At Transform 2019 last week, Microsoft CTO Kevin Scott sat down with Deepak Agarwal, LinkedIn’s VP of artificial intelligence, to discuss a range of topics. He touched on how machine learning is much easier to do now, where machine learning is heading, AI ethics, and of course Microsoft’s role in it all. But the most interesting part was arguably his explanation of AI democratization.

Agarwal asked Scott, “What does democratizing AI really mean?” Scott gave a long-winded answer (if you prefer watching above, the answer starts at 6:45):

Well, I think you all who are here at this conference today have a pretty good idea of what, AI in general, and machine learning in particular, is increasingly able to accomplish. When you imagine what’s ahead of us over the next few years, it’s not too much of a stretch to say that machine learning is going to have an enormous impact on the shape of the future of society.

If you believe that, that it’s going to be one of the foundational things directing the course of the future, then I think it’s unbelievably important that you think about the platforms that you’re building for AI. And making sure that everyone has access to those platforms so that they can use the techniques of AI to enhance their own creativity, to build their own businesses, to do their jobs inside of the companies that they work for.

I think about it a lot as sort of analogous to what happened in the early days of the Industrial Revolution. You had a little bit of Industrial Revolution happening in the late 18th century, and then the steam engine arrives and becomes this thing that massively changes the nature of productivity. Because it can do this broad new set of things that hitherto only human beings, and horses and mules and like very, very sort of physical labor had to do. It was like this brand new, powerful machine. And if you look at, like the early unfolding of the Industrial Revolution, the people who mostly benefited from this technology were folks who had the capital to like build factories and businesses around the machines and people who had expertise to design, build, and operate them.

But eventually, over very long periods of time, the technology democratized. You don’t get any sort of advantage now, as a capital owner, because you can build an engine. And what we, I think, need to do with AI right now is to dramatically contract that period of time where AI is so hard to do that only a handful of people can do it. We have to constantly be pushing to make it more open.

That’s really what I mean when I say democratizing. It has to be a level playing field. You have to have not just thousands or tens of thousands, but millions or tens of millions of people in a very short period of time using this new bag of tricks that we have to do their job. We probably have 40 million professional developers in the world. All 40 million of those professional developers need to be conversant with machine learning. Just like they can pass an interview where they can do a link list implementation, they know what a dictionary is and can analyze its complexity. You need to be able to reason about machine learning and have that in your bag of tricks. I think it even needs to expand beyond that.

After watching the rest of Scott’s talk, read our in-depth interview with Scott, which covers everything from why understanding AI is part of being an informed citizen to his upcoming book, titled Reprogramming the American Dream. The book asks the tech industry to make AI a foundation technology that benefits everyone.

When asked about finding the time to write a book, Scott chuckled: “I would not recommend trying to be a CTO of a company and writing a book in your spare time.”