Google SIMA can replace people in many games.
There is practically no intermediary that is not related to artificial intelligence, because although we are talking about a relatively modern technology, providers are almost stepping on each other's toes with various text and image generators, while models of machines suitable for creating music and videos are also beginning to appear. .
Next up, Google recently introduced a new algorithm that can create games from just images, and now another AI technology is taking the console out of our hands.
Detected by Google DeepMind SIMA (Scalable and Steerable Multi-World Agent) It is the first generative artificial intelligence that follows spoken instructions in natural language in a wide range of virtual environments and 3D video games. In practice, this means that SIMA can perform actions in a range of games based on voice commands.
Introducing SIMA: the first general artificial intelligence agent that follows natural language instructions in a wide range of virtual environments and 3D video games. 🕹️
It can complete tasks similar to a human, and outperform an agent trained in only one environment. 🧵 https://t.co/qz3IxzUpto pic.twitter.com/02Q6AkW4uq
— Google DeepMind (@GoogleDeepMind) March 13, 2024
The Google team isn't the first to teach an AI program to play, just think of OpenAI and Nvidia's Minecraft algorithms, but the special feature of SIMA is that it's not just for a specific title, but can be used in any game. And that's not all, because according to DeepMind's claim, the technology performs better in a never-before-seen address than a traditional algorithm trained specifically for this program.
To this end, Google worked with 8 studios to help train the artificial intelligence in its games. Developers such as Hello Games (No Man's Sky) and Tuxedo Labs (Teardown) participated in the program, and thanks to extensive training, SIMA is able to perform more than 600 in-game actions, such as turning in a certain direction, interacting with objects, or to navigate menus.
Although the model cannot yet handle more complex commands, DeepMind continues to improve the solution, with the ultimate goal being for the AI to be able to play completely on its own. The practical meaning of this can be debated, but for example, players with limited mobility can benefit from an advanced AI agent that can be directed by voice commands.
