Connecting Consumers to Artificial Intelligence Positively

I've been working on the development and practical application of artificial intelligence (AI) for more than thirty years at some of the largest and most advanced organizations today. In that time I've learned some simple truths:

·        Just because something is “artificial” does not make it ethically neutral.

·        AI is only as useful as the customer's reception to the improvements it offers them.

·        Organizations are ultimately accountable for how their systems behave.

AI becomes “intelligent” only by learning from us humans and the data we create. Because our biases, intentions, and preferences can imprint and shape so easily how AI behaves, our responsibilities are higher than ever. Who will be sued when an autonomous drone collides with a self-driving car? Is an insurance organization criminally accountable if its weather insights app fails to predict a tornado and people are injured or killed?

These concerns are real warnings, but even as I acknowledge these risks, rest assured my blood type is AI-positive. I studied AI in college in the 1970s during the deepest days of the AI winter. Back then, we wanted to teach computers how to play chess like a human. We tried, using declarative and inferencing programming techniques with a miniscule fraction of the computing power available today. (Today, we have commercialized aspects of AI, but we've changed what we call it: from that early AI work came “expert systems,” “business rules engines,” and “declarative programming.”) Still, we have a great deal to learn about how to bring the benefits of our technology to society.

Shortly after, in 1997, IBM’s Big Blue beat then reigning world chess champion Gary Kasparov under tournament conditions. In 2011, IBM’s Watson beat the world’s best Jeopardy players. Last year Google’s AlphaGo beat world champion human Lee So-dol at Go.

From these powerful confrontations of our greatest "minds" we've learned that, clearly, if AI is to succeed, we need to recognize that formative truth (with a nod to Marvel Comics' Stan Lee): With all great power comes great responsibility.

We can learn this lesson from what Gary Kasparov told the world about his loss to Big Blue in 1997 and shared with me in conversation. He is convinced that IBM used a “Mechanical Turk” play with Big Blue before the final match of the tournament. (The expression the “Mechanical Turk” is based on Wolfgang von Kempelen's mid 1700s clockwork mannequin that housed a small, human chess player pretending to be an intelligent chess computer. While it did defeat Napoleon, it should be stressed that dressing the mannequin as a Turk can be seen as racist and as offensive as a dime store Native American “Indian.") Between matches, Kasparov believes humans earning IBM paychecks modified the program to simulate sufficient unpredictability to put Kasparov off his game. What put Kasparov off was that Big Blue suddenly began behaving differently, making him wonder who he was playing against.

The other side should have disclosed the ruse. Kasparov has nothing against humans working with computers, in fact, he is a proponent of "advanced chess," which pairs humans coupled with AI chess bots playing other humans coupled with AI chess bots. The point is transparency, disclosure, and fairness are essential.

With consumers, our use of AI must be transparent, as that’s what they want. We must embrace the fact that "AI plus a human" is better than either of the two alone. As such, we must look for ways to leverage this new technology to make workers more productive, but not render humans replaceable. We've actually found that doing so is, quite frankly, bad business.

In a recent global survey of 6,000 consumers, we found that nearly 70 percent were open to using AI to make their daily lives easier and 40 percent believed AI can improve customer service, but 80 percent still preferred to talk to a human. The respondents were more open to the combination of AI and people, but they were suspicious of AI unchecked.

There is a time and place for opaque AI vs. transparent, and considerations for businesses about what to use when, as Rob Walker eloquently discussed recently. But that's a topic for another post.

Businesses still have some work to do to gain the consumer’s trust in their use of AI. We're asking people to let us into their lives in an entirely new way. The unfamiliarity of it acts as a block to our connecting with the consumer. We can only reach them by showcasing the qualities both we and they find essential: transparency, disclosure, and fairness.

Note: This article was originally published in Forbes France.