Army hosts global AI, machine learning event

U.S. Army researchers hosted an artificial intelligence and machine learning event May 4 that garnered the attendance and attention of hundreds of researchers.

The event, Machine Learning for Everyone: May the Fourth Be With You, provided information about concepts, math and implementation of machine learning and artificial intelligemachine learning shouldnce, or AI/ML, for Army applications.

At the kickoff by U.S. Army Combat Capabilities Development Command, known as DEVCOM, Army Research Laboratory Chief Scientist Dr. Alexander Kott, a short course on artificial intelligence and machine learning by DEVCOM ARL, and an AI/ML panel session showed the connection between the basic concepts of machine learning taught in the short course and the transformative Army research that incorporates these AI/ML concepts and featured five ARL scientists: Drs. Matthew MargeMaggie WignessShowen (Sean) HuGarrett Warnell and Priya Narayanan.

“Today, AI is a truly transformative force, perhaps the most transformative phenomenon that has happened in warfare in many centuries,” Kott said. “At ARL, we have many research projects that explore how to extend artificial intelligence and machine learning methods for solving problems of ground warfare, saving lives of future Soldiers.”

It was a pleasure to be part of an event that educated the next generation of machine learning Jedi, Marge said. His research in conversational AI is very much inspired by droids like C3PO, so he was excited to relate his work in those terms.

“I really enjoyed being a panelist in front of such a large audience, talking about my own research in aided target recognition and unconstrained face recognition, and engaging with the other panelists on Army-focused challenges in AI/ML,” Hu said. “It also gave me an opportunity to have some fun and channel the dark side of the force, relating my research on how Imperial technology could be advanced to defeat the pesky Jedi and the Rebels.”

Researchers said their inspiration for hosting the event served two purposes.

First, what better day to combine machine learning and Stars War than May 4? Host and lead of the event, Dr. Mark Tschopp, has given the short course a few times before, and after some great constructive feedback, was ready to host this next iteration.

Second and on a more serious note, Tschopp said, machine learning and artificial intelligence are being integrated with everything that the Army and Department of Defense does, so this training is crucial.

As Tschopp started to learn more and more about machine learning and talking with colleagues about it, he realized that there are a lot more people out there who want to know more about machine learning, but aren’t quite sure where to start.

“Many courses either don’t go adequately into what’s under the hood or do so, but through a week-long or semester-long course,” Tschopp said. “Many courses focus on the same datasets and the same platforms that everyone is using. I was inspired to do something different, something unique.”

Researchers agreed on what made the course so unique:

   •  Fleshed out the concepts that everyone should know about machine learning
   •  Started simple and slowly built in complexity
   •  Went through the concepts as quickly as possible
   •  Used Army-relevant examples like ballistics (bullets and armor)
   •  Used Excel to show how the algorithms work
   •  Featured an instructor who donned a full Jedi outfit for Star Wars Day

“It is important to train, educate and inspire our workforce in machine learning,” Tschopp said. “Many attendees are being asked to integrate machine learning into their applications, to work with collaborators who use machine learning, to manage programs with ML or to manage people who are delving into this area. A lot of people have questions, but often don’t know where to even start with the learning process.”

When we think of researchers, Tschopp said, we often think of talented subject matter experts who are skilled with using tools to perform critical experiments or simulations to generate data—data that drives decisions.

“Machine learning should be incorporated as a very powerful tool in that toolbox, one that allows us to transform data into knowledge, allows us to ask questions to supplement our own understanding and allows us to ultimately make those important decisions faster,” Tschopp said. “I also hope that this short course dispels some of the hype, the bias and the misconceptions surrounding machine learning. These can create lofty expectations or keep people from using it when it would be valuable – it’s not magic, it has its advantages and disadvantages like most tools.”

Tschopp is hopeful that the attendees walked away with a better sense of machine learning and understand why it is an important tool to be able to work with now.

“Not everyone needs to run machine learning algorithms,” Tschopp said. “That isn’t necessarily the point of this course; however, as machine learning and artificial intelligence are intersecting so many different domains, having a good understanding of machine learning may mean that researchers can effectively appreciate and communicate with those machine learning subject matter experts. Effectively, I hope that this understanding changes the team dynamics from “throw the data over the fence to the machine learning experts” to one where all parties understand the value that machine learning can bring to their research.”

For Tschopp, one of the most exciting elements of this course was the sheer number of people who were signing up and how many different organizations they were coming from. Not just from the large number of organizations around the Army, but around the DOD, around the country and around the world.

“It was remarkable,” Tschopp said. “On one day early on, I had 120 people register that day. I knew it was going to be a big crowd after that point. I hit 700 people two days before and was coming close to 800 on the night before.”

The full day event had 800 registered with over 450 DOD participants, over 175 from other government agencies, and over 150 academia and industry participants from all over the world.

The feedback for the course was incredible, both before the course and afterwards, Tschopp said.

“There was so much excitement about this,” Tschopp said. “I can’t tell you how many exclamation points that were included in the comments—it was a lot. While many were excited about the Star Wars theme and how fun the course sounded, they also indicated that they wanted to learn more about machine learning and the timing of this was perfect.”

Army leadership said the event fostered understanding for many researchers working in support of our warfighters.

“A great demonstration of Army leadership in the field of machine learning, and an incredible job by Mark,” said Gen. John M. Murray, commanding general of the Army Futures Command. “What started as a good idea, some initiative, and a desire to spread knowledge, was met with incredibly positive feedback. His short course showcased not only the challenges associated with building neural networks, but the Army’s efforts to harness these capabilities.”

A common theme of the feedback from the attendees was that machine learning and artificial intelligence are being incorporated into so many facets of the science and technology ecosystem, and yet most people didn’t have formal training as a computer scientist, didn’t know where to start, and didn’t have a whole week to devote to learning the ins and outs of machine learning.

“It just validated that my motivation for this course was on point,” Tschopp said. “And it certainly inspires me to keep improving this course and continue to help others in the Army, the DOD and our partners jump start their learning process about machine learning and artificial intelligence.”

The event reached hundreds of registrants from around the Army, helping educate and train Army scientists, engineers, military personnel, managers, directors and others coming into the course from different backgrounds and application interests.

This course will aid not only in current systems and programs, but in planning for future integration of AI/ML throughout the DOD, Tschopp said.