I don't know about you, but when I play video games, I like to actually play them. If you want to outsource your gameplay to others, you'll probably look at things like Let's Play or his Twitch stream. But Google is working on an AI model that can play video games for you as long as you tell it what you want it to do. It's called SIMA, which stands for Scalable Instructable Multiworld Agent, and we'll see if it works as advertised. , AI may take over your favorite hobby.
Google DeepMind, the company's artificial intelligence division, announced the new model in a blog post and a post on X (formerly Twitter). According to Google DeepMind, SIMA is the first generalist AI agent that can follow natural language instructions in a 3D environment. In other words, you can play video games based on your commands. When you say “turn left,” SIMS turns your character to the left.
Google DeepMind worked with eight video game studios to train SIMA. no man's sky hello games decomposition Tuxedo Institute. The development team wanted as many different types of games as possible to train his SIMA, as each new variable adds another skill to the model's capabilities. Google DeepMind also built a sandbox-like environment where SIMA needed to build structures to test its understanding of physics and object manipulation.
SIMA is successful because, at least in theory, it does not require any technical information about the video game itself, such as source code or API. It operates solely based on video game images and natural language commands. According to Google DeepMind, SIMA can perform more than 600 “basic skills” such as rotating in a certain direction, interacting with objects, and using game menus. However, Google DeepMind continues to develop more complex actions and commands that include multiple subtasks. It's one thing to tell an AI to climb the ladder in front of you, but it's another thing entirely to train it to accurately respond to “mining resources to build a shelter.” . The company says this is generally a limitation of large language models, where bots respond to simple commands but have difficulty performing intuitive actions independently.
Meanwhile, Google DeepMind touts the success of its multi-game training model, claiming that SIMA outperforms models trained on one specific game at a time. In fact, the company claims that SIMA can respond better to never-before-seen games than models trained on those games alone.
Although SIMA is not yet publicly available, we can imagine several potential use cases for this technology. I think this could be a great accessibility option in the future. For players who have trouble using traditional controllers, telling a bot how to control the player can be a game-changer. Of course, Google's end goal seems to go beyond this situation, and they want the AI to be able to play games on its own. This may be great for avoiding repetitive tasks like leveling up and earning money, but you might wonder, “Why play the game in the first place when you want a robot to do all the playing?” This also raises the question.
This is Google's second major foray into AI gaming. Last month, we learned that Google is working on a model that can also generate 2D platformers from natural language commands. Perhaps in the near future the company will roll out his Google Gaming. Just tell the AI which type of game you want to see and it will generate both games for you. and Play games in real time. How interesting.
