AI Library
The Age of Autonomous Scientific Discovery
Kim Kyung-jin, Attorney at Law
AI Scientists and Self-Driving Labs
This book follows how AI scientists and self-driving labs are changing the way science generates and verifies claims. It covers literature-based discovery, natural-language protocols translated into robot commands, multi-agent research systems, closed-loop laboratories, materials search, the verification gap, chains of evidence, research harnesses, journal ethics, and legal responsibility.
AI Library
A New Era of Life Sciences Opened by Artificial Intelligence
Structural Proteomics, Genomic Foundation Models, Autonomous Laboratories, and Global Governance
Kim Kyung-jin, Attorney at Law
This book is a research volume compiled with artificial intelligence. A human selected the materials and structured the work, while AI models drafted the sentences and cross-checked the facts.
AI Library
The Double Structure of Digital Sovereignty
Europe’s Departure from Palantir and the Chains of American Big Tech
Kim Kyung-jin, Attorney at Law
This is a record of 2026, when European intelligence agencies and defense ministries began removing analytics tools from America’s Palantir. It covers the replacement decisions made by France’s General Directorate for Internal Security (DGSI), Germany’s Federal Office for the Protection of the Constitution (BfV), and the Netherlands Ministry of Defense; the incident in which US export controls severed an ally’s ac…
New English Edition
Artificial Intelligence in Horticulture
Kim Kyung-jin, Attorney at Law
Across five chapters and ten sections, this book examines computer vision for crop diagnosis, harvesting robots and autonomous field systems, smart greenhouses and digital twins, precision irrigation and supply-chain quality control, high-throughput phenotyping, and predictive breeding.
New English Edition
Artificial Intelligence in Food Crop Agriculture
Kim Kyung-jin, Attorney at Law
Across six chapters and eighteen sections, the book examines digital agricultural infrastructure, remote sensing, crop diagnosis, yield forecasting, precision irrigation, genomics, molecular breeding, agricultural robotics, climate-smart agriculture, and global food security.
New English Edition
The Future of Forestry and Agroforestry
Kim Kyung-jin, Attorney at Law
Driven by Artificial Intelligence and Digital Innovation
Across five chapters and fifteen sections, the book follows satellites, drones, LiDAR, digital twins, forest-specific language models, wildfire and pest forecasting, forestry robotics, agroforestry, timber traceability, and forest carbon markets.
New English Edition
Smart Livestock Farming: AI Enters the Barn
Kim Kyung-jin, Attorney at Law
Sensors listen, cameras watch, and artificial intelligence helps farmers decide.
Across five chapters and fifteen sections, the book follows precision livestock farming from animal health and reproduction to robotic milking, virtual fencing, digital twins, methane reduction, welfare, and data ownership.
Table of Contents
Han Dong-hoon, Busan Buk-gu Gap: A Record of the 100 Days Before and After the Election (Mar. 26-Jul. 3, 2026)
Kim Kyung-jin
Table of Contents and 13 sections
From March 26 to July 3, 2026, this record follows the spring after expulsion, the Busan Buk-gu Gap by-election, victory as an independent, and the first bill submitted in the National Assembly.
Table of Contents
Artificial Intelligence and Medicine
Kim Kyung-jin, Attorney at Law
AI in clinical care, hospitals, education, and research
AI in medical imaging, risk prediction, treatment planning, hospital operations, education, and research, with patient safety, privacy, and accountability.
[AI Library] Chapter 7. Heron Systems' Breakthrough: AI Armed with Reinforcement Learning
Chapter 7. Heron Systems' Breakthrough: AI Armed with Reinforcement Learning
Heron Systems' sensation: Lockheed Martin was the strong candidate to win the AI competition armed with reinforcement learning. They created the most powerful fighter planes in existence, such as the F-22 Raptor and F-35 Lightning II, and knew better than anyone else about the physics of fighter aircraft and the doctrine of air combat. Decades of aerodynamics know-how, thousands of engineers, and an astronomical research budget. Anyone could see that they had the advantage. On the other hand, Heron Systems was a small software company with about 30 employees. They had never built or flown a fighter plane.
It seemed odd that this small, Maryland-based company would stand alongside the defense giants. But they had a secret weapon. It was a fanatical obsession and dedication to deep reinforcement learning. The principle of reinforcement learning is simple. You choose an action, see the consequences, receive a reward or penalty, and choose a better action next time. Repeat this at a frantic pace. It's similar to a child learning to walk. Fall, get up, fall again, get up again. But AI is different from a child. You don't get tired, you don't get frustrated, and you can fall and get up a million times a day.
