Pentagon\'s Top Technologist Joins I.B.M.

Last Friday afternoon, a half-hour after he said his goodbyes at the Pentagon, where he was assistant secretary of defense for research and engineering, Zachary Lemnios was discussing why his new job made sense as the next step in his career. “Exactly the right thing to do at the right time,” he said.

Mr. Lemnios joined I.B.M. Research on Monday, as vice president for research strategy. Mr. Lemnios, 58, is leaving his post in the Obama administration, after nearly four years (political appointees typically last two to four years). Before that, he was chief technology officer at M.I.T.‘s Lincoln Laboratory, a federally financed research center for advanced technology with national security applications, and previously a senior official at the Pentagon's futuristic research arm, the Defense Advanced Research Projects Agency.

His own career charts a path from being a chip guy who would later champion and fund artificial intelligence research. Mr. Lemnios holds four patents on semiconductors that use gallium arsenide, an alternative to silicon.

Of course, he noted, the years of progress in microprocessor design and sensors are essential to the recent advances in artificial intelligence. “The hardware substrate is certainly part of it ,” Mr. Lemnios said. “But it is the software for reasoning and learning that really pushes this forward.”

In late March, for example, Mr. Lemnios announced $60 million in new Pentagon-supported Big Data research projects, as part of a $200 million administration initiative in the fast-growing field of trying to use smart technology to make sense of the explosion in data from the Web, sensors and streaming into traditional databases.

Today's technological limitations, he said, are no longer the data collection tools but the technology for making sense of it. “The real challenge is to handle, understand and use big sources of data,” Mr. Lemnios said.

At Darpa, he promoted initiatives for “cognitive systems,” an approach to artificial intelligence that embraces a learning model of computing rather than more purely statistical methods. I.B.M.'s labs have been at the forefront of “cognitive computing” research in recent years, including project s financed by Darpa.

Mr. Lemnios had praise for I.B.M.'s best-known research project, the Watson question-answering computer. “It moves toward what we think of as understanding information, which is the start of another revolution,” he said.

But Mr. Lemnios said he had also been impressed with the company product and services offerings, which include large contributions from I.B.M.'s research labs, for traffic management, energy conservation and crime prevention. They are part of the company's Smarter Planet projects. “There is a lot of deep technological meat on the bones of Smarter Planet,” he said.

Having smart people is a crucial ingredient in a successful research operation, but so is research strategy and management, Mr. Lemnios noted. On strategy, he said, “balance across a portfolio is important” - that is, balancing work on current products, research bets that may pay off in five to 10 years, and further out exploratory research.

In good times and bad, he observed, I.B.M. has had the managerial patience to continue financing long-range, exploratory research. That is a model, Mr. Lemnios suggests, that the federal government would do well to follow.

Given the need for budgetary belt-tightening in government, Mr. Lemnios said, “There is great pressure to take funding for exploratory research to pay today's bills. That can prove to be a shortsighted mistake.”