President's Commentary: Developing a Road Map for AI Is as Vital as the Research Itself

Artificial intelligence (AI), and its companion discipline, machine learning (ML), have emerged as the key to future high technology innovation and exploitation. We as a nation must recognize the value of an AI environment and promote its advancement as it increases its influence on all aspects of our lives. Failure to seize the initiative with AI could leave us at the mercy of foreign rivals or even AI itself.

AI is not new. Mathematicians and computer scientists have been exploring it since the 1950s. Interest in it waxes and wanes. But over these past six decades, it has changed. Now, AI is here to stay and is part of our technology roster. The question facing us is how we use AI.

Both AI and ML are critical because of the amount of data we have to use in our daily lives. That data tsunami is likely to grow, and we will need AI to sort through it and identify which data is relevant for whatever purpose we need—whether military operations or our daily lives.

AI is not a monolithic capability moving forward in a steady march into the future. Research into its development can be divided into two tiers. The first, which we are undergoing right now, focuses largely on its ability to shape practical applications. The second, which lies some time in the future, will transform AI research into more esoteric areas in which it will have a different meaning. The transition between these two will appear to be a natural evolution, but it actually will be more of a revolutionary leap leading the way to a broader and more uncertain future.

For its current phase, AI is serving as an enabler for solutions. AI development is being driven by need rather than by the state of the possible. It not only must sort through terabytes of data for relevancy but also must be able to choose courses of action from that data to support increasingly complex systems.

In this vein, it is important for AI to be able to give us some sense of predictability. It must be able to pick up on trends and repeatable actions and present that information in a useable form, either for systems to react to or for humans to act on.

Above all, a significant challenge looms. AI must be trustworthy enough that we can believe the veracity of the processed information it presents. Different organizations must be able to share the same data to build a common operating environment, and we must be able to trust what we’re seeing in the results that AI and ML are giving us.

AI also cannot be expected to solve everyone’s problems out of the box with minimal risk. Organizational managers should introduce AI into their programs in small steps. That will help flatten the learning curve while building familiarity within the workforce. And that workforce can expect major changes in duties and processes as AI assumes key roles—some of which humans cannot do. All these changes will take place in the shadow of continuing research into AI that pushes its capability horizon far afield.

Today, almost everything at home is programmable from a smartphone. That was unimaginable only a couple of decades ago. Tomorrow, AI and ML will take us to a new level. We must think about the unimaginable, and our emerging leaders with their new ideas will play a big role in defining that reality.

I don’t share the fear of many others that AI will someday establish supremacy over us. There will always be a human in the loop to partner with it and keep it under control. But I do share the fear that we risk losing the race to develop and tame AI to researchers in other nations, particularly China. It is imperative that we continue to push AI and ML work at full speed. Whoever controls AI controls the future.