Facebook is teaching an AI Minecraft

Credit: Mashable India

Researchers at Facebook are teaching an artificial intelligence to play Minecraft.

The idea is that the project will create an AI that is good at multitasking rather than individual tasks. The team at Facebook Research is using Minecraft to train AI so that it can interact with humans and perform a range of tasks.

So far, AI has been used to play Go and other games. However, an AI naturally cannot complete a job it isn’t trained for.

Whilst the world of Minecraft is expansive, the rules are particularly simple. This makes training an artificial intelligence to play the game straightforward, whilst there are many benefits to it.

Commands are reportedly going to become easier. As the AI experiences more gameplay, it will learn how to translate a user’s request.

“Instead of superhuman performance on a single difficult task, we are interested in competency across a large number of simpler tasks, specified (perhaps poorly) by humans,” says Facebook Research.

“The set of things a player could possibly do in the game is enormous; in the most naive sense, it is all possible ways of placing all the possible blocks into as big a world as fits in RAM.”

“We expect that the distribution of player requests of an assistant will be concentrated on a tiny fraction of what is actually possible in the game,” they add.

This is not Facebook’s first foray into AI assistants. In 2015, Facebook launched M, an AI assistant within the Messenger app. It was shut down after two and a half years, with one of the limitations being that it required a lot of human oversight to complete tasks.

Minecraft is the best-selling video game of all time. It has sold over 170 million copies and over 90 million users play every month.

Luke Conrad

Technology & Marketing Enthusiast

Steve tells us about Zumtobel products and services

Delia Salinas • 18th March 2022

We interviewed 19 of the most relevant companies: Fujitsu, Acronis, D2IQ, Enlogic, Platform SH, Tencent Cloud, Dragos, App Gate, Total Data Centre Solutions, Incepteo, Canonical, Netskope, Cloud Q, Microsoft, Vertiv, Ionos, Mitsubishi, Ionos, Dragos