What do successful Kubernetes migration projects have in common? A clear strategy, a strong culture, and the proper resources to execute the plan. Check out this expert advice
How to explain machine learning in plain English
What is machine learning? What is ML vs. AI? What data problems could ML solve for your organization? Here’s how to discuss the key issues in plain terms
Machine learning is already pervasive: Most people probably don’t realize it.
“Whether or not you know it, odds are that machine learning powers applications that you use every day,” says Bill Brock, VP of engineering at Very. “Machine learning has revolutionized countless industries; it’s the underlying technology for many apps in your smartphone, from virtual assistants like Siri to predicting traffic patterns with Google Maps.”
Perhaps you care more about the accuracy of that traffic prediction or the voice assistant’s response than what’s under the hood – and understandably so. But as machine learning use cases continue to increase, you will find yourself needing to explain at least the basics of the technology to folks outside of IT, whether it’s to get buy-in, to showcase the work of your team, or simply to build better communication and understanding between departments. Your understanding of ML could also bolster the long-term results of your artificial intelligence strategy.
If you’re not using AI or ML yet, you soon will be evaluating its potential. “AI as a workload is going to become the primary driver for IT strategy,” Daniel Riek, senior director, AI, Office of the CTO, Red Hat, recently told us. “Artificial intelligence represents a transformational development for the IT industry: Customers across all verticals are increasingly focusing on intelligent applications to enable their business with AI. This applies to any workflow implemented in software – not only across the traditional business side of enterprises, but also in research, production processes, and increasingly, the products themselves.”
[ Read also: AI vs. machine learning: What’s the difference? ]
What is machine learning? Use examples
It’s not just maps or virtual assistants. Consider this example from “An executive’s guide to AI,” our recent research report conducted by Harvard Business Review Analytic Services. The report highlights how machine learning was used to solve a problem at Beth Israel Deaconess Medical Center: Its operating room capacity was stretched thin.
“Machine learning using data from a million patients – including OR times of the past, procedures done, and patients’ disease, gender, age, comorbidities, medications, etc. – determines how much OR time is needed for any given patient,” the report reads. The medical center freed up 30 percent OR capacity as a result.
This is not pie-in-the-sky futurism but the stuff of tangible impact, and that’s just one example. Moreover, for most enterprises, machine learning is probably the most common form of AI in action today. People have a reason to know at least a basic definition of the term, if for no other reason than machine learning is, as Brock mentioned, increasingly impacting their lives.
So let’s get to a handful of clear-cut definitions you can use to help others understand machine learning.
[ How does RPA fit in with AI and ML? Read also: How to explain Robotic Process Automation (RPA) in plain English. ]
Machine learning definitions
“At its heart, machine learning is the task of making computers more intelligent without explicitly teaching them how to behave. It does so by identifying patterns in data – especially useful for diverse, high-dimensional data such as images and patient health records.” –Bill Brock, VP of engineering at Very
“In classic terms, machine learning is a type of artificial intelligence that enables self-learning from data and then applies that learning without the need for human intervention. In actuality, there are many different types of machine learning, as well as many strategies of how to best employ them.” –Fran Fernandez, head of product at Espressive
“Broadly, ML is a subset of computer science which involves applying statistics over observed data to generate some process that can achieve some task. This encompasses both the structure of ML (taking data and learning from it using statistics) and the impact of ML (use cases like facial recognition and recommender systems).” –Michael McCourt, research scientist at SigOpt
Machine learning vs. AI vs. deep learning
These are good big-picture definitions of machine learning that don’t require much technical expertise to grasp. Things get more detailed – and more complex – from there. Brock notes, for example, that ML is an umbrella term that includes three subcategories: supervised learning, unsupervised learning, and reinforcement learning.
(Brock previously shared the difference between supervised and unsupervised learning with us in this story. He notes that reinforcement learning borrows from psychology experiments: “The machine attempts to find the optimal actions to take while being placed in a set of different scenarios. These actions may have both short-term and long-term consequences, requiring the learner to discover these connections.”)
You can also dig down into related sub-disciplines such as deep learning. (If you want to do just that, read our story: How to explain deep learning in plain English.) For folks outside of the IT field, though, this stuff can become confusing in a hurry. Which begs the question: How much do they actually need to understand about ML?
“I don’t think non-technical people need to understand the basics of machine learning,” says Fernandez from Espressive. “Instead, I believe they need to understand the benefits of machine learning. Rather than saying, ‘machine learning means xyz,’ they should say, ‘Because of machine learning, our enterprise has been able to achieve xyz.’”
[ Read also: How to explain deep learning in plain English. ]
How does machine learning work? Discuss data problems
If there’s one facet of ML that you’re going to stress, Fernandez says, it should be the importance of data, because most departments have a hand in producing it and, if properly managed and analyzed, benefitting from it.
“If you want to give yourself more time in the future and become more efficient by leveraging machine learning, you should think about the data that you generate as you work and how that data can be accessed and structured in a way that machine learning can leverage,” Fernandez advises.
Indeed, this is a critical area where having at least a broad understanding of machine learning in other departments can improve your odds of success.
“If people knew more about machine learning – maybe not the details, but at least the underlying concepts – then they would understand that ML does not ‘just work’ on its own,” McCourt from SigOpt says. “It takes guidance, structure, data, and time (especially in the case of big data), and it takes someone to interpret the outcomes, both during development and after deployment.”
Another motivation to help others understand the basics, especially in terms of the importance of data: Complete ignorance might increase the risk of bias and other issues. “It’s easy to blindly trust the results of a machine learning algorithm, but the results are only as good as the data the algorithm is trained on,” Brock says.
[ How can you guard against AI bias? Read also AI bias: 9 questions for IT leaders to ask. ]
How much explaining you do will depend on your goals and organizational culture, among other factors. But an overarching reason to give people at least a quick primer is that a broad understanding of ML (and related concepts when relevant) in your company will probably improve your odds of AI success while also keeping expectations reasonable.
“ML can solve problems, but your company adopting ML tools will not simply fix everything,” McCourt says. “ML, by itself, is simply the process of clustering, approximating, classifying, or designing; by learning some about the process by which ML works, less-technical people can realize that ML is only part of a fully successful process for making smart decisions and taking smart action.”
How can you best evangelize for ML? Let's look at some examples: