Context: On Feb 25, the College of Engineering at the University of Michigan organized a ceremony for my 'installation' as the Arthur B. Modine Professor of Engineering. It was a wonderful and warm event, and for me, an opportunity to celebrate the people and culture behind this collaborative achievement. I gave a talk titled 'Candles in the Dark: A Journey towards Computational Intelligence'. In a couple of posts, I'll recount some of the key messages from the talk. In this one, I'll explore my inspirations and some philosophy (perhaps some of it in a Steve Jobsian "connecting the dots backwards" sense)
Every researcher has an origin story. Mine was a profound moment as a 2nd year undergraduate student. I remember picking up a book on finite element analysis by Zienkiewicz & Taylor and coming across an image: it showed a computation of an object impacting a hard surface right next to a picture of the actual experiment. Seeing that you can look at a complex physical problem, write down some equations, solve them on a computer and get so close to reality simply blew my mind; I was instantly hooked and decided that this was what I wanted to do for the rest of my life.
Couple of years later, at the Indian Institute of Science, I distinctly remember the lecture in which Prof. Balakrishnan introduced flux vector splitting and flux differencing, and my interests almost immediately shifted from FEM to CFD. That initial spark was further solidified when I discovered the incredible work being done right here at the University of Michigan. Around 1999, I found research from the Michigan Aerospace CFD group, led by Bram van Leer, Phil Roe and Ken Powell (see postscript below). These were giants in the field. The website is preserved by the internet archive and I can't believe that I remember a couple of theses from 25+ years ago (even some figures in them!).
What truly inspired me about the research coming out of that lab is how to approach hard problems and use what appeared to me as "mathematical cleverness" and physical intuition to transform them into something elegant, computable, and useful. For me as a student back in India, this work was a revelation: they instilled in me that computation wasn't just a way to solve problems or a career path, but a beautiful intellectual pursuit.
Fun stuff: One of the things that fascinated me about the CFD group at Michigan was the wine snob's guide . That was circa 1999. In 2018, for Phil Roe's 80th birthday, I had him read it out loud. Enjoy. Now the time you invested in visiting this post has paid for itself. You're welcome.
That sense of fascination and the lessons I gleaned there, focusing on what I call as the "computational balance" continue to shape how I approach research today. I learnt a few things at Maryland during my PhD with Jim Baeder, and by the time I got to Michigan, I had the good fortune of exploring a wide range of topics. My time at Michigan has deepened and broadened this exploration beyond my dreams across the entire spectrum of computational science.
Fun stuff: One of the things that fascinated me about the CFD group at Michigan was the wine snob's guide . That was circa 1999. In 2018, for Phil Roe's 80th birthday, I had him read it out loud. Enjoy. Now the time you invested in visiting this post has paid for itself. You're welcome.
That sense of fascination and the lessons I gleaned there, focusing on what I call as the "computational balance" continue to shape how I approach research today. I learnt a few things at Maryland during my PhD with Jim Baeder, and by the time I got to Michigan, I had the good fortune of exploring a wide range of topics. My time at Michigan has deepened and broadened this exploration beyond my dreams across the entire spectrum of computational science.
I also spoke about the "candles in the dark" philosophy, a metaphor for an evolutionary and incremental approach to both research and education, which I adapted from Phil Roe's final MICDE talk.
The central premise is that approaching a new problem or education is like entering a dark room.
- Instead of seeking a way to illuminate the entire space instantly, a researcher’s task is to find a single candle and light it.
- Lighting that first candle does not reveal everything; it only provides enough light to see a small, specific portion of the room.
- The most valuable "candles" (or research findings/teaching moments) are not necessarily those that provide the most light, but those that provide enough visibility to help you find the next candle to light.
The idea here is to focus on the process of inquiry rather than just the answer. Absolutely relevant in the age of AI (learning requires 'friction' and some level of struggle). This encourages a mindset of computational balance and mathematical cleverness to turn hard, complex problems into something elegant, computable, tractable and accurate.
In a practical academic setting, this philosophy dictates how I have mentored doctoral students. Rather than presenting a student with a cleanly defined long term goal, the process begins by lighting a few initial candles to see what they reveal. As the research progresses and the room becomes more visible, the project's direction is shaped by the student’s specific aptitudes and the new information discovered along the way. This makes the PhD journey a collaborative discovery rather than a task.
At CASLAB, we also apply this kind of approach at a different scale.. to explore all of computational science. We apply this philosophy to three specific "groups of candles" or fundamental questions in our march towards computational intelligence:
- Distilling Reality: How can we translate the vast complexity of the physical world into a computable form that a machine can process?
- Predictive Consistency: How do we ensure that models remain consistent with the underlying representation and physics constraints while still being powerful enough to make accurate predictions?
- Updating Internal States: How can we effectively and efficiently update the internal models of a system ?
I view computational science research as a lifetime commitment to curiosity, where each discovery/ rediscovery (candle) serves as a bridge to the next level of understanding. As a microcosm of this, I expose students to various candlelights in the 'Science of Predictive Modeling' class which I am teaching this semester.
-----
P.S. It was cool to see Ken Powell and Bram van Leer in person, and I believe Phil Roe was online. Here's something fun: Bram had retired just around the time I interviewed at Michigan. During my interview visit, I found this 'memorial' plaque (ah! our quirky department) in the restroom and photographed it (pic below from 2013) and remember thinking "Now that van Leer has retired, I can be the poor man's van Leer". Couple of coincidences :
- van Leer was the holder of the Modine Professorship before he retired
- The ceremony was exactly 13 years after (to the day) that I arrived in Ann Arbor for the interview.
