Just as the check and mate in chess represent the ability to create a diamond tool in the video game Minecraft, which is one of the game’s highest level challenges, the same can be said about the development of artificial intelligence.
Furthermore, AI is developing the function of memory that is quite impressive in the computer game that is very popular.
AI applications have been highly advanced to play Minecraft without human intervention with big investments in all kinds of approaches. To illustrate, OpenAI, the developer of ChatGPT, has spent a large amount of money to employ human players of the game to be recorded so that it can be used to create AI that plays the game by copying the movements of people.
A research group headed by Zihao Wang of Peking University in Beijing in February presented what the group considers to be “the first multitasking agent which can deal with 70 plus Minecraft tasks.
Nevertheless, the modern technology is making huge progress. A team from Nvidia announced last week that it had accomplished the first “lifelong learning agent” that was made to adapt its strategy to the game through experiments with different techniques and then stored the results of its achievements in a library of techniques.
The technology, with other automatic systems, is complemented to achieve the first results in Minecraft faster.
The project, Voyager, is described in a paper that was posted on the arXiv preprint server by Guanzhi Wang of Nvidia and his colleagues at UT Austin, Stanford and Arizona State University. A consultant for the group is Nvidia’s senior director of AI research, Anima Anandkumar. (The article and the additional stuff are also on the companion site which is also posted by Nvidia.)