The Limits of Artificial Intelligence

The 2018 FIFA World Cup was an unqualified success. While the number of goals scored per game was the same as four years previously in Brazil, the entertainment value was way ahead, as were the number of close games.

One of the pre-tournament favorites, France, won, and deservedly so. Still, luck played a significant role: both in the absence of technology in the build-up to the first goal, when a free kick was incorrectly awarded, and for the second – the result of a controversial VAR penalty decision. This served to remind us that technology is only as good as the humans using it.

Prior to the tournament, all the big investment firms used data analysis and artificial intelligence to predict the eventual winner. Goldman Sachs ran over a million simulations and concluded that Brazil would emerge victorious. Another corporate giant, UBS predicted Germany would win after running ten thousand virtual tournaments through its software.  Well Germany was knocked out in the group phase, while Brazil only made it as far as the quarter-finals, only to be knocked out by Belgium.

Recognizing the limits of Artificial Intelligence (AI), and technology in general, is important. Predictive tools may be excellent at setting the correct price for a product, or service, according to consumer demand and competition (full disclosure – one of my clients is an award-winning pricing analytics firm), or at predicting how many teachers will be needed in a given school district. But when it comes to predicting the future of our economy (the Crash), our culture (#MeToo), our politics (Trump), or even a sports tournaments like the World Cup, it does not perform so well.

Yet there is no shortage of effort to turn messy and complex scenarios, into neat algorithms and software programs, in an effort to control or predict those variables that make us human. 

For example, human resources departments increasingly use AI and data analysis to track and predict employee performance, run training programs and vet potential employees.  Political campaigns also use past behavior and demographic data to bombard us with targeted messages, playing to our fears and hopes. In the echo chambers of social media, we are constantly subjected to commercials for products or services curated and micro-targeted to our (supposed) needs. 

There are those who argue the problem is not with the technology, but with the data – if we had access to better data than our predictions would be more accurate. This may be true, but only to a point.

This still would not take into account unpredictable occurrences, or the flashes of inspiration which can make ordinary people do extraordinary and unusual things.

The truth is, past behavior, and success, is no guarantee of future behavior or success.  Technology is most life-changing and effective when it is used as an enabler of human performance, not as a predictor. This is good news for those among us who are put off by the artificial constraints imposed on us in the daily doses of propaganda, curated specifically for each individual by machines.

The more we continuously train ourselves to think and act independently the more we prepare ourselves for an uncertain and unpredictable world.

I, for one, am hoping to be the next Croatia, the one that few saw coming…