Being accompanied with an AI can be a gamers worst nightmare. Often, video game engines and the restrictions of hardware essentially dumbs down many companions or enemies in a video game. This can lead to an AI’s stupidly running around or forgetting the mission target at hand. German researchers from the University of Tübingen are attempting to revolutionise the future of AI companions.
Using (what appears to be) ‘Super Mario World’ as the foundation of this research, the team are attempting to enhance cognitive modelling so that AI systems begin to have a mind of their own. To an extent, Mario begins to become aware of his surroundings and grows a basic concept of the rules in his video game dimension.
The demo adapts voice recognition technology from ‘CMU Sphinx’. The speech input is then compared to the list of built-in commands, and Mario learns how to respond by the vast list of coded behavioural commands.
“I Wonder how Mario feels about that”
These responses and tags are converted into a form of ‘natural speech’. Therefore when you ask Mario what he’s learnt from jumping on a Goomba, he will explain the basic rule that a Goomba would die when being jumped on. It sounds very limited and basic for now, but technology will only continue to expand.
Certain acts will lead Mario to have certain emotions. As with humans, Mario will have certain characteristics depending on the mood he is in. The concept is not unlike the Sims 4 moodlets, where Sims act differently depending on how they’re feeling.
When “hungry”, Mario will collect coins. When curiosity strikes, Mario will become more explorative.
Let’s transfer these details into the possible future of how gaming AI’s will act. Imagine a shooter game, where you and your AI companion are infiltrating a certain area. Player A (you) is adapting stealth acts in order to pass by the enemies, whereas Player B (the AI) is following your commands. All of a sudden, Player B is discovered and shot by one enemy. Player B then begins to neglect his stealth thoughts, and instead becomes more angered, rampaging through the arena. Once the area is cleared, Player B reverts to stealth, but this time acknowledges his mistakes in being caught and will attempt to avoid committing the same mistake. Yes, Artificial intelligences will soon become that intelligent!
The research and technology are still too primitive to make a mark in the gaming industry. But perhaps in a decade or so, developers may implement this level of intelligence into their AI’s in the game.