Is Google Really That Smart? (Or Why Decision-making Based On Averages Is Dangerous)
As artificial intelligence (AI) expands to fill every part of our lives, we are getting increasingly comfortable relying on it. For example, we buy a product on Amazon and get a glimpse of other products that Amazon recommends. Many of them are indeed of interest to us….and we click to see more. There is a lot of value in this feature. Amazon knows what other things we have purchased or even only looked at. It knows what other people with similar interests have purchased. There are numerous books I have purchased as a result of their recommendations. This AI feature of Amazon creates value for me…and Amazon.
When I moved to Las Vegas a couple of years ago Google Maps became one of my most useful apps. Over time I became frustrated with it when I realized I was no longer able to make up time on 215 driving from Summerlin to Henderson. Then it dawned on me that Google was learning my driving habits and calculating the ETA based on how I drove, not just average speeds. Upon realizing this I now knew that I could no longer rely on making up time on those longer trips. On the surface this is pretty smart AI on the part of Google.
But there is one driving situation where I have learned to ignore the route advice of Google Maps. On trips to certain locations in the Henderson area, Google suggests taking 95 rather than 215. 95 is significantly shorter in terms of distance and a little bit quicker in terms of time. However, experience has taught me that there is greater risk of slowdowns on the 95 route. Google selects a single point forecast of drive time ignoring any history of variance for that particular route. I choose to take 215 knowing from experience that the 3 or so extra minutes is often going to be significantly less than the 30 extra minutes if there is an accident on the suggested route.
Google Map’s AI engine does not factor in the probability distribution for drive time on those two routes. Common sense tells us that the range of drive times is greatly skewed to the high side for travel time. Experience tells those of us who frequently drive along 95 that Google’s single point estimate of drive time on that route will more likely be an underestimate than the drive time on 215. No doubt Google’s AI will eventually take this probability into account, possibly even offering a setting to take into account your personal risk preferences. But for now, we all can find examples of how our own common sense and experiences are “smarter” than Google’s.
In arriving at your own business decisions, don’t make the same type of mistake that Google Maps does. We live in an uncertain world. Decisions typically involve forecasts. Too often, we rely on a single point forecast. Consider the probabilities surrounding that single point forecast. How wide is the range? Is it skewed to one side? Often, we call it intuition when we unconsciously take such things into account. In reality, our brains are quickly recalling past experiences. Add to that recall a careful consideration of other factors that can influence the distribution of possible outcomes.
One way to gain a broader understanding of these other factors is to join a CEO peer group. This gives you access to the experience and knowledge of other business owners and CEOs. The unique perspective each member brings can improve your awareness of the variability in potential outcomes.