In recent years, the field of artificial intelligence has exploded, enabling leaps in computer image recognition, speech translation, robotic movement and a number of fields with wide-ranging implications. Based on improved techniques originally developed in the 1950s, researchers have been able to train computers to recognize patterns using large data sets and complex statistical techniques. Companies such as Google, Microsoft and Facebook have led the push, hiring hundreds of researchers in the field, implementing AI in their products and creating research groups focused on tackling major issues in the field.
The explosive growth in AI led by the private sector has heavily impacted the computer science department over the last five years, leading to ballooning AI course enrollment, increased faculty numbers and an evolving sabbatical system.
Ballooning Course Sizes
CS recently surpassed Economics as the most popular concentration at the University. As more students have flocked to the concentration, rising course enrollment has placed an increasingly heavy burden on professors.
Previously, CSCI 1430: “Computer Vision” had 28 students enrolled in Fall 2009. Now it has over 100 students. Similarly, enrollment for CSCI 1460: “Computational Linguistics” jumped from 19 to 104 students between 2010 and 2016.
Professor of Computer Science Michael Littman has witnessed these changes firsthand. Littman has taught and conducted research in AI since the 1990s, when the field was still relatively niche. This semester, Littman is teaching CSCI 2951F: “Learning and Sequential Decision Making.” A graduate-level class initially capped at 20, the course now contains 44 students.
Assistant Professor of Computer Science George Konidaris has felt the tidal shift as well. Konidaris teaches CSCI 1410: “Artificial Intelligence,” an introductory undergraduate AI class. The class had 48 students in spring 2009, but its numbers have more than doubled to 99 this fall. During his time teaching the class and its equivalent course at Duke University, he has seen students’ interests change. Previously, students would get interested in AI through their coursework, which is heavily focused on the topic’s theoretical aspects. “Now, students every day are reading about it in the New York Times. There’s a new article every week,” Konidaris said. As the subject’s audience has widened and its importance grown, Konidaris has adapted his teaching to broaden the scope of the class beyond theory to give students a “flavor” of what AI has to offer.
The rapidly increasing pace of AI and the computer science field in general has placed staffing burdens on the department as well. Five or 10 years ago, topics like AI and data science were barely on the radar, said Ugur Cetintemel, chair of the computer science department. Today, the department is racing to hire faculty to stay on track with peer institutions. This year, the department added Daniel Ritchie, assistant professor of computer science, and Ellie Pavlick, who will join the faculty in spring 2018 as an assistant professor.
The department has also increasingly relied on lecturers, who solely teach and are hired on more tentative contracts. The department also seeks to add a new lecturer for some of the introductory courses next fall, Cetintemel said.
But these additions are not enough to keep up with course sizes and other universities. Peer institutions’ CS departments are bigger and growing faster, Cetintemel said. For example, while Brown added only one new tenure-track professor for CS for this semester, Princeton added eight, he added.
The department has been forced to reckon with the lure of industry as well. Offering salaries that are sometimes double or triple what Brown pays, according to several University sources, large tech companies like Google and Microsoft are increasingly attracting PhD students studying AI who might otherwise enter academia. These firms have many other material advantages over academia as well, including large data sets, huge processing power and a less restricted budget to bolster their AI endeavors. With a large talent pool and research teams focused on improving products, tech companies have contributed large amounts of high quality research.
Tempted by these attractions, professors have increasingly turned to the industry for sabbaticals and, to a lesser extent, full-time jobs. The attraction of private industry is not new to computer science; during the tech boom of the late 1990s and early 2000s, computer science departments across the country were often left gutted by large companies looking for top talent, Littman said. Nowadays, while some similar poachings have taken place at universities like Carnegie Mellon University and the Massachusetts Institute of Technology, some of the field’s brightest minds are now shared by academia and industry. Last year, Littman went on sabbatical to tour companies to get a better sense of the work they were doing. Stefanie Tellex, assistant professor of computer science, has considered taking a similar sabbatical, Cetintemel said, adding that she will potentially explore a robotics research project in the private industry in Boston.
While the department has sabbatical procedures in place, students and colleagues are often left to scramble when a faculty member leaves. Graduate students, who are typically dependent on a single professor as their advisor and mentor, can be left without strong guidance. Additionally, professors are often forced to switch their teaching responsibilities at the last minute.
Cetintemel said the department needs to reevaluate its sabbatical procedures in light of the changes brought on by AI. If faculty members are increasingly taking sabbaticals, the department will need to establish a new model that covers these professors’ teaching loads and graduate students. Graduate students could potentially work under multiple professors instead of one, leaving them with at least one advisor should another decide to go on sabbatical. Additionally, the faculty size can be expanded to adjust for those on sabbatical.
Though the recent boom in AI driven by private industry has certainly affected the department, many faculty members viewed the changes as complementary to the University. Research at tech companies today is fundamentally different from university research, Konidaris said. Tech companies focus their research on short-term goals and only make incremental improvements. At universities, researchers tackle issues that have 100-year time frames instead of five- or 10-year ones. For example, researchers at Brown are able to pull from research conducted across the University, said Pavlick, who does work in natural language processing. By doing so, she can complete projects that would be impossible at a private company, she said.
The field of artificial intelligence continues to explode, both at the University and across the world. Locally, the Humanity-Centered Robotics Initiative, co-directed by Littman, recently partnered with Hasbro to bring robotic AI-powered pets to the elderly. At Google, researchers have created machine learning algorithms that train themselves. While no one knows how long this current boom will last, Brown is forced to adapt to these changes on the fly, fighting to remain competitive.