By A. Sharmen Mir ’15, THURJ Staff

Gajos_3In a sunny office in Maxwell Dworkin, a whimsical and rather fitting cartoon of the words “Things to Build” is framed opposite to a desk and shelves lined with volumes of books and gadgets. Indeed, current Director of the Intelligent Interactive Systems Group at the Harvard School of Engineering and Applied Sciences and recently named 2013 Alfred P. Sloan Research Fellow, Professor Krzysztof Gajos has constructed a myriad of projects spanning the fields of artificial intelligence, machine learning, and human-computer interactions.

The Early Chapters

Dr. Gajos began his career in computer science as an undergraduate at MIT, where during his third year, he joined an artificial intelligence (AI) lab, which became a formative experience for him. He fondly recalls, “My mentor at the AI lab, a graduate student at the time, had a curiosity that was contagious, and he created a really good lab environment.” The experience propelled him to stay on as a Master’s student and then again for another two years as a technical manager. After finishing his Ph.D. at the University of Washington, Gajos joined Microsoft Research. He was still attracted to the nature of academia and was ultimately drawn in because of the mentoring opportunities through which he could work one-on-one with colleagues and students. Reflecting on his university years, Gajos enthuses, “Being a graduate student is sort of the best kind of job. After all, you are getting paid to make discoveries—how cool is that?”

Gajos spent time teaching introductory AI at Ashesi University in Ghana before joining Harvard in 2009, where he now teaches the undergraduate-level course “Computer Science 179: Design of Usable Interactive Systems” and its graduate-level counterpart “Computer Science 279: Research Topics in Human-Computer Interaction.”

His current and past projects span many different applications and subfields. In his approach to interdisciplinary projects, Dr. Gajos and his team first begin with a particular question already in mind, and this usually leads them into a certain discipline that they then further investigate. One particular example was a landmine detection app, Pattern Enhancement Tool for Assisting Landmine Sensing (PETALS), which one of his students had been working on and which was motivated by real issues affecting countries like Sri Lanka and Cambodia. For psychology-related problems, Gajos and his team often take the reverse approach, where, rather than start with a preformed question, the CS research is motivated by the psychology findings themselves. One such project is Lab in the Wild, which examines how the environment can impact visual perception and how this understanding can be applied to real-world design choices. For Gajos, a particular goal of interest is to create broader and more equal access to computing, taking into account cultural differences since, as he notes, almost all current knowledge about the design of user interfaces involves mostly Western participants.

The Future is Now

Discussing the different fields in CS that he specializes in, Gajos explains, “Human-computer interactions started from human factors in psychology, and from there it came into computer science, whereas artificial intelligence started from cognitive science. For a long time, the two fields had a hard time communicating with each other, as one took a more rigorous mathematical approach.” What links them together now is their objective. Gajos remarks, “We want machines that are better and are more pleasant to interact with. When we interact with a computer, we want to have a very good idea of what to do; we want something where we say, ‘Find me the best flight to Paris,’ and have it know all of our likes and dislikes and is intelligent without being surprising. We want them to be independent. Designing something that satisfies all of this turns out to be incredibly hard.”

In Gajos’s view, “The turning points [in the fields] have already happened—it is just that we have not seen all of the consequences yet.” He cites the Roomba robot in many homes as “an example of a very complex AI system that appears simple, controllable, predictable, and robust.” As another example, he describes a nested thermostat that has been developed with a very simple design and learns from the user’s schedule and preferences. Gajos contrasts these current, innovative devices with earlier examples, like the old Microsoft Clippy, which Gajos describes as “something that tried to be intelligent” but in reality had interacted with users in a way that was not actually well developed.

Curiosity Reigns: Insights for the Next Generation 

Dr. Gajos observes, “When I work with undergrad students, they say yes to an ‘impossible’ project. Research is a side thing for undergrads, but undergrads happen to do risky projects.” The aforementioned landmine clearance project, PETALS, which has now evolved into an entire startup, is just a single case in point.

However, Gajos also notes that some students who are risk-seeking as undergrads “suddenly become risk-averse because now their life depends on research. Many young graduates forget they are graduates to begin with because they end up turning a passion into a tedious career. My advice is this: remember you are here because of curiosity, and you will do better research.”

While it is hard to generalize the particular challenges, Gajos advises, “A problem that is obviously valuable and is one that everyone thinks has a particular solution that will lead to success will have been done already. For some other projects, you are not sure whether it will make a difference and whether the answers will be fruitful. Students who choose to do research sometimes consider practical skills of secondary importance, but they are incredibly important enablers. Have strong and diverse skills so that you can make progress quickly, and so that engineering and technical challenges are not a barrier to making conceptual progress.” He adds that it also then becomes “easier […] to throw out things that are obviously bad ideas.”

Curiosity is unquestionably crucial in research. Gajos remarks, “Between CS 179 and 279, in CS 279 we almost entirely forget about grades, and the focus is to understand the research process with the motivation to develop something interesting. The nutrition app PlateMate, in fact, started in 279. It was such a good project. The students working on it were excited about it and continued for another semester and turned it into a successful research paper. Even in CS 179, we have been successful in creating a strong atmosphere.”

While the research is heavily applied in nature, Gajos stresses the importance of the pure theory of computer science as well. “As an undergraduate,” he fondly recalls, “I was looking at the most fundamental building blocks and loved them all. CS is an incredibly powerful intellectual toolkit, and what’s under the hood is fascinating and something I would really encourage all students, including CS minors, to expose themselves to. It really opens up your eyes, and the elegance is incredible.”




This site uses Akismet to reduce spam. Learn how your comment data is processed.