Interview with Kelvin Lwin, NVIDIA Deep Learning Institute

After spending nearly a decade at UC Berkeley, Kelvin decided to repay his debt to the public education system by helping build UC Merced. He spent seven years teaching 4,500 students across 55 classes while redesigning the undergraduate Computer Science curriculum. He is now busy designing curricula at NVIDIA’s Deep Learning Institute (DLI) to democratize access to the latest technologies across many disciplines, industries and geographies. Kelvin helped DLI reach over 100K developers worldwide directly and in collaboration with Udacity and Coursera/Deeplearning.ai. He continues to search for ways to leverage AI to solve the Paradox of Progress.

At the Deep Learning Summit in San Francisco this January 24 - 25, Kelvin will be speaking on the Education and AI stage, exploring Human Insight Project - 21st Century Education System. In advance of his talk, we interviewed Kelvin to hear more about his current work:

How did you begin your work in AI and education? What came first?

I came to the US as a 12-year-old from Burma and was given a mission by preceptor monk to bring back the education system. Consciously or unconsciously, I’ve followed his directive by tutoring in High School, TAing (paying my way through college) in Undergrad, Teaching minor in Grad school and as “Teaching Professor” in UC Merced for 7 years. Then the opportunity to work with Deep Learning Institute came so I followed the adage that teaching is the best way to learn since I knew that AI was key to solving my life mission.

Tell us a bit more about your team and your goals at NVIDIA. How are you using AI?

Our goal is to teach all Developers, Data Scientists, and Researchers how to apply AI into their domains. We teach fundamentals but specialized in Industry Verticals where we see rapid adoption of AI. Our job is first to teach about how to even think about AI and make decisions about where to apply; equally important where NOT to apply AI.

What are some of the challenges you’re currently facing in your work?

3 personas I've listed above illustrate the challenge since their skills have very little overlap. AI right now is a field that requires depth and breadth which nobody has especially when it comes to ethics and societal/psychological impact it will have. Everyone wants to rush their prototypes out and deploy without space (or frankly incentives) to consider the bigger picture.

Where do you see the biggest transformation happening in education in the coming years, and where will AI have an impact?

1:1 personalized learning has been the Holy Grail of education and AI will finally enable it. The question is, are we as humans ready for the radical transparency it will bring? I think the definition of education will expand from just children to every human being and be applied to the whole person including things like lifestyle, diet, spirituality and other nutrients we need to give meaning to our lives instead of just money or power.

What does the rapid pace of tech development mean for society?

The pace is only increasing because it’ll be so pervasive and nobody can escape it. So instead of addressing just a certain segment, we better address the whole curve, especially the long tail. Because the power of each individual is growing rapidly also with the tech and destruction is much easier than creation. So if we disenfranchise or leave a certain group behind then the whole interconnected world has never been easier to destroy than now and into the future.

What can be done to ensure we’re not ‘following data blindly’?

Bringing all the domain experts in with historical, scientific and social backgrounds to analyze the data to fret out biases and make sure that “source of truth” is something that will benefit for most everyone. This is the biggest failure of the Big Tech companies since we aren’t trained in those disciplines and we think those other disciplines move too slowly. We got to this outcome through Hacker way but it won’t be successful for AI. The responsibility and the task is just too big.

Wikipedia is a good example of democratizing both access and contribution to the masses. But you still see editors and curators who care enough to guard some pages. We have to engage everyone but there should be means of control so it’s not just anarchy. When you create an interconnected world like that, positive social impact naturally follows I believe. AI is merely a tool but one that can scale for any individual to make an impact. The role of education is to enable these newly empowered humans with critical thinking and conscientious to wield the power responsibly.

Where will we see AI benefit society the most?

It’ll free all of humanity for the first time in known civilization to be fully human instead of being judged based on economic utility. Basic mundane tasks (assuming we solve the energy issue) will be automated and basic services can be provided for society en masse. The difficulty will be how does each individual figure out their “worth” and choose to spend their time. It’ll bring on potentially massive existential crises for the masses who had never had to grapple with it. But I think with guidance and social safety net, everyone can find their way. Maybe it’s a curse of being a teacher but again I see 1:1 education for every human as the real benefit for their own personal growths.

What does a typical day look like for you?

A read a lot whether it’s emails, articles, labs, books or messages. We have a lot of events happening worldwide so many things need attention and dealing with all that while remaining balanced. I’m always searching for the patterns and ways to automate so always thinking about how to leverage AI to solve these problems as I go through them.

What’s next for you in your work?

Building a coalition for Human Insight Project so we can seed future AI with the balanced and inclusive dataset that represents all of humanity.