Facebook wants an AI capable of beating the hardest game
image: Facebook AI

Facebook wants an AI capable of beating the hardest game

Developing games is one of the fertile training methods for building high-end artificial intelligence systems of late. AI has become smart enough to beat humans easily in almost all games including StarCraft II and Dota 2 to Minecraft and Go.

Now Facebook wants to beat NetHack, one of the hardest games to take down. The company has asked the AI community to come for its aid for this mission. It will also help computers learn to simulate instances faster without many resources.

A game that doesn’t let you win

NetHack is a game that involves tactical gameplay. It was developed in the 1980s but is still pretty popular and hard to beat. The game doesn’t expect the players to win – it expects the player to die in the game. And, that happens very often!

What’s more frustrating is the fact that once the player perishes, the game resets back to square one. The only way to win in the game is a player’s ability to combine luck, quick problem-solving skills, and learning for mistakes players have made before them in NetHack.

As part of the NeurIPS 2021 NetHack Challenge, Facebook wants teams to build and train AI systems that can “develop agents that can reliably either beat the game or (in the more likely scenario) achieve as high a score as possible,” reads Wednesday FB blog post. The challenge starts this month until October 15th and the winners will be announced at NeurIPS in December.

New solution based methodology

Facebook believes this challenge will demonstrate the NetHack Learning Environment as a good reinforcement learning system and also pave the way for potential AI/ML solutions based on both neural and symbolic methodologies.

“The candidate agents will play a number of games, each with a randomly drawn character role and fantasy race,” the Wednesday post explained. “For a given set of evaluation episodes for an agent, the average number of episodes where the agent completes the game will be computed, along with the median in-game end-of-episode score. Entries will be ranked by average number of wins and, if tied, by median score.”

Disclaimer: The above article has been aggregated by a computer program and summarised by an Steamdaily specialist. You can read the original article at engadget
Close Menu