Researchers release artificial intelligence into Minecraft game
Recently, in South Africa, researchers let two AIs have their way in the popular “sandbox” video game Minecraft. The objective? Observe their evolution in this environment and better understand in general the possible flaws of AI when they are not guided by humans.
A special AI model for Minecraft
As a reminder, Minecraft is a video game released in 2011. This open world integrates a crafting system focused on the exploitation and transformation of natural resources. The objective is to make constructions using small cubes . Minecraft surpassed the 200 million copies mark in mid-2020 for a community of more than 125 million monthly active players. It is thus the best-selling game in history.
For Steven James, a researcher at the University of the Witwatersrand (South Africa), Minecraft can also be a great playground for young AI . The intelligences involved here are indeed multitasking and learning is the key to their creation and evolution. This is therefore a question of machine learning, a process that allows AI to learn from its errors to gain performance when solving tasks without having been initially programmed to do so .
The South African researchers' AI model is none other than MinePlanner, created especially for testing in the game Minecraft . The results of the study were then pre-published on the arXiv platform on December 20, 2023.
Fairly low performance
The researchers created two AIs using MinePlanner. Each had to ensure a total of 45 constructions separated into three difficulty levels : easy, intermediate and difficult. For the AIs, the most complicated thing was to cope with the imposing mass of information contained in the game. They then had to ignore the data deemed useless for the realization of their constructions and therefore progress on their own.
The results were not really conclusive. Indeed, the first AI managed to complete fourteen easy constructions and three intermediate level ones. The second AI only managed to complete five easy and only one intermediate construction. According to the scientists, the second AI particularly suffered and exhausted its memory on most tasks before it could even carry them out.
The authors explain that there is still a technological gap to overcome the resolution of certain problems. Nevertheless, such work could help develop new approaches and possibly in even more complex areas. Thus, the future will tell us whether it will be possible to better understand AI learning using the MinePlanner tool.
Comments
Post a Comment