Artificial Intelligence

Three things they should have put in the manual (if there was a manual), by Jason Riesa

by Jason Riesa

I was asked to contribute an article about my time at ISI as a graduate student, but when you get your PhD you are no longer supposed to always do what you are told, so I thought I would modify the scope a bit. Young researchers are often bad at maintaining structure and good at digging and diving deep into their narrow research topic and staying there. Neither of these are helpful for crossing the boundary from young researcher to seasoned scientist. Here are three things I learned while a PhD student at ISI that helped me along.

1. Read new things. Daily.

Most days at 4pm I stop what I’m doing and pick up a paper to read for an hour. It could be a paper I am reviewing or something just on my reading list, classical or new, in my research area or in a related area – perhaps some new topic I’ve been meaning to learn more about. For example, I concentrate on machine translation, but lately I’ve been reading a nice survey paper on topic modeling. Setting some consistent reading time aside for yourself helps to keep current with what’s going on in your field and better informed about related areas.

2. Find balance. And keep a schedule.

Some weeks before I defended, I confessed to Daniel Marcu I had had a great evening in Hollywood with several friends on a recent Saturday, and that I hadn’t done so in quite a while. Daniel chuckled approvingly and replied, “Why aren’t you going out every Saturday?” He had a point: Be very productive during the week so you can take some personal time on the weekend. After all, you’re only young once. You will come back in recharged, refreshed, and re-energized to make tons of progress the next work week. There will always be deadlines and extenuating circumstances, but try to find your balance. And, keep a regular schedule. After you finish classes, graduate school becomes very unstructured. Build your own structure by finding a routine. Time management is key.

3. Do an internship. Or two.

The expectations in industry can be different than in academia. It’s good to be exposed to industry if you are at all considering this option for employment after grad school. Implementation skills, communication, and teamwork are highly valued. You may want to apply for an internship early in your PhD studies to experience a different culture and potentially be motivated by or get research ideas from contributing to work in new problems you wouldn’t otherwise be exposed to at ISI. Doing an internship later in your career also has its benefits – you will have a deeper understanding of your field later in your studies making you more valuable and more productive in a research team in an industrial lab. Your host will also be more likely to remember you when you are looking for jobs.


Jason defended his PhD dissertation, entitled Syntactic Alignment Models for Large-Scale Statistical Machine Translation, in March 2012. He will be joining the Google Translate team in the fall.