Grad Student Conversations: Q & A with Randy Olson, Data ‘Genius’
Randy Olson is a doctoral candidate and a dual major in Computer Science and Ecology, and Evolutionary Biology and Behavior at MSU’s College of Natural Science. He has worked as a graduate assistant in the Adami Lab under the guidance of Dr. Chris Adami. Randy is interested in evolution and how it interrelates to machine learning and artificial intelligence. Randy’s latest project with Discovery News, “How to Really Drive Across the U.S. Hitting Major Landmarks”, recently garnered significant media attention.
I chuckled when I saw that. The press has called me several things over the past year, and as I’ve pitched these data-oriented stories, they wanted to give me some sort of fitting title and seemed to group me into this “data genius” category. As these stories hit the media, I’ve had several top-notch professors ask me how I’m getting such notoriety for what amounts to solving a relatively simple problems. The reason these stories seem to be getting attention is because they are something people can relate to and actually use in their everyday life. Up until January, I was doing very complicated research with data analysis and machine learning. But I wanted to do something that the everyday person could recognize as useful and interesting. Quite often fundamental research helps us to understand a concept that might apply or help in another field, but if there isn’t a practical outcome, people don’t seem to understand it. When people hear about research, sometimes they need to understand what’s in it for them.
Your projects are curiously random (football predictions, Where’s Waldo and a U.S. road trip.) Where do you get your ideas and inspiration?
Early on, I would have random conversations with friends, family, and lab mates about the latest news and I’d wonder if there was any research on the topic so I could find the answer. So I’d go home at night, do some research on the Internet, and if I couldn’t find anything I’d try to find the answer myself. Recently, my projects have been partially guided by my collaborations with the media since they have a great feel for relevant trends. The road trip project was done in collaboration with Discovery News. They contacted me after the Waldo project was in the news.
Your blog indicates you’ve already followed up the U.S. road trip with a trip across Europe. What’s next?
A few days after I published the U.S. road trip, I started working on the European version to demonstrate that this process doesn’t apply only to road trips in the United States. If the Google Maps data exists, you can optimize routes anywhere in the world. And this doesn’t only apply to road trips, either. It could be used to optimize bike routes, walking paths, city paths for touring a city, and even combining public transit with walking. There are numerous applications for this technology. We’re essentially building a layer over Google Maps, so if the data is available, it can be optimized. Google can do this now, but only for a maximum of 10 way points. This model can handle hundreds of way points, so the possibilities are endless.
How has working in the Adami Lab helped fuel your creativity or changed your thinking?
It’s really been a lot of fun. I’ve been there for four years already and Chris has been great as an adviser and it’s partly because of him that I’ll complete my PhD in four years. He is fairly hands off and allows our group to explore things on our own, yet he mentors and gives advice when we need it. I like to work independently and explore my own creativity, so it’s been a very positive experience. The core of my research is using evolution to shape intelligent behavior and artificial intelligence (AI). Being in this group and working with Chris — who works at the forefront of using evolution to create artificial intelligence — has been invaluable.
Before I started in his lab, I thought that programming a robot to walk was artificial intelligence. But really, that is just programming a repetitive task – telling the robot to put one foot in front of the other. Working with Chris guided me to better understand the real principles of artificial intelligence. I am looking to create a system that can learn and is not simply hard coded behavior – it learns like a baby through experience. The movie Chappie is a good current example. My interest is to develop AI that can learn and adapt to the world in real time.
Where are you currently in your educational career and what are your overall career aspirations?
Well, I’ll be graduating this spring and finishing my dissertation on how we can use evolution to teach individual agents to work together as group. We don’t just want one of these robots or machines to work alone; we need them to work together as a team to solve problems. I’ve accepted a post-doctoral position at the University of Pennsylvania where I’ll be continuing my research in AI and working in bioinformatics to solve human health issues. Eventually I hope to find a tenure track faculty position. I haven’t found many companies working on adaptive AI since most companies are working on short-term applications that can be used in the very near future. I plan to stay in academia so that I can continue working on the cutting edge of AI research.