Innovative AI in games

GoodAI is exploring breakthrough applications of AI in games. Firstly, after collaborating in the VeriDream project, we developed a solution that allows players to quickly build and animate walking robots. Secondly, we are developing a video game that uses large language models to power in-game interactions with the player.

GoodAI and Keen Software House are sister companies founded and lead by Marek Rosa. The most notable products of Keen Software House are video game Space Engineers with 4M+ copies sold and the fully volumetric VRAGE engine that powers it. GoodAI is an AI research and development company with a long-term goal of developing General Artificial Intelligence. In GoodAI, we have found that video games are a platform that allows for creating meaningful and achievable milestones on the path to this ambitious goal. At the same time, we know that AI is going to revolutionize game development and games. We present two use-cases that support this vision, which is further catalyzed by our proximity to Keen Software House.


Player-created robotic walkers in Space Engineers
The aim of the project, developed within the EU-funded VeriDream collaboration, is to allow players of Space Engineers to build and operate walking robots within the game. The game runs a physical simulation that calculates properties of mechanical motion in realistic detail, allows driving vehicles or flying spaceships, all in fully destructible voxel-based terrain. Building a legged robot in the game engine has been a possibility since its release. Making such a robot move while respecting the laws of physics has however always been a challenge reserved for the skilled. Thanks to this project, the players of the game will be able to quickly put together a walking robot of their own design that they will be able to control using keyboard and mouse, without having to implement the movement controller or understand the intricacies thereof. The content is in the process of integration into the game. Internally, the robotic walkers rely on Quality Diversity and Reinforcement Learning algorithms that are an outcome of collaboration with project partners from Sorbonne university and DLR in Germany.


AI Game: large language model-powered game interactions
The aim of this project is to develop a new game - code name AI Game - where freeform dialogs between player and game characters take a major role. In the game, players interact with game characters controlled by large-language models, where the language models simulate the behavior of people in both responses and actions, displaying complex personalities with goals, emotions, and memories. How does it work? Large language models are neural networks trained on internet-scale text corpora with one objective: to predict the next word. During gameplay, the game feeds the game character’s personality (described by designers in plain text), observations, recent events, and dialogues to a large language model, which then predicts what this person in this situation would do, think, and say (emulating their personality). This output is then translated to game actions. The current system is hybrid, relying on large language models, but also on traditional planner technology. In the future, we expect that the large language models will do planning too. This new approach to AI-driven characters leads to non-scripted behavior, complex and richer personalities, and more emergent gameplay.

www.goodai.com
www.veridream.eu