• Strategy

Artificial Intelligence for Beginners

‘Prediction Machines’ a new book by three Rotman professors demystifies AI for the confused and worried

Friday 25 May 2018


By downloading this resource your information will be shared with its authors. Full privacy statement.

For the last few years the talk has been on Big Data. The world has taken a step forward again – and now the buzz is on AI – artificial intelligence. The very phrase conjures images of Hollywood Robocops, Hals and Hers. But what is the reality?

The reality is that AI is undoubtedly coming to a business near you very soon and it will have significant impact on how your and other businesses do business. However, it will not be the all-singing and all-destroying automaton of our dreams or nightmares. As I describe in the book, 'Prediction Machines', co-authored with two colleagues from the Rotman School of Management, the Creative Destruction Lab, AI is essentially the ability we have to make use of the torrents of big data now flowing into many businesses to use complex arithmetic to crunch the data and make predictions from the patterns that emerge from that.


The co-authors of ‘Prediction Machines’, Joshua GansAjay Agarwal and Avi Goldfarb will be leading the AI Primer program for executives, in Toronto, with 1.5 day programs running in June, September and December this year... LEARN MORE HERE


When electric light was created it changed the world. Previously burning a candle or lantern for an hour had a material cost, approximately 400 times what it costs today to get the same amount of light. This meant that evenings could become productive for everyone, for the first time in history production was not generally restricted to daylight hours. We are entering a similar situation with prediction. Where the creation of predictive patterns would have cost an enormous amount in both data collection and analysis, now it can be achieved at minimal cost. This means that it can be done for seemingly trivial or as yet not fully understood outcomes, and so it will spread.

The complexity of AI is in the algorithmic code, not so much its results. As it advances, more data will be able to be inputted, and its breadth of understanding and ability to learn will increase. Put an AI machine in a car with a human driver, with the correct number of visual and audio sensors and it will learn that at junctions the driver puts the indicator on. What it will not be able to tell is whether the car can turn left or right without the indicator being used – until it sees this happening. The AI is restricted by what it knows.

AI is good at making predictions where there are known unknowns, it is no good at all where there are unknown unknowns and can be sent down the wrong track entirely if there are unknown knowns. Humans still have a considerable edge at dealing with ambiguity.

The London taxi-driver who has to pass a rigorous test on the quickest and best alternative routes around the city before getting their cab license, has been significantly impacted by the arrival of Uber drivers who rely on AI-driven GPS mapping. But get into a London cab and say “I need to get to that hotel near Madame Tussauds where Justin Timberlake stayed”, and the Uber driver’s GPS won’t help you, but the cabbie just might.

Equally computers are not great at interpreting statistics when the right questions have not been asked of it. Abraham Wald was a mathematician who was famously asked where the limited amount of armour-plating should be added to fighter planes in World War II and gave the counter-intuitive answer that it should be put where bullet holes had not punctured the aircraft. His reason being that, the aircraft being analysed had all returned, and so not been fatally damaged. The ones that failed to return had presumably been hit somewhere more damaging – and that is where needed to be protected.

It would be very difficult to program a computer to give this response as the answer was not in the data directly but the wider context.

For executives who are scanning the horizon for threats and opportunities to and for their businesses, it is important to have a solid appreciation of what AI can and cannot do. Undoubtedly it will continue to grow in scope, but for the coming years it is unlikely to be able to make value judgments or predict anything with data not clearly and logically linked to the core data set. The known knowns.

Rotman School of Management is Canada’s leading business school and has Canada’s largest group of management faculty. It is home to some of the most innovative research institutes in the world



Change Is Not Always Good


Industry Beyond Digitization

Two renowned authorities on strategy and innovation explain how the fusion of physical products and services with real-time data and AI will shape the industrial future


The World Beyond the Horizon


Decisions, Decisions, in the AI Era


Delivering Strategic Change

Google Analytics Alternative