Heron Systems' approach was completely different from that of traditional defense companies. Teams like Lockheed Martin and Aurora have tried to inject the knowledge of fighter pilots into AI. I tried to teach them rules like bite your tail and conserve energy. This is called an expert system. But Heron was different. They didn't teach the AI anything. Instead, they were thrown into a virtual environment and made to repeat the process of killing and killing countless times. Heron's AI agent was called Falco. The name comes from the falcon, a type of falcon.
Falco grew up through self-play, constantly fighting against himself in a virtual space. The process of today's me developing new tactics to overcome yesterday's me, and tomorrow's me breaking them again, was repeated infinitely. According to Ben Bell, a machine learning engineer at Heron, Falco fought billions of dog fights against a league of 102 different AI agents over a total of about five weeks. More than 4 billion simulation steps were performed. Converting this to human flight time, it amounts to over 30 years of accumulated flight experience.
Considering that it is difficult for an actual pilot to fly more than 2,000 to 3,000 hours even if he or she flies a lifetime, this was a triumph of compressed time that transcended the constraints of physical time.
During this process, an interesting phenomenon was discovered. During the initial learning phase, the AI tried to mimic standard maneuvers taught by human instructors. It was a textbook move to manage energy and maintain turn rate. However, as learning continues, AI begins to abandon human doctrines. Instead, we sought the extreme efficiency allowed within the physics of the simulation engine. The human pilot tries to control the aircraft smoothly so as not to miss the enemy. Heron's AI was different. The aircraft was controlled by constantly modifying the control surface in fine and rough ways. Dozens of corrections per second.
It was a movement impossible by human hands. This creates an incredibly precise shooting angle. Two sophisticated techniques were behind Heron Systems' success. This is reward formation and curriculum learning. Reward shaping is a way to provide more frequent feedback to AI. Rewards are not only given when you shoot down an enemy, but also when you gain an advantageous position or when you get close to the enemy's tail. This will help the AI learn faster in the right direction. Curriculum learning is a method of teaching starting from the easy things.
At first, you'll fight enemies that only fly in straight lines, and then you'll face increasingly smarter enemies. It's like going from elementary school to middle school to high school. Heron combined these two techniques with the Colosseum's self-confrontation system to create a powerful performance. Enemy AI agents developed by APL ranged across a broad spectrum in complexity and capabilities. The simplest zombie simulated a cruise missile with straight, horizontal flight. Rosie, the primary agent, implemented minor time-based altitude and speed changes.
As a script agent, BUD FSM recognized the engagement situation and implemented a predetermined response. Alphamarve0, the highest level reinforcement learning agent, is an agent developed solely through self-confrontation without human input and simulates an expert pilot. On the third day of the competition, as the semifinals and finals progressed, Heron Systems' AI showed off a shocking tactic. It was a frontal attack, a head-on gunshot. This is a tactic where you face the enemy plane head-on, run towards it, and fire your machine cannon. Training regulations prohibit this maneuver for human pilots due to the risk of collision.
Usually do not shoot at angles greater than 135 degrees. There is a high risk of colliding with each other, and fragments of destroyed enemy aircraft may be sucked into my engine.
However, for an AI without fear of death, this was the most likely winning formula. Heron, who met Team Aurora in the semifinals, rushed towards the front of the enemy at breakneck speed as soon as the battle began. Glock, who was in charge of commentary, was astonished. “Heron has no hesitation. As soon as the starting signal is given, he shows his will to live and aims for the bridge of the enemy’s nose.” Heron's AI put this tactic to good use in the final against Lockheed Martin. Lockheed Martin's AI mirrored human doctrines, attempting to manage energy and dominate positions. It was an elegant and standard move.
But Heron's AI refused such a gentlemanly fight. Like a mad dog, they rushed at each other and preyed on the opponent's weaknesses. The Heron aimed its cannons with incredible precision. The muzzle of the gun stuck to the enemy plane like a magnet. Once I caught a target, I never missed it. In a split second, he dealt a fatal blow to the enemy plane. Even in a situation where the human eye judged that it was impossible to shoot from that angle, in the AI's calculations, the chance of hitting was over 90%. In the end, David defeated Goliath.
Heron Systems defeated Lockheed Martin in the finals with an overwhelming margin of 16 wins and 4 losses to become the champion. A small team of 30 people defeated a slow defense industry dinosaur with tens of thousands of engineers. This was an incident that showed that data learning ability can be more important than domain knowledge. The methodology for training AI was more decisive than the experience of building a fighter jet. Heron Systems' breakthrough was an event that proved that AI can go beyond human imitation and create a new grammar of victory that humans have not discovered through data and simulation.
And now, waiting before them was their final opponent, a human pilot.
Kim Kyung-jin
Attorney · Former Member of the National Assembly · AI Policy Researcher
© 2026 Kim Kyung-jin. All rights reserved.