Cognitive Science Alumni

Our Department isn't defined by the physical space we occupy on the campus of UCSD - it is defined, rather, by the remarkable individuals who make up our community. The lifeblood of any community is its people, and this is especially true of a community that relies on ideas and innovations. In recognition of this, the Department of Cognitive Science at UCSD presents our new "Alumnus/Alumna of the Month" feature.

The goal of the Alumnus of the Month program is to celebrate some of our outstanding Alumni, while giving all students - past, present, and future - an opportunity to meet some of our graduates, and to see some of the amazing things that people from our community have accomplished - in the field of Cognitive Science and beyond.

The hope is that it will both put a more 'human' face on Cognitive Science, as well as be a testimony to the wide range of interesting things that one can do with a Cognitive Science background. In addition, of course, it's a means for CogSci alumni to see what other alumni are up to.

Alumna of the Month: Dan Liu, 2008

Dan Liu

Dan Liu was born in Beijing and got her Bachelors of Engineering and Masters of Science at the Department of Automation at Tsinghua University. She was also a visiting student at the multimedia group at Microsoft Research Asia in Beijing while getting her Master's degree. She came to America in 2003 to join the Cognitive Science department at USCD, working with Professor Emo Todorov on the computational modeling of human motor control. Dan also did two internships at Google in the summers of 2006 and 2007. She joined Google full-time in 2008, after completing her Ph.D.

An Interview with Dan Liu

by Tristan Davenport

Could you talk a bit about what your research was at UCSD?

I was working on computational modeling of human motor control with Emo Todorov. On one hand, I designed and wrote programs to run the experiments. On the other hand, I did statistical analysis on the data and applied optimal control tools to model the data.

A quick question: what are optimal control tools?

They're a computational tool. Basically, you define your goal - in this case, moving an arm to a certain place - into an objective function, and you try to find the best action to achieve the best performance.

What was your prior background in computational modeling?

Before I joined UCSD, I was a visiting student at Microsoft Research Asia. That's where I started doing machine learning stuff, and I learned to use visual C++. That helped later, because during my Ph.D., all the experiments were written in C. I also read many books on coding towards the end of my Ph.D., when I was preparing for my interview at Google!

How did you decide on the Cognitive Science Program at UCSD?

At the time when I chose grad schools, I had several options including cogsci and computer music at UCSD, computer science at UBC, and some others with no full financial support. I was actually considering computer music since my M.S. was on applying machine learning to music recognition, but I was worried that I would have to take too many music classes. Meanwhile, I got the chance to talk with Emo and Jochen Triesch in CogSci. I learned a lot about the program and knew that I could do computational modeling of human behavior, which is what I am interested in. Most importantly, I exchanged many emails with Emo about his research, which I really liked. In addition, he's made a big impact on the field, and is very nice and considerate. So I chose CogSci in the end. Emo turned out to be a great advisor. He is the smartest person I have ever worked with. He always knows what the key issue is, and how to connect everything together. He knows the brain very well, and has an extraordinarily strong math background. And he's so good at asking the right questions. After many years in academia, I have seen many sad stories between grad students and their advisors. I feel so lucky to have worked with Emo Todorov.

Why don't you tell us about your current job?

I am an analyst in the ads department at Google right now. Basically, we need to better select ads and rank/show them in a proper way to both fit users' needs and of course to make money. So what I'm doing now still has a lot do to with running experiments, and applying statistical analysis to the results - although in a totally different field.

And did you know all along that you were looking for a job in industry after your PhD? Or did you also consider academia?

I was hoping to become a professor when I started my Ph.D. But by my third year, I had been to a few conferences. I realized that it's hard to compete with so many geniuses. Plus, although my research went well, most of the time I was following Emo's ideas. I don't think I have the vision he has, and I may never have it.

Do you feel that the things you learned in the UCSD Cognitive Science department are valuable to you in your job at Google?

I'm still using matlab to draw beautiful 3-D pictures! Mostly, though, I am using C++, Python, and R now, which I didn't use much at UCSD. I have to admit my coding skills are not as good as the computer science people's... I don't think the classes I took at UCSD are directly useful here, but the learning skills acquired during the Ph.D. process have definitely been important so far. Google has a lot of very complicated systems, and learning them is a big challenge for every new employee. During my research, I was constantly learning new stuff and becoming more and more confident and efficient facing new tasks. Also, my research experience with Emo taught me how to ask the right questions during meetings, and then figure out the details on my own later. This has turned out to be very useful when I'm working with other people in a large group. One class I took that actually was very useful was the second- and third-year project class, taught by John Batali and Rik Belew. I learned so much about presentations then, and it's been so useful during my presentations at Google.

How is the food at Google, by the way?

Awesome. We have lobster now.

Wow. But that must keep you at the company for a long time every day, right? Do you find it hard to keep a balance between work and the rest of your life?

Not really. My boyfriend is still living in San Diego, so I don't cook at all. Having three meals a day at Google is very helpful to me. I usually go to work at 9:30am to catch breakfast. At around 6, I go to my yoga class (in the gym at Google) or play tennis (I am playing on a Google ladder). I come back for dinner and then go back home. So I'm not working all the time.

What do you think people should consider when choosing a graduate program?

I don't have any wise opinions on choosing the right graduate program. I guess it's important to consider your research interests and your career goals. If you want to go to industry, you should consider labs with industry connections and try to work on projects that are more applicable. If you want to stay in academia, getting a well-known professor as your advisor would be more important. Finally, you may want to talk with grad students in the program/lab you are interested in. I hope that's helpful.

What advice would you give to current first-year students in the UCSD Cognitive Science program?

Understand quickly whether you really want to spend five or more years on getting a Ph.D. Avoid quitting after three or four years. Also, if you do want to go into industry, getting an internship (like the summer internships at Google) is very, very useful.

To Nominate Someone

To nominate someone as an alumna/alumnus of the month, or if you would be interested in being featured yourself, please contact us at alumnusofthemonth@cogsci.ucsd.edu.