Machine learning is a powerful computational technique based in mathematics and statistics. First proposed in the 1950s, it began to blossom in the early 2010s as computers finally became powerful enough to work with the volume of data and complex calculations that the technique requires. Machine learning is one specific tool in the realm of artificial intelligence.
Rather than the traditional method of providing step-by-step instructions to program computers, machine learning teaches computers how to complete complicated tasks by providing examples and training them. This approach enables computers to do things that they could never do before.
So what does that make it good for?
Machine learning is such a powerful tool that it can do all kinds of astonishing things. It enables computers to:
- Understand speech, both words and meaning. This enables voice-driven agents like Siri, Alexa, and Cortana, and organizations use this capability to automate customer service or transcribe speech.
- Make predictions or extrapolations based off of historical data, such as stock prices or customer buying behavior. This enables organizations to anticipate and act on things before they happen.
- Analyze images and video. This has all kinds of applications, from optical character recognition (OCR), facial recognition, and analysis of medical imagery to detect disease.
- Classify interesting things (either good or bad) in data such as spam, network intrusions, or fraudulent credit card transactions.
- Make recommendations for new products, services, or actions.
That’s a pretty wide range of things we can do, and we’re just scratching the surface.
Fundamentally, machine learning solves problems and performs tasks that are too difficult, expensive, dangerous, or tedious for humans.
So if those are the kinds of problems that machine learning is good at solving, what do those problems look and feel like in your world? After all, this is a relatively new thing, and things were just fine before it arrived. Why should we worry about it now?
These are some of the symptoms that you may be experiencing if you’re in a situation where machine learning can help:
- You’re spending a lot of time and money doing manual analysis, and/or you’re not sure if the results are any good.
- You’re not catching fraud or errors, and then when you eventually find them it’s panic and chaos.
- You feel like you’re sitting on a vault full of useful information if only someone had the key (or it’s all in Excel).
- You’re missing opportunities to cross-sell to your customers or clients (leaving money on the table).
- You feel like your competitors are outmaneuvering you and that makes you nervous.
- You see an opportunity to outmaneuver your competitors, but you don’t know how to make it happen.
Perhaps you’ve already identified machine learning as something that can help you succeed, but don’t know what to do next. You may have read books or articles on the topic, started an online course, or tried delegating it to the IT team in the hope that they understand it. You may have poked at a product or two. Frequently, we find that the issue is so big that many organizations do nothing, waiting for clarity to appear.
Even worse, there are some paths that most organizations should avoid:
- Buying a product that doesn’t actually achieve your goals
- Hiring a generic IT staffing company that doesn’t understand machine learning as a specific discipline and area of expertise
- Hiring a gigantic “we do everything” consulting company
- Trying to go it alone with insufficient skills
Machine learning is not
for everyone. Click here
to read about the four prerequisites for any
machine learning project to succeed, and learn about the unique approach we use to ensure that every machine learning system we build creates astonishing results.