Tweet about this on TwitterShare on FacebookShare on LinkedInShare on Google+

This is part 2 of our series deconstructing the different components of artificial intelligence. In case you missed it, check out our primer on natural language processing (NLP) in part 1!

What is machine learning, really?

“Machine learning” and “deep learning” get thrown around a lot, whether in technical articles or as hashtags on Twitter. Machine learning refers to the specific technology that comes closest to matching our pop culture perception of AI. HAL 9000, SkyNet, The Matrix: machines able to enhance their own intelligence and capacity without human intervention. In the movies, machine learning may lead to disaster, but in practical applications, it leads to tremendous gains in efficiency.

Think of it as a combination of complex pattern recognition and advanced game theory. We program machines to recognize patterns in large, often unstructured groups of data, then teach them how to make assumptions or predictions by comparing those patterns to previous patterns encountered. The “learning” happens when they test their predictions, evaluate, and refine their original assumptions. Imagine the scientific method, but at a rate of trillions of decisions per minute.

Why does it matter?

When you can program something to  apply human-like intelligence at inhuman scale and speed, you can begin to tackle previously insurmountable challenges like real time analytics in video games, or analyzing the trillions of data points mobile phone users generate every day.

Machine learning is the difference between smart programming and true artificial intelligence. It’s a background technology, essentially a complex network of algorithms, but you can see its fingerprints on many familiar applications:

  • Financial fraud prevention: looking for the imperceptible irregularities among huge swaths of data
  • Marketing automation: creating and sending hyper-personalized communications and optimizing communication intervals
  • Adaptive websites that change content based on real time user behavior

Machine learning for your business

For a familiar example of machine learning in the wild, take a look at how IBM is marketing Watson (not just a game show gimmick anymore!). Nearly any business system has the potential to become a learning system. While some require vendors who specialize in specific industries or your applications, there are more turnkey solutions available to help transform your marketing, sales and customer service data into valuable insights.

Part 3 is coming soon! In the meantime, check out part 1 of our series on the building blocks of AI.

Posted by on Thursday, April 13, 2017.


Leave a Reply