Game AI Daily · Wednesday, October 28, 2026
GAMEAI HOT
Epic plans EULA change to allow AI training on Unreal Engine and MetaHuman outputs
Epic CEO Tim Sweeney said on X that Epic plans to revise its EULA to remove restrictions on using Unreal Engine and MetaHuman outputs for AI training.
Model releases & updates
DeepMind unveils next-generation world simulation model Genie 4
Google DeepMind has unveiled Genie 4, a next-generation world model that generates interactive simulated worlds in real time, with native audio and conversational interaction. Users can change the weather or camera angle, or add characters and objects, using natural language, and Gemini adjusts prompts to keep the scene coherent. It is currently available as a research preview to trusted testers. Google AI Ultra subscribers and early Gemini API partners will gain access in the coming weeks, though visual detail and object persistence may still drift over time.
Hyper3D Rodin's WorldGen coming soon with Agentic Mode
Hyper3D announced that WorldGen will soon add Agentic Mode, which generates a complete 3D scene mesh from a single image. The output includes lighting, materials, and separate object meshes, all of which can be edited and refined. The original post says it is aimed at uses such as game development, design, and robotics Sim2Real.
Engines & tools
NVIDIA RTX Spark powers neural rendering and AI characters in games
NVIDIA RTX Spark is a Windows PC platform for AI development, content creation, and gaming, designed for thin-and-light laptops and compact desktops. It combines a Blackwell RTX GPU with a Grace CPU, offers up to 128 GB of unified memory, and includes the full CUDA stack along with RTX technologies such as ray tracing and DLSS 5. The article is aimed at game developers and says it will cover related technologies, but the text is truncated, so the details are not shown.
Research
WorldPager: zero-stall tiered memory for long-horizon world models on a single consumer GPU
WorldPager stores the int4 KV memory of interactive world models across GPU, DRAM and NVMe tiers and prefetches it based on camera pose. On a single RTX 4090, it continuously generates 37.5 minutes of video without stalls, with host memory staying constant and a speed of about 9.5 frames per second. On a closed test set, int4 storage showed no measurable quality loss. On long-interval revisits, image quality dropped by about 1 MUSIQ point.
Text is the bottleneck: a small-scale replication of LLM game playing in Riichi Mahjong
A small-budget replication of LLM Riichi Mahjong gameplay research
How humans learn unfamiliar worlds: a think-aloud dataset on ARC-AGI-3
Researchers built a think-aloud dataset that records how humans explore and learn unfamiliar game mechanics. It contains 144 human trajectories from 64 participants across 25 public games, plus 250 LLM agent trajectories. The analysis shows that qualitative representation makes up 50.0% of successful human trajectories versus 14.3% of successful LLM trajectories, and humans more often explore without a specific hypothesis.
Industry
Rockstar staff to vote on union recognition in November, just as GTA VI launches
Rockstar Games workers to vote on unionization right as Grand Theft Auto VI launches
How-tos & opinion
How enemies and bosses decide what to do in video games: a primer
This Unity official primer covers three common ways game enemies decide what to do: separating behavior into states, behavior trees checked by priority, and scoring actions by situation. Using a castle guard, it explains four states — patrol, investigate, attack, and flee — and stresses that enemy perception and attack telegraphs should feel fair to players. The article also breaks boss fights into two design approaches: testing player skill and advancing in phases.
When AI becomes your gaming teammate: Artificial Agency CEO on agent-based gameplay
The Naavik podcast features Brian Tanner, co-founder and CEO of Artificial Agency, discussing how to build agent-driven gameplay and autonomous virtual players. The interview notes the company has cut inference costs to single-digit cents per player-hour and plans to integrate with Owlchemy Labs' Dimensional Double Shift, where three AI teammates would replace human players in a cooperative VR experience.
Shigeru Miyamoto on AI, productivity and creativity
Shigeru Miyamoto says AI makes execution easier, so creators must think harder about what counts as creation. He separates conception, which builds the overall structure, from execution, which fills in details and polishes the finished product. He warns that polished execution alone does not guarantee a work's value. Miyamoto believes AI may take part in nearly every stage of production, but human judgment about what is worth creating remains the most important factor.
Start with a feeling: How to define your first video game
Unity's official blog suggests starting a game by deciding what feeling you want players to take away, then letting that shape mechanics, art, length, and the other choices. It offers the sentence "My game makes players feel [feeling] by letting them be [character] and [core action]", and recommends breaking it into three or four design pillars to filter new ideas. It also notes that with AI tools, the more specific the description, the closer the output stays to the design direction, and introduces Unity Spark, Unity's browser-based AI game creation platform.