Interview with Blake Elias, AI Resident
Blake Elias, AI Resident
Tell us a bit about your background.
I recently completed the Bachelor of Science and Master of Engineering degrees in electrical engineering and computer science at the Massachusetts Institute of Technology (MIT). I always wanted to understand and augment human intelligence. As a result, I took symbolic AI courses, like Professor Patrick Winston’s “The Human Intelligence Enterprise” and Professor Gerald Sussman’s “Adventures in Advanced Symbolic Programming,” and read books like Marvin Minsky’s Society of Mind, which were absolutely inspirational.
Because I also wanted to take advantage of MIT’s other unique opportunities outside computer science, I took classes like “Biological Circuit Engineering Laboratory” with Ron Weiss and Tim Lu, and “Principles of Neuroengineering” with Ed Boyden. These disciplines promise to help us understand and enhance the human experience as well. For my master’s thesis, I developed a platform for low-cost, high-throughput automated DNA assembly using robotic pin tools.
Before this, I was the first employee of IdeaFlow (opens in new tab) [CrunchBase (opens in new tab)], which is building a human-AI hybrid “shared brain” for organizations. I’ve also done research in the MIT Media Lab on augmented reality for enhanced learning and memory, and was a technical program management intern at Google.
How did you find out about the AI Residency?
I had heard about similar programs through a jobs mailing list at MIT, so I started investigating those. At one point I saw a forum post discussing AI residency programs and someone mentioned the [then] newly launched Microsoft Research AI residency.
What made you decide to apply?
I thought an AI residency program would be a good way to improve my skills and work on some awesome projects. I liked that there would be opportunities for my research to go into products and/or free software projects, with an eye towards human-AI collaboration and having an ethical impact on people and society.
I was specifically interested in Microsoft Research because of its Neural Program Synthesis and Deep Program Understanding projects. While the AI world has been moving towards machine learning, I see a lot of value in the older symbolic AI that I learned at MIT. I’ve been wanting to combine these two approaches and these projects seemed like a great domain for that.
Six weeks in, what are your initial thoughts on Microsoft and your experience?
Microsoft is a great place to do research. We have the full R&D pipeline, from basic theoretical research to applied end-user applications, and everything in between. It’s interesting to see this all under one roof.
My current project combines SAT solvers with neural networks, to teach computers how to write simple programs. This is a core problem that asks “how can we combine statistical and symbolic knowledge – pattern matching with reasoning – when neither is perfect on its own?” I think progress here can illuminate new approaches in many areas of AI.
Microsoft Research is doing a wide range of research in many areas of computer science, and there is an opportunity for AI researchers to collaborate with other areas (in my case, for example, program analysis and verification).
What do you want to get out of the residency?
I want to improve my skills designing deep neural networks and develop insight on how to combine machine learning with symbolic AI. I also hope to learn guiding principles for developing technology with moral impact and that truly makes our lives better – not just faster.