Teammate AI: Google’s SIMA learns to play alongside you

Google Logo
Google Logo

Imagine you are playing a game and your game partner doesn’t bother with winning but simply chooses to adapt to you and follow whatever you tell him to do. By employing the latter, Google DeepMind’s artificial intelligence development for the future was envisioned: SIMA. While we can easily teach an AI to win any game, Sima pays attention to the small details that make a complex game exciting, such as open-world games that have no fixed ending.

Rather than replace AI, our SIMA model seeks to be on a common platform where the solution can be used across a team in a way that can be easily integrated into an existing party. It takes your voice commands, interprets the 3D environment and recognizes objects while it’s in the game. “Real winning is not the main goal,” says researcher Tim Harley; “Once you learn the game and lead this way, that’s when the real success story will reveal itself.”

Such collaborations with game developers like Hello Games and Coffee Stain Studios helped Google pave the way for SIMA’s training process. Game Sima learns universal methods of gameplay by sampling a variety of games. It is then able to simulate but never completely emulate the human element in gameplay because humans do not behave exactly as they do in reality. Thus, the goat simulator, in which all the principles of chaos are valid. Symbolic interactionists like SIMA require this highly “unstructured” style to conduct their research on spontaneous behaviors.

Here is the simple training method: A representation of a world where agents are self-trained for their desired reward, this is Google’s virtual environment, and this leads to a technique for unsupervised learning of object manipulation to create sculptures. Is. Then, the scientists made in-game recordings of one participant keeping a goal while the other was issuing instructions. This resource is one of the strongest as it lists the actions taken by the user and also explains the reasons. SIMA also takes into account the ability to know everything that happens later in the game, it is **well-informed!**

Similarly, SIMA is now able to take advantage of basic features such as heading, ladder climbing, and navigation. long-term goals? For example, resource gathering and base building are both relatively more sophisticated. Attempts to remove SIMA from the gender gap, of course, do not mean human replacement – that is another player entirely, which affects the outcome.

Although voice commands are a possibility, Sima’s current purpose is to control the game and uncover new scenarios. SIMA is now a concept and prototype, but Google is confident of this project that one day it will become a powerful artificial intelligence partner that will guide you to victory and your victory.

Source