..Stories About Industrializing In-Game Machine Learning Solutions
summary
Ubisoft builds a modular Machine Learning Framework that moves inference into the engine to stop Python-over-sockets lag and let prototypes ship as game code. Core pieces are in-game inference, messaging and debug views, and Python trainers for offline or online workflows; inference backends include ONNX Runtime plus LibTorch and a proprietary engine. Demos: RL vehicle driving in Watch Dogs, navigation using jump pads in HyperScape, Forerunner combat bots and imitation aiming; online RL also runs auto-tests. Integration lives in Snowdrop as of Nov 2023.
AI-written summary · about
chapters (7)
- 0:00Industrializing in-game ML
- 3:55From Python to in-game inference
- 6:06ML framework for production
- 11:06Industrialization in Ubisoft
- 15:24Cross-team ML adoption
- 20:22Modularity drives ML innovation
- 24:36Q&A on AI difficulty
related talks
48mMachine Takeover - Improving Your Game With Machine Learning & AI
29mThe Present & Future of AI and Games
52mThe Journey of AI in Games: A Technical and Philosophical overview
41mIntroduction to Deep Learning Models in Unreal
20mStudio Speaks: The Reality of AI Adoption in Games
01h 01mIntroduction to Practical AI
32mMachine Learning for Game Developers
30mBuilding, designing, teaching and training simulation environments for Machine Learning