ANGELA ZUTAVERN the Mathematical Corporation
An expert in machine intelligence describes the potential of ‘the mathematical corporation’.
You believe that the world of leadership has hit an inflection point. How so?
As useful as popular mental models and heuristics are, machine models now outstrip human performance in about half of the portfolio of cognitive tasks. Going forward, we will be able to beat machine models less and less often — except in the realms of imagination, creativity, problem-solving and some kinds of reasoning.
Enter the ‘mathematical corporation’. How do you define this new type of organization?
In the past couple of years, there have been major advances in machine learning, bringing massive new potential across industries. The mathematical corporation is a term that describes how organizations will need to operate in the future, to embrace these advances. Simply put, machine intelligence — built on data science — enables us to see patterns, anomalies and associations that were previously unidentifiable, and this emerging ability requires a new form of leadership. In a mathematical corporation, people actively collaborate with machines. AI basically has a ‘seat’ at the boardroom table.
Is this about embracing Big Data?
Of course, leaders of mathematical corporations use analytics and Big Data as well as AI and other advanced technology. But the capabilities of the mathematical corporation extend beyond mining data sets — an endeavour that has focused narrowly on answering known questions by querying specific piles of data. The mathematical corporation also focuses on uncovering new questions by querying a universe of data. Searching for and answering questions outside the spotlight of conventional thought can provide knowledge about the future.
If the recent past was about analytics and Big Data, the future is about the ‘big mind’ of the mathematical corporation, which comes from combining the mathematical smarts of machines with peoples’ imaginative intellect. This is what will trigger the next leaps in organizational performance.
Does intuition have any role in the mathematical corporation?
Intuition has served leaders well — and still does — principally because the human mind absorbs and understands more detail and substance than we consciously know. But with each passing day, machines are catching up. So, when should you trust a decision to your gut, and when to data? Most of us recognize that biases, politics and wishful thinking distort our perception; but as machines start to comprehensively reflect the real world, their biases dissipate. So, we need both.
Former Rotman School Dean Roger Martin has said that data can only be used to understand the past—not to predict the future, because the future hasn’t happened yet. How do you react to that statement?
I think Roger is probably talking about the critical skill sets that humans bring to the table in our data-filled world, which include things like creativity and imagination, strategy-setting and vision. A machine is not going to do those things for us, but machine intelligence can certainly be very helpful in other ways.
For instance, when Intercontinental Hotels pitches a new offer to its loyalty club members, its marketers still devise the content, but the computer chooses which individuals get which offer. We will see more and more such partnerships between human and machine.
In its manufacturing processes to make vaccines, Merck uses a four-stage process. In the past, they looked at data for each of the four processes separately and optimized within that process. But they continued to have problems: Batches were going bad and they didn’t understand why. So, they brought together the data across all four processes to optimize across the system. That’s where they uncovered some really good insights. They had suspected that raw materials were causing the batches to go bad, but it turned out that wasn’t the case at all. It had to do with the elements of the fermentation process. The machine intelligence model predicted when a problem was about to happen, so that they could stop it from happening.
In manufacturing, companies are now able to predict when their machines will break down and do preventative maintenance before that happens; and in sports, we’re able to use sensor data to predict when an athlete is about to get hurt, what kind a pitch a baseball player is about to throw, and what’s the best way to train and condition an athlete for a certain type of performance. So, there are already lots of examples where data alone has been able to predict the future.
Are humans even required in a mathematical corporation?
Absolutely! I don’t believe the doom-and-gloom scenarios where AI takes over the world. Humans are absolutely critical for functions that machines just can’t do. Machine intelligence is great at things like number crunching, recognizing patterns, organizing information and remembering things; but, as indicated, people still rule when it comes to problem-solving, reasoning, creativity and imagination. Human leaders are absolutely critical for setting strategies, for asking the big questions, and putting together the right combinations of backgrounds to create a team that will come up with new breakthroughs and solutions. It’s not about choosing one or the other. It’s got to be both: The best organizations will have human and machine intelligence working together.
Machine intelligence enables us to see patterns, anomalies and associations that were previously unidentififiable.
What is your favourite example of a mathematical corporation?
