This article introduces a new learning artificial intelligence called AIRIS, which can learn independently in a complex 3D environment of Minecraft. Unlike AI that was tested in simple environments in the past, AIRIS started with "nothing", learning how to play games through game feedback, demonstrating its strong learning ability and adaptability. Its learning process involves obtaining input information from the environment (5x5x5 3D square grid and its own coordinates) and performing operations, and gradually building internal maps to learn to navigate and deal with various obstacles. AIRIS has a wide range of successful application prospects, such as automatic errors and stress testing for software, which will greatly improve software development efficiency.
A new type of learning artificial intelligence has emerged in Minecraft called AIRIS (Independent Intelligence Strengthened Inferred Symbols). The AI learns how to play games through practice, basically starting with "Nothing" in Minecraft, teaching it using only the game's feedback loop.
Early versions of AIRIS were tested in a simple 2D mesh world puzzle game environment. However, in order to test the system, developers need to test the system in a more complex and open 3D environment. Minecraft fits this description very well, it is a very popular game and has all the technical requirements needed to plug AI into it.
AIRIS works by getting two types of input from the environment and a list of actions it can perform. The first type of input is a 5x5x53D grid surrounding the block name of the proxy. This is how the agent "sees" the world. The second type of input is the current coordinates of the agent in the world. This gives us the option to specify the proxy where we want it to arrive.
AIRIS starts with “Free Roaming” mode and tries to explore the world around you. Build an internal map that records where it has been and can be viewed using the included visualization tool. It learns how to navigate the world, and when encountering obstacles like trees, mountains, caves, etc., it learns and adapts to them.
Successful use cases for AIRIS may include automatic errors and stress testing of the software. Assume that AIRIS can run throughout Fallout 4, for example, error reports can be created when interacting with NPCs or enemies. While quality assurance testers still need to check what AI records, it will speed up the tedious and frustrating process of development.
The emergence of AIRIS is the first step in the virtual world where artificial intelligence independently learns in a complex and all-round world. This should be exciting for AI enthusiasts as a whole.
The application of AIRIS in Minecraft demonstrates the huge potential of artificial intelligence to learn independently in complex environments, providing new directions and inspiration for the future development of artificial intelligence technology. It is worth looking forward to its application and breakthroughs in more fields. Its application in software testing also indicates a more efficient software development process in the future.