AI & Machine Learning

Sony received 8 patents covering AI-driven gaming experiences across content generation, player interaction, and quality assurance. The first patent describes a system that converts gameplay telemetry into personalized post-game content like storybooks, podcasts, highlight reels, and memes by chaining two generative models together. Another application addresses automated quality assurance, using machine learning to detect performance issues during game execution, diagnose root causes, apply fixes autonomously, and commit changes to version control before escalating to human developers. A third patent outlines a system for converting player body movements into text descriptions that remote friends can attempt to mimic, while simultaneously mapping those motions to in-game characters when contextually appropriate. The fourth application enables AI-powered haptic feedback by feeding text descriptions of game events into machine learning models that generate tactile controller responses dynamically. Sony also filed a patent for an ML system that fuses camera, inertial, and microphone sensor data to interpret player intent for controller-free gaming. Another application uses machine learning to reverse-engineer which controller buttons were pressed by analyzing gameplay video alone, overlaying this information for viewers watching tutorials or streams. The seventh patent covers an in-car entertainment system that feeds real-time navigation data into a large language model to generate interactive stories tailored to specific routes, displaying visuals on transparent windows for passengers while keeping driver interaction audio-only. The final application describes an AI system that allows NPCs to understand player speech and execute in-game commands, with a feedback loop that converts action results back into text to keep NPC dialogue synchronized with actual game state rather than generating disconnected responses.

Adeia received 1 patent for a time-aware deep learning framework that fits gaming sessions into user-defined time windows while maintaining consistent difficulty throughout. The system embeds time constraints directly into the Monte Carlo tree search algorithm through modified confidence calculations and uses dual neural networks that optimize simultaneously for game outcome and session duration, avoiding the resource waste and experience degradation that comes from switching AI models mid-session or adjusting difficulty on the fly.

NetEase received 1 patent for an AI-driven offensive zone selection system in virtual sports games. Rather than boosting player attributes or relying on manual positioning, the system intelligently pairs virtual player objects with optimal attacking zones through a structured matching algorithm that considers both player selection and spatial positioning simultaneously.

Truist Bank received 1 patent for a player re-engagement system that combines in-game behavioral data with real-world financial transaction data from a partner banking application. This cross-domain approach trains machine learning models to predict and prevent player churn more accurately than systems relying solely on gameplay metrics like session duration and frequency.

Binary Kinetics received 1 patent for a stateless AI Game Master architecture that reduces per-session memory requirements by approximately 99% through transaction-based context reconstruction. Instead of replaying full conversation histories to maintain LLM context, the system reconstructs prompt-compatible semantic context from compact action logs using event sourcing principles, combined with provider-agnostic LLM orchestration and multi-provider failover for scalability.

Microsoft received 1 patent for a photorealistic avatar system using Gaussian splatting with prior model optimization. The approach integrates a statistical prior model trained on human appearance data to constrain and accelerate avatar reconstruction, reducing artifacts and enabling high-fidelity digital twins from sparse or low-quality input data for applications in VR, gaming, and video conferencing.

Playtika received 1 patent for a dynamic player segmentation system that personalizes gaming experiences in real-time based on behavioral data. The infrastructure decouples player segmentation from core game architecture, allowing operators to implement temporary or permanent experience modifications quickly without deep code changes, enabling granular behavior-driven segmentation and rapid deployment of differentiated experiences across large player populations.

Google received 1 patent for an AI coaching system that explains strategic reasoning behind optimal gameplay moves through interactive chain-of-thought dialogue. The system processes actual gameplay video and images as multimodal input, allowing players to interrogate the AI's reasoning in natural language conversations rather than simply receiving opaque action recommendations.

Meaning Machine received 1 patent for an AI system that dynamically updates video game world lore in real-time as NPCs improvise dialogue. The system uses semantic vector comparison to merge emergent, AI-generated lore back into the master context database, maintaining consistency across concurrent multiplayer sessions without exceeding LLM context window limits, moving beyond static query-only retrieval approaches.

EA received 1 patent for an AI-powered gameplay assistance system that uses geometric feature extraction from virtual environments to help players navigate obstacles in real-time. Rather than relying on player performance metrics or scripted hint systems, the approach determines when and how to assist players based on understanding the spatial challenges they face in the game world.