I’m happy to consider the world an algorithm: Microsoft’s Madhu Sudan
AI expert Madhu Sudan on chip design, accessible mathematics and the intellectual challenges of Solitaire
Madhu Sudan, principal researcher at the Microsoft Research lab in New England—one of Microsoft Corp.’s 12 global research facilities—and adjunct professor at the Computer Science and Artificial Intelligence Laboratory of the Massachusetts Institute of Technology (MIT), hopes his current research on reliable human communication will eventually help computers talk to each other in different environments and with fewer errors.
In an interview in Mumbai, Sudan said he would rather encourage young research scientists to chase their own dreams than incrementally build on other people’s ideas. He also touched upon how mathematics is increasingly being used to verify algorithms, ensuring that programs do what they were designed to. Comfortable with the idea that nature itself is an algorithm, Sudan believes playing Solitaire is as intellectually challenging as chess. Edited excerpts:
What got you interested in mathematics and theoretical computer science?
I was doing well in mathematics, which always helps if you like a subject. Besides, mathematics is also a language of precision with very little ambiguity. As for computer science, I drifted into it. When you are in IIT, some people have a vague understanding of what engineering is—something you need to build buildings. And that was the extent of my understanding too. I knew mathematics and engineering were coveted, but I actually had no idea of what I was getting into.
What did you work on as a research staff member at the IBM Thomas J. Watson Research Center from 1992 to 1997?
I was looking at the theory of optimization. There are different varieties of this concept. Take chip design, for instance. Some chip designs require a smaller amount of space, while others require large surface areas. The challenge is to optimize the design of the chips so that you make them smaller in size. Some optimization problems are so hard that even today, computers will take years to solve them. I have never worked directly on any project. I have always looked at the theory behind these problems, written papers on how well such issues can be solved, and maybe sometimes developed algorithms to solve these problems. We highlight what methods will work, or not work, and this approach ends up saving companies a lot of time and effort since they may not want to waste time trying things that won’t work.
But isn’t it true that every researcher wants to become famous and be known for some discovery?
Not all have this ambition. My ambition is to do relevant work—mathematics that a large number of people can understand, that can help develop new technologies. For instance, when I was working at IBM on optimization, I also drifted into error-correcting codes. This was not what IBM hired me to do. But the underlying mathematics for topics like optimization or error-correcting codes for that matter is similar. This also explains my current drift towards researching reliable human communications.
So this is what you’re working on at Microsoft Research?
Microsoft has a significant collection of researchers. I work in the Boston area. Our goal is to push the frontiers of science. The secondary goal is to be useful to Microsoft. But in my case, my research is so futuristic that I do not see Microsoft benefiting immediately from it. I’m working among a collection of inter-disciplinary researchers—for example, with social media researchers who help me in understanding how humans communicate, which helps me in my research on reliable communications.
You simultaneously teach at MIT...
The company also wants me to be active in my field, and here is where my academic connection with MIT helps. So I happen to be in a very fortunate setting.
How futuristic is your current research?
During the time of Shannon (born on 30 April 1916, Claude Elwood Shannon was an American mathematician, electronics engineer and cryptographer, and is also known as the father of information theory), information was not considered to be mathematical, which is what he made it. Uncertainty itself can be modelled mathematically, and we can measure it.
My work is focused on reliability—how to properly communicate what I’m thinking about. Even when we are sitting across the table, I could misunderstand the context since no two human beings have common backgrounds. This leads to a fascinating variety of challenges to overcome. We have had 70 years of excellent research that covers reliable communication when we talk about the first order of information—for example, how to send an email. But when we talk about software being compatible across devices, we still have problems. Computers behave very differently from the way humans communicate. For instance, if you want to go to a new cafe, you do not call up the owner to ask how to order a coffee. You simply speak. Computers almost never have a Plan B. Most of the time, they do not even know if Plan A worked. Humans can do these tasks better. So there are learnings that can be implemented in computers.
Have you exchanged notes with Naom Chomsky, given that he’s done extensive work in some of these areas?
Yes. We did exchange emails with him. We wanted to explore why we have ambiguity when communicating—one word having different meanings—and we were trying to come up with a mathematical model to explain this. That’s when we wrote to Chomsky. We were privileged to get a detailed response. In fact, he said the word ambiguity itself is ambiguous (laughs).
You won the Infosys Prize 2014 in Mathematical Sciences for your contribution to theoretical computer science, especially in the area of error-correcting codes. What do you plan to do with the prize money?
I haven’t been actively thinking about it. It’s good to have a nest egg, though. More than the award money, it’s the prestige and visibility that the Infosys prize has got me.
Do you keep emerging markets in mind while doing research, especially given your association with India?
I do not tailor my research to suit any geography. But I stay connected with researchers at academic institutions across India. Microsoft also has many young researchers at its lab in Bengaluru, whom I mentor periodically.
What advice do you give research students?
I advise them not to try to play catch-up, because you stand to lose. Even in research, many are good in playing catch-up with increments on top of a body of research that has been published. These people are very essential to the ecosystem. But you’re also better off chasing your private dreams. It is feasible to find this out with a trial and error process. You can write a paper and submit it at a conference to find out whether your idea has been worked upon.
What do you have to say about the hypothesis that our world may, in fact, be an algorithm?
An algorithm is basically a sequence of steps. Each of which is simple, but may produce complicated outcomes at the end of the process. Now sometimes the rules of how you make the simple steps are designed by you. Those are the algorithms that we design and run on computers. Nature is also another algorithm—see the way it mutates the genes and DNA a little bit to see if it works, then mixes it together with a population and shakes it a bit. I’m happy to consider the world as an algorithm.
Do you, then, believe in God?
I’m agnostic to that concept.
I presume you play chess, given your interest in maths...
I do not like competitive games—two-person games—where one person has to lose. I like one-person games like Solitaire. And let me say there is research to prove that Solitaire is no less intellectually challenging to play than it is to play chess.
Unlock a world of Benefits! From insightful newsletters to real-time stock tracking, breaking news and a personalized newsfeed – it's all here, just a click away! Login Now!