By George Leopold, Datanami.com
/ Published March 22, 2016
Cognitive computing is being enlisted in the latest effort to fix the hidebound U.S. weapon acquisition process.
Specifically, IBM’s Watson has been tapped by the U.S. Air Force to develop new artificial intelligence tools to help manage cumbersome acquisition rules while helping procurement officials and prospective contractors navigate the swamp that is federal purchasing guidelines.
According to a report in the Washington Post, the Air Force was to deploy Watson as a potential “bureaucracy buster” as the service responds to growing calls in Congress to fix the way the Pentagon buys weapons and services. Acquisition reform efforts tend to resurface about once a decade, but this time lawmakers are concerned that faster technology cycles will make costly weapons obsolete by the time they are deployed.
The Air Force, which is currently overseeing the most expensive weapon program in history—the F-35 fighter aircraft, which is estimated to cost about $100 million per copy—is taking the lead in the latest acquisition reform push. Air Force officials to the Post that Watson is uniquely suited to helping program managers and contract bidders navigate the formidable Federal Acquisition Regulation, which currently totals nearly 1,900 pages of bureaucratese.
An Air Force acquisition official noted that Watson could be used to digest the massive and often impenetrable purchasing rules. Watson’s insights could then be used to help program managers determine, for example, how to structure contracts in order to get equipment into the field faster while eliminating costly duplication and schedule overruns.
The newspaper reported that the Air Force has hired two contractors to “train” Watson. The first step is loading the acquisition rules, then building up its knowledge base by asking the platform questions that will, program officials hope, help the cognitive computer understand context and nuance buried in U.S. acquisition rules. The effort also will attempt to take machine learning a step further by combining cognitive computing tools with human understanding of often-inscrutable procurement regulations.
The technology push comes as legislation has been introduced in Congress in the latest attempt to fix the broken Pentagon procurement system. The proposed Acquisition Agility Act would attempt to rein in soaring weapon costs by imposing stricter milestones on development while providing new incentives for contractors to deliver weapons on time and within budgets. It also would expand input from the uniformed services that actually use new weapons, meaning service acquisition officials would have “more skin in the game,” according to the Washington-based National Center for Policy Analysis.
The first test for the reform proposals and for emerging tools like Watson is a massive Air Force program to build the next-generation long-range bomber. Each plane could cost more than $500 million. In a sure sign that the weapon procurement system is broken, the losing bidders in the competition to build the B-21 bomber—the team of Lockheed Martin and Boeing—immediately protested the Air Force decision to award the prime contract to Northrup Grumman Corp.
Indeed, losing bidders for most major government contracts often file a protest with the Government Accountability Office, further slowing the procurement process. At this point, it isn’t clear how Watson could be trained to account for these vicissitudes of government contracting.