factors or scenarios when it is developed. In order to carry out its mission under any
circumstances, the military AI system must have the ability of self-learning and quickly
react to sudden and extreme situations on the battlefield.
Third is the problem of security. AI system is facing many threats in battlefield,
including physical damage and cyber attack. Comparing with the hard damage caused
by physical attack, the soft damage caused by cyber attack may be more serious and
more difficult to prevent. The US’s “Resilient Military Systems and Advanced Cyber
Threat” describes the situation of missile control system being hacked and may be
directed against own troops [3]. The software nature of AI system determines the
inevitability of cyber attack. This is because the intelligent function of AI system comes
from the software composed of various branch logics, variable tables and parameter
tables [4]. The more complex the task and the more diverse the environment, the higher
the complexity of the software, the wider the scope involved, the more vulnerabilities
and weaknesses, the greater the possibility of being broken.
Fourth is the problem of credibility. In civil AI products, as long as the algorithm is
effective, the others are not so important. How to allocate drivers and determine routes
for Uber? How does Google generate search results? How does Tesla’s autopilot work?
Many users are not particularly concerned [5]. In military applications, commanders
often need to know the reasons for AI system decision-making, because of the problems of life and death, war ethics and so on, before they can make up their minds.
But AI cannot explain its thoughts and actions to commanders, and commanders
cannot fully understand the decision-making process of AI system and cannot distinguish the logic behind a specific action of AI system, which often leads commanders to
distrust the decision-making and action made by intelligent system. Therefore, credibility is the key to the wide application of AI system in battlefield.
3 Solution Analysis
In view of the above problems in the application of AI in military, through the analysis
of the research status at home and aboard, especially the USA, solutions to the above
problems are suggested as follows.
3.1 Using War Games to Simulate, Generate and Collect Data to Solve
Data Source Problems
Game AI has developed rapidly in recent years. AI has defeated humans in most
games, including Atari, StarCraft, Dota and so on. The rapid development of intelligent
game technology has attracted the attention of the USA and taken a series of related
measures. At present, the US Marine Corps University has been experimenting with the
“Athena” deduction tool, a simulation and deduction platform dedicated to training,
education and testing future AI applications [6]. Using Athena’s war game, an air
assault task can be planned to control the interception position to support amphibious
landing. The game requires the user to complete the planning process when talking to
the voice assistant. The program will prompt defensive forms, definitions of different
tactical tasks and related historical cases. As the game progresses, an AI application
Key Problems and Solutions of the Application …
409
circumstances, the military AI system must have the ability of self-learning and quickly
react to sudden and extreme situations on the battlefield.
Third is the problem of security. AI system is facing many threats in battlefield,
including physical damage and cyber attack. Comparing with the hard damage caused
by physical attack, the soft damage caused by cyber attack may be more serious and
more difficult to prevent. The US’s “Resilient Military Systems and Advanced Cyber
Threat” describes the situation of missile control system being hacked and may be
directed against own troops [3]. The software nature of AI system determines the
inevitability of cyber attack. This is because the intelligent function of AI system comes
from the software composed of various branch logics, variable tables and parameter
tables [4]. The more complex the task and the more diverse the environment, the higher
the complexity of the software, the wider the scope involved, the more vulnerabilities
and weaknesses, the greater the possibility of being broken.
Fourth is the problem of credibility. In civil AI products, as long as the algorithm is
effective, the others are not so important. How to allocate drivers and determine routes
for Uber? How does Google generate search results? How does Tesla’s autopilot work?
Many users are not particularly concerned [5]. In military applications, commanders
often need to know the reasons for AI system decision-making, because of the problems of life and death, war ethics and so on, before they can make up their minds.
But AI cannot explain its thoughts and actions to commanders, and commanders
cannot fully understand the decision-making process of AI system and cannot distinguish the logic behind a specific action of AI system, which often leads commanders to
distrust the decision-making and action made by intelligent system. Therefore, credibility is the key to the wide application of AI system in battlefield.
3 Solution Analysis
In view of the above problems in the application of AI in military, through the analysis
of the research status at home and aboard, especially the USA, solutions to the above
problems are suggested as follows.
3.1 Using War Games to Simulate, Generate and Collect Data to Solve
Data Source Problems
Game AI has developed rapidly in recent years. AI has defeated humans in most
games, including Atari, StarCraft, Dota and so on. The rapid development of intelligent
game technology has attracted the attention of the USA and taken a series of related
measures. At present, the US Marine Corps University has been experimenting with the
“Athena” deduction tool, a simulation and deduction platform dedicated to training,
education and testing future AI applications [6]. Using Athena’s war game, an air
assault task can be planned to control the interception position to support amphibious
landing. The game requires the user to complete the planning process when talking to
the voice assistant. The program will prompt defensive forms, definitions of different
tactical tasks and related historical cases. As the game progresses, an AI application
Key Problems and Solutions of the Application …
409