Artificial Intelligence: A conversation with Brig. Gen. Balan Ayyar, USAF

By: Taylor E. Watson

Inside the flight simulator at the dogfight competition run by the Defense Advanced Research Projects Agency (DARPA). The competition was created to determine if man can beat machine in the sky. (Photo courtesy of DARPA)

TEW: Please share a little about your Air Force career.  

BA: We had a spectacular 26 years. Wouldn’t change a thing. Our family loved every minute of it. My career was everything a young immigrant banana head like me could have hoped for. I was a pilot. I had the opportunity to fly jets in India with the Indian Air Force. I loved my time as commander of the Air Force Recruiting Service, and our family had a totally capstone experience at Randolph AFB. My culminating experience was commanding forces in Afghanistan. Coming out of that assignment, it was time for our family to make a new life together. Whatever might happen, I knew coming out of my service in Afghanistan that whatever work I would do next, I wanted to improve the way we were using technologies like AI and machine learning to protect and save the lives of our servicemen and servicewomen. 

TEW: Can you please share a little about, and the work you’re doing now? How do the public and private sectors interact around innovations and developments in this space? 

BA: When I left the military, I had the great opportunity to work with another government services company as their COO and ultimately CEO. Then, I went to Silicon Valley to pitch this national security and intelligence company that would be a center of excellence for AI. I wanted to operationalize this technology in a way that would advance our interests and protect our values, culture, and nation rather than put them at risk. At the convergence of national security and AI technologies, it was clear that leaders in this space had been so inward facing there was inoculation against innovation. CACI, one of the leading companies with government contracts in this area has thousands of employees, but not a single patent. There was an opportunity for government services companies to try and build capabilities in this business space through partnerships with innovation leaders in the private sector. We went to Silicon Valley to recruit scientists, engineers, and product developers who had helped develop Google Maps technologies and worked with companies like Uber. My top scientists are about five individuals that were leaders in these technology areas; together they hold dozens of patents and have written hundreds of papers. Our main product is Mirage, and what it does is accelerate analysis of hours and hours or several years of multi-source data that would consume an overwhelming amount of a human analyst’s time. There’s so much data that needs to be processed and the quantity of that data is accelerating. Using a tool like ours to deliver focused input quickly regarding areas or persons of interest that are pertinent to the mission can help national security agencies, the intelligence community, or military operators more effectively make decisions. Private sector innovations are critical to bring capabilities to bear for our Air Force and the profession of arms more broadly. 

TEW: Broadly speaking, what is AI? What is important for the public, more specifically the military aviation community to know or understand about AI?  

BA: To summarize – it’s spectacular. There’s no such thing as general AI. No atomic detector is out there like in the movies. A machine can detect a watch, but not know it’s your watch, right? It’s just one in a class of watches. That’s the nature of machine intelligence – it can do things the human brain doesn’t really like to do, even though the human mind is its own cosmos. In terms of processing vast quantities of data, our neuron pathways are like those of a small lizard compared to what a tool like Mirage can bring to bear. 

Because it’s massively paralyzable – there’s much more data than the human can understand at one time. What artificial intelligence can do for humans is help us properly use machine learning so that we can reach a higher and faster level of understanding than would be possible otherwise. Artificial intelligence can process large quantities of data at lighting speed in ways that our brains are more like lizards – but compared to the human mind machine learning has the neural processes of a snail. It is effective at helping humans contextualize much faster than previously.  

In many ways, we’re already in the future. Google is already in every part of your life. Whether it’s a FICO score or your search engine, results are being tailored to you based on what you like and how you behave. It matters in the national security and intelligence space because there’s too much data. We have satellite images and street cameras and video feeds. There is so much information available that nothing can be understood fast enough to be relevant.  

You don’t just keep putting more and more instruments in the cockpit. At some point, the pilot is completely undone. There’s too much to look at it – you can’t do it. There’s a certain point where more information doesn’t help her make decisions. There’s a need to have something triage all that information to get what’s important to the pilot. 

TEW: Why has AI been introduced as an important technology for military applications? 

BA: This is critical technology for military applications because precision and decisiveness are critical to success. The ability of AI to help us understand the world allows us to be more precise and intentional as we make decisions. Just because you can listen to every frequency in the world doesn’t mean you should. You have got to listen to parts relevant to your mission.  

When I was serving in Afghanistan, we had a situation where there was a truck that exploded near a secure location, resulting in some deaths and loss of life. Fortunately, the truck or its driver had experienced some confusion and had done a lap around where we were before the explosion took place, which reduced the impact. As it turned out, we had perfect intelligence right after the event happened. Civilians had been moving away from the truck, there were indicators that things about it weren’t right. Now if we had this intelligence before the attack, we could have taken steps to prevent the tragedy. 

