Andrew Ng has accumulated a uniquely impressive list of titles. One of the world’s foremost experts on machine learning and neural networks, he has been a professor at Stanford, founder of Coursera, founder of Google Brain, and, most recently, head scientist at Baidu, guiding a massive, intercontinental team of researchers for the company’s AI efforts.

Since his departure from Baidu earlier this year, however, he has been quietly at work on something all his own. In June he announced a website, deeplearning.ai, which was then little more than an online “under construction” sign, consisting of a single, cryptic black page with the words “Explore the frontier of AI” and a promise of more to come soon. Now, at last, the project has officially launched, and Ng has pulled back the veil on his newest, and perhaps most personal, project to date.

We spoke with Ng to find out what brought him to this new project, and where he hopes it will lead.

As it turns out (and contrary to what some might have been expecting), Ng is not launching a business venture of his own. Deeplearning.ai is instead a series of five courses being offered in partnership with Coursera, each meant to be completed over about a month or less with just three to six hours of commitment per week.

They are not pure beginners’ courses (enrollees are meant to already have familiarity with Python, for instance), but otherwise are intended to serve as a gateway to understanding such deep learning techniques as convolutional and recurrent neural networks and long short-term memory.

Neural networks have been studied for decades, but whereas the more traditional and basic type of neural network has three layers of figurative neurons—an input layer, output layer, and a so-called hidden layer in between them—within the last decade or so new classes of neural networks with multiple hidden layers (hence the “depth” that gives deep learning its name) have come to prominence, overcoming challenges in areas such as computer vision and language processing that had stymied other approaches for years.

That is largely what has catalyzed the surging interest, and faith, in the potential of AI over the last few years, and deep learning has already found practical and commercial applications. It is what lies behind Amazon’s Alexa and similar voice assistants, Baidu’s self-driving technology, and most advanced facial recognition.

Ng has other, previous course offerings on machine learning through Coursera, but as he explained, machine learning is a vast field with many specialties, and those earlier courses are meant to provide more of an overview—breadth, rather than depth. Ng says that the courses being presented through deeplearning.ai, on the other hand, are more focused, and also perhaps a little more purposeful. Through these courses, people should come away not just with a general knowledge of machine learning, but the practical skills to at least begin putting deep learning to use in applications of their own.

The question, though, is why this? Given Ng’s background, when he left Baidu he could have had his pick of jobs with any world-class tech company, or even founded one of his own. Yet instead he has chosen to launch an educational project.

It isn’t that Ng is stepping away from active research and development work entirely, and he noted that he has other projects he has been working on, though he isn’t quite ready to reveal them yet. “The other thing I’m trying to do, and this is more business oriented, is build some of these AI projects myself,” he said, adding that he still has an interest in working internationally. “One aspect I’m very interested in is collaborations with China … In some ways I think AI will take off faster there than in the United States, both because of the culture and the government support.”

“I think the US is still ahead in basic research, basic algorithm development. I think China is actually much better than the US in taking ideas to market, building things that work and shipping things to product. The US and China should both learn from each other.”

Meanwhile, Ng has been asked to advise any number of companies working towards AI. He currently serves on the board of directors for Drive.ai, a self-driving technology startup. “I think the team has invented some ideas and technology that I don’t think any other company has invented yet.”

“I think there are lots of good opportunities [for AI]. I think some industries are more ready for that transformation. I think healthcare is exciting, I think education is exciting. There are a few industry verticals where … the options to help people are already quite clear.”

If teaching, rather than doing, should seem somehow humbler than all of those alternatives, it may ultimately prove more consequential. Aside from any research or development work he has done personally, Ng’s greatest contribution to the field of AI thus far may well be in having built several of the world’s leading teams.

“Compared to other executives I think I invest more in training the teams that I manage. I’m really proud that all the teams I’ve built have continued to do great without me,” he said, citing those at Coursera, Google Brain, and Baidu for their ongoing success even after his departure.

And that, perhaps, best reveals the value of Ng’s deeplearning.ai project. If Ng’s plan succeeds, and by turning to online courses to broadcast both his message and his knowledge on AI he is able to enlist hundreds of thousands or millions of developers into creating AI applications, the impact will be larger than that of any one company, or even any one industry, and could help push practical knowledge of AI beyond the elite tech hubs of the US and China to every other part of the world.

“I’m going to say something that might sound cliché, but is completely sincere, which is that the thing that excites me most these days is developing an AI powered society. Ten years from now I want to live in an AI powered society where all of us have access to self-driving cars, where access to healthcare is much more affordable and much higher quality, where every child has access to a personalized education. But to make that happen, no one company can do all the work that needs to be done. So I think we need millions of people all around the world to understand the truth of deep learning.”