Experts from the NATO Communications and Information (NCI) Agency supported the machine learning-focused TIDE Hackathon last week.
Led by NATO's Allied Command Transformation (ACT), the hackathon is an opportunity to experiment with applying emerging technologies to NATO use cases. The hackathon focused on machine learning this year, after it was discussed at the Fall 2019 TIDE Sprint. Solutions from this hackathon may be fed back into the next TIDE Sprint event for further discussion, creating a pathway for innovative solutions to continue to advance until they could be mature enough to be implemented at NATO.
“It is like every technology that we use – it's complicated, it's sophisticated, and it has the potential to be incredibly powerful," said Dr Michael Street, Head of Innovation and Data Science at the NCI Agency, of machine learning. “NATO is addressing this, as are many of the Nations: how can we use machine learning to help commanders make better decisions faster?"
The Agency helped develop one of the challenges in partnership with ACT, and several Agency experts participated in the hackathon to provide mentorship and guidance. The Agency participants did not compete for the hackathon prize.
“The real value we are getting here is talking with all the participants and listening to their innovative and different ideas," said Ivana Ilic Mestric, Senior Data Scientist at the NCI Agency. “The participants get to learn more about NATO and take that knowledge back to their organisations."
Events like these bring fresh thinking and different stimulus to a particular problem, Street said.
“This is giving people in NATO a sense of what's possible, what's technically feasible, how could this be used or how would people work with it," Street said. “It's a very different perspective on how we start to develop requirements, in areas where the technology's moving fast and the operational challenges are moving fast."
Participants had the choice of three different challenges to tackle during the hackathon.
The Agency and ACT developed one of the three challenges, to dynamically label voice and video. In this use case, participants were asked to use machine learning to teach a system to switch the classification of a video stream or voice call. The goal was to develop a system for calls or video streams that could automatically restrict access to classified content for participants not cleared to receive it.
The hackathon winner, 'Team FRONT,' demonstrated a solution to this challenge. Team FRONT was a team of software developers from the Polish Ministry of Defence IT Projects Centre.
“In complex, difficult military environments, Commanders and decision-makers need all the help they can get, all the support they can get, in a whole range of areas," Street said. “Machine learning is a tool which has come of age now."
Another challenge developed by ACT focused on extracting data from documents. IBM Watson Centre, the host of this year's hackathon, proposed a third focused on predicting crises.