes the data it begins to recognize patterns and extrapolate those patterns to data it has never seen before.
Here is a simple example: Spotify uses machine learning to improve music recommendations. It does this by feeding massive amounts of data about people’s listening habits into its machine learning systems. The systems then detect patterns in listening behavior that allow it to recommend new music to you based on your music selections. The more people listen, the more learning data is created and the greater chance you will enjoy the recommendations.
We find another example in healthcare. Researchers can feed massive amounts of data, including genomic, lifestyle, behavior, geographic, and environmental data into learning systems to find patterns in disease development and build hyper-targeted drugs for a person’s exact situation.
On a smaller level, companies are starting to mix their customer data with public data (like that available from the U.S. government census) and use ML to find potential customers that look like their most profitable customers.
Many business leaders think ML is too expensive and complex a technology for their organization. To dispell that myth, take a look at this article from The Guardian in which the author, not a technologist but a journalist, was able to set up an ML server on Amazon for $.70/hour, train the server using content from Guardian articles, and run a rudimentary ML proof of concept. While the results are not jaw-dropping, the ease with which the author was able to build a prototype ML project should impress anyone.
So why is ML becoming relevant now? The math required to build ML systems has been around since the 80’s maybe even earlier. So why now? The short answer is the Cloud. All that data storage and processing takes tremendous computing resources. The cost of the data processing and storage for even the most rudimentary ML application was prohibitively expensive when you had to buy, configure and manage a large number of servers. Now that those servers are available from cloud service providers like Amazon, Google, and Microsoft on a “pay for what you use” model, everyone benefits from the financial and computing power economy of scale. This puts machine learning within reach of even an individual journalist with limited technical skills.
It is important to understand that machine learning is not just a marketing fad. Machine learning is that rare technology that will create more impact than the marketing says it will. It is also not a technology that the business can ignore while I.T. uses it to make magic happen. ML requires the business and I.T. to work closely together to get the right data in the right form and ask it the right questions.
While it doesn’t require an enterprise budget and doctorate in data science, machine learning does require a mature I.T. infrastructure to support the data. Want to know if your technology is ready to support your use of ML? Contact us at info@GreystoneTech.com or call 303-757-0779 and schedule a consultation today!
