A Mercyhurst College study on the future of predictive analytics, which examined the outlook for intelligence analysis in the computerized age, shows machines not yet capable of detecting deliberately deceptive data.
The artificial intelligence program "Watson" may have outsmarted human competitors on the television quiz show "Jeopardy!" recently, but it would have to go a long way to best an intelligence analyst, according to Kristan Wheaton, J.D., associate professor of intelligence studies at Mercyhurst College.
On Feb. 14-16, the reigning champions of Jeopardy! Ken Jennings and Brad Rutter faced a formidable new competitor a supercomputer named Watson under development by IBM for four years. Watson defeated his adversaries handily.
Wheaton's graduate students recently completed a study on the future of predictive analytics, which examined the outlook for intelligence analysis in the computerized age, including emerging technologies similar to IBM's "Watson." They delivered their findings Feb. 21 to a representative of one of the nation's leading analytic organizations, which sponsored the study as part of Wheaton's Strategic Intelligence class. For security reasons, the organization declined to be publicly identified.
In their study, Wheaton's students' discovered that the technologies that would completely replace humans with machines in the field of intelligence analysis are not looming on the horizon.
"What you saw on Jeopardy would not play out the same in the world of intelligence analysis today because none of these new technologies including Watson - deal well with deliberately deceptive data," Wheaton said.
Lindy Smart, one of the student analysts on the project, explained why. "While there are technologies that can extract data from unstructured sources like e-mails and blogs, they are unable to identify the validity of those sources," she said."For example, let's say you want to do a search about mining practices in African countries. The software could extract data from every single type of source identifying mining practices in Africa. However, it would not identify what data came from state-run news sources that would most likely have skewed the data. Humans are still needed to identify and weed out which sources are deceptive and how reliable the data is."
Wheaton added, "The technologies are improving rapidly, though, and there might be machines capable of mimicking human intelligence at some point, but the last job I think they are going to get is that of the intelligence analyst."
Explore further: Programming smart molecules: Machine-learning algorithms could make chemical reactions intelligent