Sign up for the free weekly IndyTech newsletter delivered straight to your inbox
Sign up for the free IndyTech newsletter
Google's DeepMind lab has announced Sima, a new AI gaming companion that behaves like a human player and can play alongside gamers. This also raises concerns about the future of the online gaming community.
Sima stands for Scalable, Instructable, Multiworld Agent and is designed to perform video game tasks according to the gamer's instructions and play alongside the gamer.
This differs from the non-playable characters (NPCs) commonly found in video games, instead acting as fellow gamers who help the human player accomplish what they tell them to do.
Although currently in the research phase, DeepMind said AI will eventually learn to play any video game, including those set in a virtual open world.
“Sima is not trained to win matches; he is trained to run it and do what he is told to do,” the company added.
The AI acts like another human player and can influence the outcome of the game.
“Sima only requires images provided by a 3D environment and natural language instructions provided by the user,” DeepMind explained.
Google DeepMind launches Sima to train AI agents to learn gaming skills
(Google Deep Mind)
While such AI players can be cooperative companions in multiplayer games with story modes, Sima can also provide gamers with an unfair advantage.
For example, in a game that involves tedious, repetitive tasks to earn XP, the AI could be told to do all that work, freeing up the human gamer to develop a better strategy. there is.
Gamers took to social media to share their concerns about losing to undetectable AI agents when playing online games in the future.
“When these kinds of agents improve to real AGI levels, people will no longer be able to compete with them,” one user posted on Reddit.
“Anyone can use an agent because they will eventually be able to master, learn, and master any new game faster than anyone else, and then be able to do things far beyond our capabilities.” You will be able to win without being noticed,” the user said.
Google worked with several game developers, including Hello Games, Embracer, Tuxedo Labs, and Coffee Stain, to train and test Sima and incorporate AI into their games. like We'll use No Man's Sky, Goat Simulator 3, and Valheim to teach you the basics of playing these titles.
DeepMind noted that from training to play video games, the AI now has around 600 basic skills, such as turning left, climbing ladders, opening menus and using maps. .
“Learning to play even one video game is a technical feat for an AI system, but learning to follow instructions in a variety of game settings can unlock AI agents that are more useful in any environment. possible,” the company said.
Eventually, with more training, Shima could be instructed to perform more complex functions in the game, the AI Institute said.
Researchers hope to translate the functionality of advanced AI models like Sima into actions that are useful in the real world.
“We hope that Sima and other agent researchers can use video games as a sandbox to better understand how AI systems can help,” they said.
