Deep in the intelligence community, a project seeks to build a model for detecting the future. Its goal is to someday give analysts a way of knowing what technologies are gestating in labs and research houses. If analysts knew which ones would emerge as successful products, such early knowledge might give the U.S. an edge in maintaining defensive superiority.
But first, can such a crystal ball technology work? That’s the question confronting Dewey Murdick, the geeky program manager running the Foresight and Understanding from Scientific Exposition, or FUSE, program within the Intelligence Advanced Projects Research Activity. FUSE, part of the Office of Incisive Analysis, fits right in with IARPA’s stated mission to invest “in high-risk/high-payoff research programs that have the potential to provide our nation with an overwhelming intelligence advantage over future adversaries.”
The FUSE project awarded contracts to BAE systems, Raytheon, SRI International and Columbia University. Along with a slew of academics and specialty small businesses, they’ll scan, by computer, 30 years’ worth of technical, academic and trade literature published in a variety of languages. They will develop algorithms that look at patterns and changes over time to see if a successful technology’s eventual emergence can be detected far in advance.
Initial work won’t try to predict anything, Murdick explained. Instead, the researchers will take technologies known to have emerged, and test theories of emergence evidence using hundreds of millions of pages of material published in the decades before to see if a model can successfully detect the emergence.
He cited flash memory, a hugely successful technology now found throughout business and consumer electronics. It is based on electrically-erasable programmable read-only memory (EEPROM and EPROM), which was theorized in the 1970s, became a lab specimen in 1980 and was publicized widely starting in 1984. What might be a mention in some obscure journal eventually blossoms into published material by competing scientists, then engineers and marketers. Finally it might hit market. So Murdick’s work aims to see if patterns in vast data troves are discernible and later translatable in a way useful to intelligence analysts.
“We’re trying to find out if this is even feasible, to prioritize what an analyst should look at,” Murdick said. “We’re trying to develop the capability to help avoid future surprises.”
He contrasted this type of inquiry from a standard search.
“If you want to learn about anything today, you go to Google or any of the search engines. You get an immense amount of information. But what if you don’t know what you’re looking for? Can you find what you need to know without knowing what you’re looking for?”
Murdick said FUSE embodies three basic areas of inquiry.
First, Murdick said, is, “Can you identify what it means for a technology to ‘emerge?’” Second is developing text indicators of emerging technologies. This involves speed reading ⎯ by machine ⎯ those hundreds of millions of pages of material to see if there are patterns in the way things are stated or laid out that indicate future emergence.
Third, and most exciting to Murdick, is discovering whether the importance of emerging technologies can be determined, and future emergences ranked according to significance, together with the “why.”
Murdick said that within a narrow domain ⎯ memory, disk drives, some branch of software ⎯ a given expert can scan the subject-related literature and have a good idea of what’s going to happen. But FUSE is thinking bigger than that.
“What if your responsibility is to look over the entire world for things that could impact national security. You don’t know what will,” he said.
Murdick said it will take time to build such an engine capable of (in IARPA’s words) detecting technical emergence, presuming it can be. “This is really hard. IARPA is taking it on for five years. I’m excited, sanguine about it. I believe it will happen,” he said. But, he added, he is trying to manage the team of contractor researchers to maintain scientific rigor and cost discipline in the project.
Murdick, who is as a computational physicist, first came into the government as a scientific and technical analyst for the Army. He’s been a programmer, developing payroll and inventory tracking systems in COBOL. He’s even taught elementary school, briefly. His education background predicts the kind of deep scientific inquiry he’s involved with now:
He received his PhD in engineering physics from the University of Virginia, where his advisor was an accomplished inventor, Hayden N. G. Wadley. Murdick’s dissertation was: “Simulating the atomic assembly of gallium arsenide.” He was a member of the engineering school’s committee for the enhancement of research culture. Murdick’s undergraduate degree is from Andrews University, where he majored in physics and minored in Spanish. He was president of the physics club.
As is often the case, his mathematics orientation is accompanied by musical interest and talent. Murdick is a trombone player who has performed in orchestras and a brass quintet. Because of apartment noise considerations at one point, he also took up guitar, where his main medium is bluegrass.
But most of his time today is taken up as program manager in the realm of federally backed science. His overarching goal: “A reliable, scientifically valid way for analysts to improve national security.”