From an airman’s perspective, these technologies can confirm where there’s potential collateral that may be at risk. We never want to harm someone if we don’t have to. The ability to detect changes as they’re occurring in very near to real time is critically important for us. When you’re downrange or in a live fire environment, this is very relevant to the profession of arms.  

TEW: Within the military, and specifically the aviation community, where is AI already being used? What kind of impact has it had? What are some current, significant developments around AI? 

BA: The real thing that brought me to do this was the advent of neural nets, made possible by the enormous computational power we now have. Neural nets and deep nets allow machines to look at every frame – say 30 frames per second on a cell phone camera – and use machine intelligence to detect patterns it’s been trained to recognize like faces, objects, cars. We’re able to create a 128-dimensional space; each object has its own cosmos or constellation of points that is like a signature. By analyzing all those 128-dimensions and embedded points, we can determine that my face looks like my mother’s or my dad’s, but not so much like Taylor’s. We can merge that analysis with other constellations, compare all those embedded points and identify patterns, reembed it with say a video feed and be able to identify a face that looks like Balan. I can look at thousands of hours of still images and videos in a few seconds, and can find where Taylor and Balan have ever been seen together. How is this practically useful? 

My firm builds a platform called Mirage and we use air, space, and ground data to correlate and geocorrelate objects, persons, vehicles, in a way that helps you understand pattern of life and what historically may be helpful in planning a mission. Why does this matter to an airman? With lower earth orbit satellites, the moment the sensor touches the target area a tool like Mirage can process what it’s seeing. For example, we can create a geofence around San Antonio. A picture is taken, Mirage processes the photo and can determine “Here’s Taylor’s car at this office location in San Antonio; from the picture we can see it’s parked at this location at 5 AM and leaves at 8PM at night over a period of time”, and draw a conclusion that maybe Taylor is working too hard.  

That ability to characterize what’s happening over a period of time is very powerful. We can see, say “Uh oh there’s a lot of forces that seem to be moving west”, and then geofence around aerial ports and shipping ports. We can then program and say “Hey, let me know when the first tank is seen at point of embarkation”, since we know that’s an indicator that battalions and brigades could be deployed at that point. The first indication that our adversary is doing this helps preserve our freedom of action as we try to detect and understand in non-permissive environments.  

Before Mirage and these kinds of tools, a human operator would have to get and look at all that data, and analyze what they’re seeing. For a long time, human tradecraft was all we could rely on. Nowadays, there’s so much data everywhere, even in non-permissive environments, it could easily overwhelm a human operator. If we can understand what all the data means, there’s much more freedom of action. 

TEW: Looking forward, what could the major impacts of AI be? Is it a game-changer? Why or why not? 

BA: In the world there’s structured data and unstructured data. The difference between what Google does versus what Balan’s firm does is a framework engineered around what mission problem we’re solving. Our goal is to accelerate and elevate what the human can understand. What the machine can learn is strengthened by human experience.  

For example, in the case of Russians operating in Syria… by reviewing all of the data we have, we can learn more about the movement and intent of Russian forces. In tandem with a human SME expert, they can learn from each other in a reinforcing way. Before AI, say you have a geospatial analyst who is your top person on Russia. If they were the only agent analyzing all of your Russian data, much of your knowledge and information in that space leaves if they leave the company. Because AI can learn from your analysts and how they work, what patterns they detect, and build on their knowledge, your organization is learning from its people in a way that makes your enterprise value skyrocket. An organization like the National Intelligence Agency can hold on to a lot of the learning that’s already been done to help others improve and make better decisions in the future.  

Platforms like Mirage can also be repurposed for other uses, like climate change, without the expense of additional analysis. Say we were monitoring a Russian air base to evaluate the readiness of Russian forces by evaluating how many armed and loaded bombers are on that base. If we can detect this before the enemy knows we can, we have increased understanding and better leverage as we make choices and decisions. But, in gathering and processing images of that base, we can also see how quickly the lake next to that Russian air base is changing, helping us understand the impacts of climate change in that region. All that is visible from space. With AI, human understanding from visual intelligence is at an all-time high. 

TEW: What, if any, practical challenges and/or pitfalls around the use of AI for military applications do we face? What questions and concerns, if any, are involved in the discussion of the ethics around the use of AI for military applications? 

