Google is training AI to play video games — and why this could be problematic

Computer-controlled video game characters are nothing new. Non-playable characters, or NPCs, have long been a staple of gaming.

However, Google’s artificial intelligence research laboratory DeepMind has Announce Working on something unique and novel in computer-controlled gaming: SIMA, which stands for Scalable Directable Multi-World Agents.

Unlike NPCs, Google DeepMind’s AI-controlled gaming companions are not programmed to act in a certain way or react to certain human-provided actions within a specific game. As DeepMind describes it, SIMA is a “universal AI agent for 3D virtual environments.”

SIMA is not an NPC

SIMA plays with gamers, just like human players, following verbal commands. As Google explains, “SIMA is not trained to win games.” It is prepared to work alongside human players and perform actions based on natural language prompts.

This element is crucial to SIMA’s uniqueness. Again, SIMA is not an NPC designed specifically for a game to provide challenges to players. SIMA acts as a player companion, trying to accomplish whatever the human player tells it to do.

“SIMA only requires images provided by the 3D environment and natural language instructions given by the user,” Google said. “With mouse and keyboard output, it can assess 600 skills covering areas such as navigation and object interaction, such as ‘turn left’ or ‘Cut down the tree’.”

According to Google DeepMind, SIMA was trained and tested on nine different video games of different game types, including no man’s sky, Goat Simulator 3, tear downand shaky life. Training does not require any form of game source code or API access.Google deep thinking explain It focuses on open-ended games rather than narrative games to facilitate SIMA’s training of more general gaming skills.

SIMA has some positives and some very obvious negatives. An AI gaming companion can be good when one wants to play multiplayer games or offline two-player story modes without the need for another person to choose a controller.

However, it is clear that the ideas that drive SIMA research can also be used for malicious purposes. Gamers can use SIMA to gain an unfair advantage in online gaming. Games where players have to perform monotonous tasks to gain resources or complete tasks to gain XP can obviously be manipulated by humans by simply telling an AI companion to do all that work for them.

But for now, SIMA isn’t ready for prime time. It’s still in the process of learning to play video games.

Source link

Leave a Reply

Your email address will not be published. Required fields are marked *