The Artificial Intelligence Primer

The machine intelligence market hot right now. There's a huge amount of excitement and hype around the technology. What we have to remember is behind that there really are true technological advances that's propelling machine intelligence into some meaningful commercial applications. And organizations of all types are seeing these advances and beginning to invest heavily both to advance the technology as well as deploy it broadly across the organizations. Simultaneously, senior executives are getting inundated from outside their firm by technology vendors and others with software and solutions that have lofty claims about what they're capable of doing with machine intelligence. In some cases, these claims are fact, and in others, they've been exaggerated to a little or large extent, and it's left up to the senior executive to ferret out fact from fiction, which is becoming increasingly difficult. We wrote the Machine Intelligence Primer to help executives and others understand machine intelligence and really separate the fact from fiction. We are deeply passionate about machine intelligence technologies, and we really want people to understand what it is, how it works, and how it's going to affect how we work and live. At the same time, we're a bit pragmatic and realistic, and it's equally important to us that people understand the limitations of the technology. Machine intelligence is not a panacea that will solve every business problem. Although it can add significant value to solving those problems if used properly. Our hope is the Machine Intelligence Primer provides its understanding and empowers people to create the future in which machine intelligence is harnessed for the collective benefit of all of our lives and our society. 

Machine intelligence has existed as a field of study for more than 60 years, so it's been around for a very long time. But there have been a few key developments that have enabled machine intelligence to really ascend in the past few years. The first of these trends will be obvious to most everyone. It's the proliferation of digital data. As all of us are aware, data now is being produced by all of our devices all the time from our smartphones to our connected refrigerators and even toasters. There's now a huge amount of data on which to train the machine learning algorithms that make machine intelligence work. The second major development has been advances in computer processing power. This is really important because for machine learning algorithms to interpret all of that data requires a huge amount of processing power. The third piece that's really enabled machine intelligence to take off is advancements in research around machine learning. When taken together, these three trends help explain why machine intelligence has seemingly overnight become so transformative and disruptive across such a wide range of sectors. 

Machine learning comes in two major categories. One is supervised learning; the other is unsupervised learning. Supervised learning is where you have examples of what you're trying to train the machine learning algorithm to find. Unsupervised learning is where you don't have any of that; you just have the data. It's interesting because you might find patterns that you never knew you were even looking for. Stuff in the data that was, as they say, unknown unknowns. But it's challenging because you can't tell whether or not the algorithm is doing a good job because there's no flash card to turn over and say, 'yeah, you got it right; you found what I was looking for.' 

Machine intelligence, or MI, has captured the public's fascination within the past few years. There are lots of problems that can be solved with machine intelligence, but there are some that don't fit so well into machine intelligence solutions. Let's look at a variety of the different tasks that are well suited for machine intelligence. For instance, if an organization has rote repetitive tasks that take up a lot of human employees' time, techniques such as robotic process automation can be used to automate those tasks. Tasks such as data entry or completing forms can very easily be automated using these techniques. Also, if an organization's data is largely pattern-based and requires human judgment to determine outcomes for that data, there's a good chance that supervised machine learning can be used to automate that decision making process. While that's the good news, there's plenty that machines are still not capable of fundamentally. Although machines can understand patterns in the data, they don't have a true understanding in a human sense as to the meaning behind data. As a result, machine intelligence isn't great for tasks that require complex judgment or higher order thinking. One of the weaknesses of today's machine intelligence algorithms is that the decisions that they make can't be readily explained. Finally, machine intelligence algorithms are data hungry. If you have a problem that you're trying to solve, but insufficient data or insufficient understanding of the data required to solve that problem, you may want to look for other solutions or invest time in building a larger collection of data to shape your problem space.