AI Not Only Learns, but Must Be Taught

If you’ve been following the news lately, you know that the media is obsessed with reporting on AI, and particularly people’s reactions to it. Some sources like The New York Times are doing excellent work in covering the topic and its growing influence, but most are, predictably, giving us little more than celebrity soundbites and dystopian predictions.

Being in the middle of this chaotic news cycle—are we doomed? are we saved?—can be like sailing a rowboat into a hurricane… without a paddle. While I’m certainly no better than the next person at fortune telling, I hope to provide you an oar (with a side of pragmatic positivity): AI will not doom us, but neither will it save us. Like all things, the answer is likely somewhere in the middle.

The reality is that we humans are no better at predicting the future of AI than we have been at predicting the future of any human invention. Robots and automation have threatened to replace human workers for at least a hundred years—“Man Devoured By His Machines” reads a 1921 headline—and yet we remain remarkably, persistently employed. We have a long history of distrusting that which we don’t fully understand: the lightbulb, the typewriter, the telephone… the list is obviously a mile long. Understanding that we are responding to AI the same way we have responded to new technology throughout our history is the first step in realizing that you needn’t embrace crippling anxiety and fear as the only way forward.

Having said that, we must also understand what does make AI unlike the inventions of the past: it’s the first man-made technology that has the ability to learn. Up until this moment, we have created machines to do something, and they have done it, at speeds much faster and accuracy rates much higher than humanly possible, but still doing what we programmed them to do. With AI, we’ve created a machine intelligence that can and does learn, which is a significant differentiator because it means it not only does what we prompt it to do, but it can and will do much beyond that. In short, there is a whole lot more unpredictability compared to the inventions of the past. And therefore the stakes are higher—in positive ways and negative ones. And we must respond by developing new ways to deal with it.?

So if we understand that the way to view AI is with neither unbridled exuberance nor incapacitating fear, how should we stay informed and find a balanced viewpoint going forward? My philosophy is to treat AI like you would a child.

Because AI is the first technology to learn, it’s also the first that we need to teach—I would argue that we have a responsibility to teach. I have two kids at home, ages 3 and 5, and they can be both joyous and wild. When one of them throws a tantrum or starts wailing on the other one, I don’t foresee this as a prophecy that he/she will be a violent sociopath as an adult! I view it as a teachable moment—I need to do better, so they can do better. And as they do better, I can adapt the way I teach them.

This, too, is something we must understand about AI: We don’t have all the answers now, because we must be nimble enough to change our course of action in order to respond to issues and tendencies. We’ve seen this play out already, in how we’ve already witnessed the biases that exist in AI’s current iterations. It’s been based on human data, and humans are biased. So we must course correct, and teach it better. I fully support rules and regulations here—in the same way we use those to guide our kids—but I guarantee that restricting AI’s use too forcefully, whether that’s in the government or schools or at work, will never be successful. We run the risk of regulating it to the point of minimizing our own creativity in its applications.

And that creativity is precisely what we should be focused on. If you’re a leader weighing the decisions around the use of AI, I want you to get excited about the opportunities for augmentation here. AI can positively augment all functions in all industries from art to hospitality to health to software development. This is our opportunity to be creative around these possibilities and start piloting AI for efficiency and speed gains.?

But you must be prepared for the displacement stage post augmentation, where it’s inevitable that AI replaces some, or a lot, of the jobs in the same way we’ve seen happen with so many technologies before this. The impact of this stage can be dramatic, on a strategic business level and an interpersonal one as well. Currently, it might not be apparent how to prepare for this phase but ignoring (or straight out denying) this phase will help no one.??

But the final stage is the creation stage, where AI can, and in my view will, help us create new jobs and new business opportunities in current and yet undiscovered industries. The key to this phase is value—we must employ AI to create new value for humanity.?

The bottom line here is that how we view AI, how we utilize it in our personal and professional lives, is a choice, and as leaders (and parents!) we have an obligation to mitigate the challenges and amplify the opportunities. I do believe the positive will far outweigh the negative, but I know that no new technology is all “good” or all “bad.” Our relationship with figuring out how to teach AI is a multi-year journey, perhaps a multi-decade one, and we are only at the beginning. We are teaching, but we are also learning—just like parenting.