BA: Foundationally, AI is no different than the story of all humankind. As humans progress, humans and machines work together to accomplish an end. Whether learning to fly, or taking us to space, for generations we can see humans and machines working together to achieve a common purpose. Machines and technology help us to accomplish things that heretofore we have not been able to accomplish. In many ways the future in 1910 is no different than the future now. Not many people thought the airplane was going to be practical or relevant long-term, and look at its impact today. 

We are really at the very beginnings of human and machine teaming; things are very nascent, even though that experience has been with us for some time. Our ability to trust these new technologies and to work with them effectively will determine very much whether this is an American century or not. 

As for pitfalls, it’s important that we manage risk and work to gain an accurate assessment on how much can be understood with confidence. It’s important to recognize that every mission is unique and has its own level of risk. Our platform allows the operator to control risk and manage algorithms. Keeping that balance between human understanding and autonomous systems is important.  

Informed by our culture, our way of life, our love for our fellow man, we are trying to avoid collateral damage and minimize the risk to human life whenever possible. We want to avoid unnecessary loss of life when we’re in theater. We don’t want attacks in America like we’ve seen in France and Germany. AI and machine learning tools and platforms can help us build the intelligence to understand patterns to make more informed decisions more quickly. If we’d had some of these tools in place in these instances, some of these attacks might have been avoided. 

TEW: NSCAI just issued a report recommending acceleration in the use of AI technologies. Other groups such as the Red Cross and the EU have put forth for consideration a global ban on autonomous weapons and regulatory frameworks. Do you have any comments or perspective on this? What resources might be helpful to understand these positions more effectively? 

BA: Wonderful folks. Lots of people with great consciences. The NSCAI report that was recently released is a great report. CSIS also produced a report around the new year that was very well done. Acquistion of AI has got to be a priority for the nation; it is increasingly important for the 21st and 22nd century. We want to use these kinds of tools to challenge bias rather than allow it to be reinforced. With the world as it is today, the level of trust someone has in institutions or law enforcement is largely based on their own experience. Having the capacity to look at an ever-increasing amount of data in seconds rather than over the long-term can lead us down a more responsible path in ensuring we are doing the right things. 

While there are lots of good things about AI, there are some things to be concerned about. Russia wants to participate in the liberal order or liberal democracy, but in a somewhat malign way. They are using some of these tools to deliver false information and create tension inside America to attempt to polarize us and make us less effective, and to some extent, they’ve been successful. On the flip side, we can use AI to detect that kind of work and help prevent our country from coming apart internally due to the creation of that tension. 

TEW: What is your philosophy of public service? From your perspective, how is public service important right now? 

BA: What a question. There are three reasons why I think it’s important: 

First, public service not about the public. It’s about helping you develop character, becoming part of something bigger than yourself, and building something for the next generation. Serving others preserves our experiences as Americans. 

Second, it’s not so much about you as about knowing the meaning of rights and responsibilities and preserving a culture that respects individual effort. We create self-respect and confidence as we volunteer of ourselves to help others and our government do things that it doesn’t do particularly well. 

The third thing it does is give you a sense of ownership as a citizen that you come to feel when you volunteer and contribute to overcome and provide solutions to challenges that are real. As we do this work, we learn a sense of responsibility for our neighbor and community that can be transformative and we learn not to be trapped in a destructive, self-absorbed, me-first way of thinking.  

If you look around Washington, D.C., there are no statues to billionaires. We erect statues to remarkable individuals such as MLK. Ultimately, we honor those with strong character, and there are no better means of developing that than through serving others. The end of all education is to build character. When I have love for others, volunteer, or work in the public sector, I feel better about myself. Helping someone else rather than buying a seven-scoop sundae is the best way to feel self-worth. This work isn’t because anyone else needs it- you need it in your life every day. 

Brig. Gen. Balan Ayyar, USAF (Ret) is the Founder and Chief Executive Officer (CEO) of, a Silicon-Valley based artificial intelligence firm with a business focus on national security and intelligence missions. His final role as a general officer was as the Commanding General of the Combined Joint Interagency Task Force 435 in Kabul Afghanistan. Prior to his Afghanistan assignment, he was the Commander, Air Force Recruiting Service. He has served in four combatant commands, as the Military Assistant to the Secretary Defense, and in the White House as a White House Fellow. He is a member of the Council on Foreign Relations. General Ayyar was commissioned in 1987 as a graduate of the U.S. Air Force Academy. He served as a command pilot with more than 3,500 flying hours in the T-37, T-38, C-141B, Kiran Mk IA, KC-135R, C-17A and KC-10A. He is Daedalian Miss Watson has known the Ayyar family for over ten years.