GDM Holding filed 2 patents this quarter, both in AI & Machine Learning.
The filings cover machine learning applications for gaming environments. One patent describes a companion agent that processes gameplay video streams to provide real-time assistance and visual overlays, while the other details a reinforcement learning matchmaking system that trains AI agents through dynamic opponent pairing for complex multi-agent simulations.
The 2 AI and machine learning patents explore distinct approaches to game intelligence. One application describes a companion agent that operates outside the game itself, watching the player's video stream and responding to natural language questions by overlaying hints and guidance directly onto the screen. The system works with any title because it relies on video observation rather than developer integration, combining proactive sampling, persistent session memory, and tool access within a unified pipeline. The second patent addresses multi-agent reinforcement learning through a matchmaking framework where each training agent maintains its own distribution of sparring partners, drawing from a diverse pool that includes fixed reference policies and agents with varying hyperparameters. This structured pairing prevents the collapse into dominant strategies that often occur with simple self-play, allowing the collective pool to improve over time through exposure to heterogeneous opponents.
All data sourced from USPTO patent filings. Google Patents may take several weeks to index recent publications. If a link is unavailable, search for the patent number at USPTO Patent Public Search.