Human Adaptability Still Needed in an AI-Driven World

If you’re one of the countless people wringing their hands over the prospect of artificial intelligence (AI) taking over the world, you might want to take a closer look at the kind of world AI is supposedly inheriting.

It’s a world where human flexibility and adaptability are needed more than ever.

AI can outperform humans in various ways but dealing with unrelenting change is not one of them – especially the fevered variety we are now subject to every day.

The stop-go regime of President Donald Trump is a prime example. Just in the week or so preceding the writing of this blog post, the US slapped tariffs on imports of steel and aluminum from Canada, Mexico and the European Union, the on-off summit between President Trump and Kim Jong Un, the leader of North Korea, was apparently on again, and early talks began on a summit between President Trump and Russian President Vladimir Putin.

But the current occupant of the White House is by no means the only source of the extreme turbulence that is buffeting the world. In Europe, for example, the Brexit negotiations seem to deliver a daily diet of surprises, there are fears that a new governing coalition in Italy will stoke resentment of the country’s eurozone membership, and the Spanish Prime Minister Mariano Rajoy was deposed in a no-confidence vote. The Middle East remains intensely volatile.

There is little respite in the commercial world. The rapid pace of technological change and the ongoing advance of e-commerce are just two of the forces that are restructuring markets and rewriting the competitive rulebook.

No company is immune from these changes; not even long-established ones. For example, as the Wall Street Journal reported recently, the departure of longtime CEO Denise Morrison is a symptom of the crisis facing the 149-year-old Campbell Soup Company, which is being hit by a new generation of small brands unfazed by Campbell’s formidable supply chain reach.

AI has a role to play in these tumultuous times – but so do humans. Increasingly so.

It’s important to keep in mind that AI is essentially a backward-looking technology. Machine learning, a key application of AI, is based on the notion that machines can learn from data. Give a machine enough data (although the data must be clean and formatted correctly) and it can learn extremely quickly from experience and project forward – as long as the past’s rules have not changed.

But what if the experience delineated by the data is not a complete reflection of reality, or is rendered largely redundant by a fundamentally new situation?

An example of the first issue is the failure of facial recognition systems based on machine learning technology to accurately read African-American faces. Researchers have found that while algorithms are able to read white faces accurately, they are less competent when scanning dark-skinned faces. The reason is that darker skin tones tend to be underrepresented in the training data used to create the algorithms.

Extreme changeability of the kind we are now witnessing also can affect the performance of machine learning algorithms. Learning from vast amounts of data on what occurred in the past is all very well, but what happens if the future presents situations that are unprecedented? Experience might not provide a template for how to react.

This is where the versatility of human decision-making excels. While people use their experience of past events to make decisions, they also are wired to rapidly reboot when confronted by a totally new set of experiences.

It’s important not to forget these capabilities in the rush to adopt AI or, for that matter, digitization generally. Digitizing business processes is already delivering significant benefits, and will no doubt become more powerful as companies gain more experience and the component technologies mature.

However, taking people out of the equation is not always a good idea at a time when we need them to help us navigate through extremely turbulent times.