Google is developing an artificial intelligence system that can play video games and take commands from players like a human, and could have an impact on the real world in the future.
“This study was not about achieving a high score in a game,” the SIMA research team wrote. Posted by Google DeepMind Early this month. “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. It may be possible to cancel it.”
SIMA, which stands for Scalable Instructable Multiworld Agent, is different from the typical computer players that are built into certain games. Rather, AI agents play and learn just like humans, through image recognition and native language commands, and manipulate keyboard and mouse output.
What is artificial intelligence (AI)?
According to DeepMind’s post, “SIMA only requires images provided by the 3D environment and natural language instructions provided by the user.”
For now, the AI agent is just a research project. It is intended to act as a companion to the human player who can perform tasks.
“SIMA is not trained to win the game, it is trained to run the game and do what it is told,” Google researchers wrote.
Google collaborated with eight companies game developerHelp with SIMA training and testing, including Hello Games and Embracer. As part of their goal to make the program task-oriented, the researchers primarily trained his SIMA in open-play environments such as No Man’s Sky and Goat Simulator 3, bizarre games in which players control a havoc-causing goat. Did.
Now that artificial intelligence can clone voices, video game actors must consider their options
“This research demonstrates for the first time that agents can understand a wide range of game worlds and perform tasks within them, following natural language instructions, just as humans can,” the Google team wrote.
SIMA has learned about 600 basic skills so far, such as turning left and climbing a ladder. Ultimately, Google researchers hope to allow AI agents to adapt without having to play games they’ve never played or special training.
Nvidia develops real-world robots with artificial intelligence capabilities. (Justin Sullivan/Getty Images)
But ultimately, Google hopes SIMA’s research will build toward “more general AI systems and agents that can understand and safely perform a wide range of tasks in ways that help people online and in the real world.” , the researchers wrote. In a blog post.
CLICK HERE TO GET THE FOX NEWS APP
“Our study shows how the functionality of advanced AI models can be translated into useful real-world actions through language interfaces,” the researchers added.
However, SIMA still has a long way to go before it can even perform basic video game play, with No Man’s Sky, for example, only succeeding in a third of its tasks.
SIMA researchers say, “We hope that SIMA and other agent studies can use video games as a sandbox to better understand how AI systems can help.”
