Turing Laureate, Compiler Pioneer Jeffrey Ullman Visits NJIT for Data Science Lecture

When a Turing Award winner speaks, NJIT faculty and students listen.

Jeffrey Ullman, known as a father of the modern compiler at Bell Labs and Princeton University in the 1960-1970s, filled the largest lecture hall in NJIT's Guttenberg Information Technologies Center for his October 14 lecture on data science, as part of the Ying Wu College of Computing’s distinguished speaker series.

Ullman won his Turing Award, considered the equivalent of the Nobel Prize in computing, in 2020. He is now professor emeritus at Stanford University. He talked about his new passion, data science, which permeates New Jersey Institute of Technology and practically every major category of academia and industry too.

He noted that the original term was data mining, then big data and now data science, but it refers to the same principles. "As far as I'm concerned, the idea has been the same all along, which is you put big hardware, algorithms, the smartest programming systems and you're trying to solve problems," Ullman said.

"Statisticians think data science is machine learning and machine learning is statistics. I see data science as coming out of database systems. … There's computer science and there's domain science, and somewhere in the middle is data science."

Ullman illustrated his point by referring to a popular Venn diagram, attributed to consultant Drew Conway, defining data science as the intersection between mathematics/statistics knowledge, hacking skills and substantive expertise. Machine learning exists between the first two sections, traditional research exists between the second two, while the overlap between the first and third is just dangerous, he joked.

No algorithm is perfect and even the best ones today might not be so good tomorrow.

"Right now machine learning has big wins and in many cases has beat conventional algorithms … We need statisticians in computing [but] I think the best statisticians are thinking about algorithms in the same way as computer scientists. They won't admit it," he said.

Ullman added that aspiring data scientists must simply try to do their best. No algorithm is perfect and even the best ones today might not be so good tomorrow. An example is anti-spam software, which changes efficacy as spammers change their strategies to reach your inbox. "Just because you can't do a perfect job, doesn't mean you shouldn't be trying," he said. "Do the best you can."

Craig Gotsman, dean of Ying Wu College of Computing, was among the audience members keen to hear Ullman speak. He noted that several of Ullman's textbooks on compilers, algorithms and databases were vital to his own undergraduate computer science education 40 years ago — "Basically everything I learned in basic computer science came from Ullman," he said, “before you could find information on the World Wide Web."

After the lecture, Ullman met with Ph.D. students. He emphasized the importance of research collaboration with anyone possible, including other faculty and students beyond the NJIT campus.