Democratizing AI: It’s Time to Get on Board

by Adrian Bowles, PhD

The hype surrounding AI is everywhere, and it is still easier to put AI on the label than it is to put AI in a product or service, so buyers are right to be cautious.

The truth, however, is that some AI technologies are already becoming pervasive, and waiting for technologies to democratize AI and put it in everyone’s hands is ill-advised. As author William Gibson observed, “The future is already here—it’s just not very evenly distributed.”

Still, some see a ten-year horizon for AI everywhere. Here’s why we think that’s way too conservative.

Using AI: Machine Learning Is AI and AI Is More Than Machine Learning

AI is a broad discipline that includes technologies to perform tasks that we generally associate with human cognition: learning, understanding, and reasoning, and the ability to support biological I/O systems like perception (hearing, vision, touch, smell, taste). Today, the AI tail, machine learning, is wagging the PR dog. It is a false but useful equivalence.

Some benefits of AI are already available to anyone with a typical smartphone. Many people use AI every day, from speech recognition, to translation, to navigation apps that factor in weather and historical traffic patterns to enterprise apps that help an account manager decide which brochure to send to a prospect. So, almost everyone can use AI today without even realizing it.

Do I Need to Understand AI to Make Decisions with It?

I’ve been asked whether everyone will be able to create new, personalized solutions using AI in ten years. That’s a little more complicated. We have already seen great advances in tools to help people develop machine learning solutions without a detailed knowledge of machine learning itself. In the utopian view of democratized AI, these user/developers are often referred to as citizen data scientists. If you have enough good data, these tools will guide you through the analysis and help you gather insights from that data.

The technology behind tools like Watson Explorer from IBM or the recently announced AutoML from Google is sound. It won’t take ten years to have tools like this in the hands of everyone who wants to make evidence-based decisions using machine learning. Any knowledge worker who needs deeper insights from existing data should become proficient with these tools in three to five years. By then, this functionality will be integrated into even the most basic productivity tools.

So, What Will Take Ten Years?

The big shift in the next ten years will be toward true distributed intelligence, where the learning and reasoning tasks are performed throughout the global network on devices ranging from smartphones, to refrigerators, to automobiles, to military command centers.

AI-powered applications in the hands of everyone (everyone on the grid, that is)? It’s available now. The ability to create AI-powered solutions that are available to everyone with a modicum of technical or business acumen? Three to five years. Individuals at the center of their own ecosystem, with AI-powered devices monitoring their behavior and emotions, and responding with context-appropriate data and services? That’s a five-to-ten-year journey that has already begun.