Facebook will use Minecraft to give its AI a crash course in multi-tasking and listen to the directions of players
- Facebook is training an AI assistant with the Minecraft task
- Players can give the assistant commands such as & # 39; build a circle & # 39;
- The bot will then read the command and execute the player request
- It also learns from applications and develops new skills with the help of machine learning
Minecraft is more than just a game, according to Facebook, it is also an ideal training ground for advanced artificial intelligence.
The company says it will use Minecraft, a popular construction and adventure game in sandbox style, to train an AI assistant on one of the most important human skills: multi-tasking.
& # 39; Instead of superhuman performance on a single difficult task, we are interested in competence for a large number of simpler tasks, specified (perhaps poorly) by people, & # 39; say researchers in one proposal paper.
The bot can communicate with players and then perform tasks based on requests such as & # 39; come here & # 39; or & # 39; can you build a circle there? & # 39;
WHAT IS MINECRAFT?
Minecraft was created in 2009. At the start of the game, a player becomes a & # 39; virtually infinite game world & # 39; placed.
They can then walk around different terrains, including mountains, forests and caves.
Players can also fly up in the sky for a panoramic view of the landscape.
Players receive blocks and tools to build villages and towns.
As a player progresses, they can earn advanced tools and building blocks in different materials.
The game was originally made for the PC, but Xbox 360 and mobile versions are now available.
Players can now enter virtual reality with the launch of the game for the Oculus Rift.
For a person – even a child – these may be simple questions, but they can be monumental for the current AI.
For example, to tell an AI to build a circle, the technology needs to read the typed text instructions, know where to go, know what a circle is, where to place the blocks, how much to use, and much, much, much more .
That difficulty, say Facebook researchers, is part of the call.
& # 39; In such a & # 39; n environment, understanding the content of a task can be challenging & # 39 ;, is the proposal.
& # 39; In addition, a large number of tasks means that many tasks have only been seen a few times or even never, requiring sample efficiency and flexibility. & # 39;
Ideally, the AI learns from these requests and starts to develop newer options for storing in his repertoire.
According to them, Minecraft is an ideal environment for training bots for various reasons.
One of the strengths as a training ground is that the game has an extremely open design and offers a world of potential possibilities that the assistant can chew and spit out.
Another ideal feature, they say, is the fact that the AI is designed to be fun and as a result, players are actively encouraged to communicate with it.
& # 39; Because we work in a gaming environment, players can enjoy interacting with the assistants as they develop, providing a rich source for human-in-the-loop research & # 39 ;, they write.
Minecraft, although easy to play, can be very comprehensive. Photo above is a full-scale building made using the modular architecture of the games.
The seemingly unlimited possibilities in Minecraft make the game an ideal training ground for an AI assistant, researchers say.
Even if the AI Assistant fails to help players build an item, it's not hard to see how Facebook could use the loads of data from the bot's foibles to inform future creations.
Facebook has long relied on algorithms to make its platform more attractive, to remove content and to offer advertising suggestions to its users – a possibility that has helped drive its fundamental business people.
According to Facebook, an early version of the assistant can be downloaded here.
. (TagsToTranslate) Dailymail (t) sciencetech