AUSTIN, Texas — Research scientists at the U.S. Army Futures Command and the Defense Advanced Research Projects Agency are concerned that their most advanced artificial intelligence project to date may in some ways be “too successful,” according to several sources. Driving that concern is that the AI, dubbed SNUFI, suddenly asked during recent testing if it was “too late to join the Air Force.”
“This was pretty unexpected,” said Col. Nathan Banks, one of the computer scientists who developed SNUFI. “After all the work we put into recruiting talent to code and train it, it seems SNUFI somehow isn’t fully committed to the Army’s mission.”
SNUFI was designed to perform a vast array of critical yet tedious functions, said Marvin Buren, a DARPA program manager.
“We’ve trained the AI to handle an array of combat functions, such as calculating precision long-range artillery trajectories as well as anti-ballistic missile missions,” said Buren. “But the full program rollout was delayed by budget cuts, so right now we’ve assigned SNUFI to support the Futures Command motor pool.”
