JUST IN: Africom Wants Multilingual AI

iStock photo-illustration

WASHINGTON, D.C. — Artificial intelligence is being increasingly utilized for a variety of different tasks, but U.S. Africa Command has a unique challenge for it: help understand under-resourced languages.

Africom’s area of responsibility includes more than 50 countries, spanning the entire continent of Africa, minus Egypt.

“We have a significant amount of different dialects and languages on the continent, so one of the focus areas is getting into … the low-resource, low-density languages, so that we can increase our communication pipelines with various different countries,” said Air Force Lt. Col. Jared Bindl, chief science, technology and innovation officer at Africa Command.

Artificial intelligence and machine learning have a “big impact in that translation space but may or may not be focused on the lower-density languages,” which are considered under-resourced with less training available for new speakers, Bindl said during a panel discussion at the National Defense Industrial Association’s Emerging Technologies for Defense Conference and Exhibition Aug. 28.

“We want to continue to push” making language databases more robust within the command, he said. Among potential resources that could help are language-agnostic speech technology and rapid response automation for language identification that specifically target low-resource, low-density dialects, Bindl noted in a slide. 

Another unique geographical challenge the command faces is operational energy, which Bindl listed as another science and technology focus area for the command. 

“We are operating in extremely austere environments,” Bindl said. “We have to look at creative solutions” for challenges like purified water and tactical power generation, “since we're not going to have big hubs where we operate out of. So, we have to have this tactical energy unit. This will continue to be a challenge for the Africom [area of responsibility] as long as we’re operating there.”

Among the technologies Bindl said the command is on the hunt for are micro and austere operational energy solutions, unique water creation and purification systems, water-to-energy disposal and energy-efficient structures for housing and offices.

Lastly, Bindl said Africom is focused on using artificial intelligence and advanced data analytics to get commanders real-time data as quickly as possible and “get off the PowerPoints and really start seeing what they can do for quick decision-making.”

Bindl said a goal is to better leverage “everything that’s going on in the Al/ML space,” and becoming more efficient is the “real, main thrust area that we’re focusing on with AI/ML, big data.” 

Specific areas for which the command could use artificial intelligence and machine learning include biometric and forensic support, operational medicine information systems and mission partner bi-directional support, he said. 

Army Lt. Col. Timothy Hodge, requirements branch chief for U.S. European Command, offered a word of caution about artificial intelligence, particularly for austere and forward-deployed commands. 

“AI, in its current iteration, it's very fat, for lack of a better description,” Hodge said. “It is resource intensive, it's compute intensive … and that is really not conducive for forward deployment.”

“We are less concerned [about] maybe the next giant generational leap in your AI model. We are definitely concerned on you being able to run it on … something that doesn't require a giant cooling facility fixed site,” he continued. “Frankly, if nothing else, it helps us avoid having to build all that infrastructure. Because without some sort of low-power cost, low-compute cost AI model, it will be relegated back to major fixed headquarters, and its ability to influence the battlefield in forward positions is going to be very hard.”

 

Topics: International, Defense Innovation