The Use of AI in the U.S. Military - Types

AI is the use of computers to do things that were previously thought to only be doable by a human. There are different AI methods in play, including machine learning, deep learning, and neural networks. AI researchers are uncertain about the pace at which AI could potentially advance. Some researchers feel that innovations that are likely to bring about artificial general intelligence (AGI) are just a couple of years away. However, other researchers believe it will be at least a score year before such a breakthrough occurs (Grace, Salvatier, Dafoe, Zhang, & Evans, 2017). AGI is a self-aware AI that can work on many skills and combine the lessons learned from many areas, much like a human. When a person fears the Terminator movies, they are really fearing AGI. However, AI systems of today are narrow systems. Meaning they are really good at performing a very small and specific set of tasks. The most common of these narrow AI systems is Facebook’s facial recognition system that recognizes your friends in your pictures and offers to tag them for you. Additionally, Google’s AlphaGo Zero, created to play the game Go, surpassed all previous version of AlphaGo; including AlphaGo Lee which defeated 18-time world champion Lee Sedol (Hassabis & Silver, 2017). The question now becomes: why do militaries care about a narrow AI system?

First, many industries and military leaders know the speed at which battle is conducted is increasing. Battlefield tactics have evolved significantly. Soldiers no longer stand in a line and shoot at each other from across a field. A military must attack before the enemy is able to realize what’s going on. Striking fast is not the only tactical advantage. For example, when it comes to air defense, a computer’s speed allows a city to be defended more effectively by defense systems. When being bombarded by an onslaught of mortars or missiles, it will overwhelm a person’s reflexes, no matter how much skill and experience he or she has. Israel’s Iron Dome system already operates in this manner to provide exception defense (Work, 2015).

Second, AI will allow us to be more creative on the battlefield. New tactics, such as swarm tactics, will be introduced to allow less exposure to soldiers and pilots. This would put a human tank or airplane pilot in a master vehicle. From this master vehicle or mother ship, the pilot or driver could coordinate several uninhabited swarm nodes. The swarm nodes would need algorithms for obstacle avoidance and basic maneuvering. This allows the human to set objectives and targets without having to micro-manage all the individual swarm nodes.

Third, data is abundant, and it is extremely difficult to keep up with the processing of it all. Narrow AI systems are exceptionally tuned for data interpretation, allowing humans to focus on higher level tasks. For example, the U.S. military wants to accelerate the processing of imagery from UAV and video surveillance feeds. This is the project I mentioned earlier; where many Google employees signed an open letter expressing their concern. This narrow AI application for militaries uses the same algorithms that Facebook uses to recognize your friends in a photo you upload. When you upload a picture of a new friend, Facebook’s AI will tell you what friends it thinks are in the photo. If it doesn’t know, the AI will ask you who it is. In time, the AI will know all your friends and you will no longer need to tag all your photos. Companies use the same technology to identify vehicles, buildings, and people in the video feeds coming from military drones. Instead of having an operator or analyst tag all the object, the AI will. The analyst will make corrections, and the AI will learn to become more accurate. This will improve our ability to make good decisions.

References

• Grace, K., Salvatier, J., Dafoe, A., Zhang, B., & Evans, O. (2017). When Will AI Exceed Human Performance? Evidence from AI Experts. Journal of Artificial Intelligence Research (AI and Society Track), 21.
• Hassabis, D., & Silver, D. (2017, October 18). AlphaGo Zero: Learning from Scratch. Retrieved from DeepMind: https://deepmind.com/blog/alphago-zero-learning-scratch/
• Work, B. (2015). Deputy Secretary of Defense. CNAS Defense Forum. Washington, D.C.: U.S. Department of Defense.