Tech Juice 2511 Part II: Unsupervised Learning AI Algorithms & The Military

Background

1. ChatGPT and other Pre-trained Transformer solutions have now become ubiquitous with a chatbot-like interface, taking in questions in conversational style (like Alexa or Siri),  and providing intelligent responses formatted for direct use. When Microsoft linked this with their browser, they started an AI war with Google. 

2.  Google was forced to bring up its version of AI to stay ahead of the stock market, and Bard was revealed to the world. Unlike ChatGPT, Bard has already digested the information on the internet and kept incrementally learning. It had already formed a human-like model to transform knowledge into intelligence and respond uniquely as if it were a person. Bard shocked audiences around the world with replies to queries like "summarize the New Testament" or "translate the summary into Latin" etc.  

3.  Legend has it that Ernest Hemingway won a bet by writing a six word story "For sale.  Baby Shoes.  Never worn".   When Bard was asked to finish this story, it wrote a short story with a human touch.

4.  It even translated the single-line story into verse with very human emotions. This unsettling experience shook the whole world, and there have been calls for regulation ever since.

 

5.  Google also acquired a British startup called DeepMind. DeepMind created a chess-playing program through unsupervised self-learning, where the computer played chess with itself. In a week, it learnt chess and beat grandmasters. 

Unsupervised AI & The Military

6. In the past year, we have witnessed the explosion of Generative AI tools with image, video, audio and even movie-making abilities.  While the social aspects will take time to iron themselves out, the military applications will almost immediately become evident in asymmetric conflict zones.   In the medium term, they will start showing up as military products and enhancements for established militaries to upgrade their inventory.  Unsupervised learning AI can be highly useful in military applications for tasks involving pattern recognition, anomaly detection, and decision support.

7.  Digital Battlefield.  The modern battlefield is sometimes referred to as a digital battlefield.  Digital communication infrastructure forms the backbone of this battlefield, with sensors constantly pumping data (image, voice, video, weather, targets, etc.) to a computing system that would be AI-enabled.  Loss of digital connectivity, due to cyber threats and malware, between consumers and data sources can be catastrophic to the outcome of any battle.  AI's pattern recognition can play a crucial role in picking up malware and detecting anomalous network traffic patterns and eliminating cyber threats.  

8.  Battlefield Awareness.  Processing the mass of data that gets thrown up from an area of surveillance is also a major challenge.  Gone are the days when humans scanned monitors to identify unusual activity such as unexpected troop/inventory deployments.  AI is ideally suited for the task as long as it is properly trained for it.  Trained AI algorithms can fuse data from spaceborne, airborne and ground sensors, from multiple spectra (visible, UV, IR, MW, V/UHF) and sensors to make sense of what is happening in a highly complex battlefield.  Electronic order of battle (SIGINT), that was earlier the preserve of a separate domain called Electronic Warfare has progressively blurred to become part of the digital battle field.  Now the electronic order of battle is estimated form the data flows.  Necessary counter measures are estimated from past pattern recognition abilities of AI.  Battlefield awareness is the key to success in today's warfare.  

9. Reconnaissance in a future battlefield will be done by space assets as well as UAS' (Uninhabited Air Systems).   UAS with significant autonomous capabilities can learn patterns in terrain and recognize strategic points without predefined labels.  Based on changes in the observed areas these UAS can adapt to new environments for improved navigation and route optimization.

11.  Drone-UAV-UAS???.   I think the word drone was invented to make sure that the established militaries were not threatened, and thought of this as only required for Dirty-Dumb-and-Dangerous missions.  It was an industry master stroke to make the generals believe that no jobs would be automated and replaced.  'Unmanned Aerial Systems' was also an equally boring name as it suggested that we are using them where a man's presence and on-site intelligence is not required.  Soon, this went into trouble with the female piloting community objecting to the terminology as sexist, although I think they were equally fooled by the intent behind the name.    The focus on the aerial component belies the fact that it is integrated with a significant computing infrastructure either on the ground, in space, at sea, underwater etc. Therefore, progressively, the word vehicle in UAV became Systems and therefore UAS.   Even in the air the 'Drone' operated with other drones as a swarm.  Therefore,  I will use the UAS (Uninhabited Aerial System). 

12.  Swarm technologies have enabled UAS' to operate as swarms.  Whether it is attacking an enemy resource with intelligent swarms strapped with explosives or operating with a manned aircraft in what is increasingly being referred to as a loyal wingman (again, an industry marketing gimmick), swarms are here to stay.  AI can even be used to optimise swarms based on aggregated battlefield data.

13.  Equipment availability and predictive maintenance are vital to ensure success in an equipment-intensive battlefield.  I think there will be extensive application of AI for predicting equipment failures (digital twins) based on anomalies in the large volumes of maintenance data that will flow to these algorithms from the equipment deployed.  Digitising stocks and warehousing can help AI identify inefficiencies and optimise the supply chains based on clustering of past re-supply missions. 

14.  Social Media.  Increasingly, we observe that social media has become the preferred weapon for information warfare for governments and for asymmetric warfare for insurgents.  Social media data can give early warnings and interconnections between insurgents by advanced pattern analytics that AI can bring to the table.

15.  Wargaming. AI algorithms that constantly digest business data and are trained on business strategy have already found their way onto company board rooms.  War gaming is no different.  It involves pattern recognition based on the clustered knowledge of successful and unsuccessful historic battles, scenario generation with a red team, and detecting unexpected behaviours of red teams through real-time anomaly detection abilities.  The simulation abilities of AI can be extended to create autonomous agents performing multiple battlefield roles and collaborating for strategic decision making without human intervention.  AI simulation can be used to expose vulnerabilities in the electronic order of battle and supply chains.  Simulation abilities can also be used to estimate the impact of psychological and information warfare strategies.  AI based strategic decision support may be vital for optimising response of an operational commander engaged in a real war with data emerging from the fog of war.

15.  Conclusion.  Unsupervised Learning AI Algorithms have emerged as a key part of the augmentation in a digital battlefield.  This is a space that will rapidly develop as there are no runners-up in a military confrontation.  In the social sphere, there will be a lot of discussion before AI is applied with regulations, I am sure.  But the military is the space to watch.