One of my favourites is the Data Science Bowl, which set out to identify new ways of predicting and identifying lung cancer. We organized this with Kaggle — a platform for predictive modelling and analytics competitions — putting it out to a community of hundreds of thousands of data scientists around the world. They competed (as volunteers) to develop algorithms to find new ways of identifying lung cancer earlier than it can be identified today.
Some real breakthroughs occured. Several of the winners have gone on to receive funding to continue to research these ideas and ‘productize’ the algorithm, so it can be used more widely. I predict that we are going to see some incredible breakthroughs in the health arena. What is so cool is that the teams that are winning these competitions don’t necessarily even have a medical background. They’re able to learn enough from online tutorials to make breakthroughs that researchers in the medical community have been trying to make for years. It goes to show that when you bring together people with completely different backgrounds, you often get breakthrough ideas.
Describe how Ford Motor Company is taking steps to become a mathematical corporation.
It might be one of the oldest corporations in North America, but Ford is dedicated to embracing machine intelligence. Their new CEO, Mark Hackett, actually came from the automated vehicles area of the business — one of the key research areas for machine intelligence.
In a mathematical corporation, you have to constantly experiment, fail, learn and iterate. Ford has about 25 experiments going on at any one time. They want to learn everything there is to learn about transportation. In the past, like many other companies, Ford relied on customer surveys for feedback. But, of course, a survey is just a sampling: You never get a 100 per cent response. So now, they are using real data about actual consumer behaviour. The consumer doesn’t have to provide the data: The sensors in the company’s cars track and provide it. They’re learning a lot about how people move around in the world today, and it’s not based on what people say they’re doing, but what they are actually doing. The fact is, in most cases, sensor data is much more accurate than what people say.
What has to change about our approach to discovery in today’s world?
When we’re thinking about new products and services, we tend to reason deductively: We have a presupposed idea of what we plan to find; we might even have drawn a model on a napkin about the way something works. Through experimentation, we then check to see whether we’re right and establish our ‘rightness’ with a large degree of certainty.
Today, we need to reason inductively. We do this by examining the way the world currently works and inferring relationships between key elements in the system. We recognize that we can never say for sure that we’ve discovered the truth, but we can reach conclusions based on reasonable logic. While the conclusion of a deductive argument is certain, the conclusion of an inductive argument is probable, based upon the evidence.
We naturally engage in both types of reasoning, but when we’re creating and launching a new product, we often limit ourselves to thinking deductively, because we see that as the only way to gain a level of certainty. With machine intelligence, we increasingly have the opportunity to reason inductively and discover new knowledge that we never could have found on our own.
For example, in marketing, discovery in data sets can unveil patterns in customer defections and predict means to reverse them; in logistics, discovery can examine daily or seasonal flow of materials or products and guide transportation planning; in HR, discovery can identify unhappy employees and guide changes in organizational practices; in manufacturing, it can process warranty data for patterns of product failure to suggest engineering improvements. This type of discovery promises to turn businesses — and lives — around.
You also believe we need to learn how to ‘re-frame questions’ for this new era. Why is that so important?
If we don’t break away from how we’ve done things in the
Searching for and answering questions outside the spotlight of conventional thought can provide knowledge about the future.
past, we will never be able to make the discoveries that machine intelligence makes possible. One example that comes to mind is the work that the U.S. Census Bureau is doing. Over the years, it had always conducted census surveys in the same way, by relying on enumerators — people going door-to-door, collecting data. But recently, the bureau’s leaders completely rethought everything about how to conduct the upcoming 2020 ‘Decennial Census’. They realized that they can pull data to avoid in-person visits. For instance, sometimes they need to verify that an apartment building is still there; today, you can do that using satellite imagery.
In addition, in the past, enumerators were just left on their own to figure out which routes to take, what time of day to visit, what order to go in, etc. Not surprisingly, they often found that people weren’t home and had to make repeat trips. This time around, there will be a mobile app that enumerators can use to predict the best time to make a visit, in what order they should be made and what route to take. That is a key example of why it’s so important to be open to new approaches. And importantly, leaders need to recognize that in many cases, the new approach they develop will be completely different from the way they have has operated in the past.