Data Machina #220
AI and Gaming. MindAgent Gaming. eastworld GenAI for Games. AI for Gaming Books. AI in a Box. Model Merging. GPT Pilot App Dev. Chain of Verification. BoolFormer.
AI and Gaming. I’m doing a consulting gig at a company which happens to have a human Games Master (GM.) The GM’s main task is to evangelise the application of gaming in life and business. Obviously, I’m very much inclined to propose that they replace this human GM with a proper, solid AI Games Agent. I suppose this would significantly increase randomness and serendipity in terms of “game-ability.” In the meantime, let me share some interesting bits on AI and Gaming:
MindAgent: Emerging Gaming. A new, interesting paper in which researchers propose a novel infrastructure to evaluate planning and coordination emergent capabilities for gaming interactions. The paper introduces a new gaming scenario and related benchmark that dispatch a multi-agent collaboration efficiency and supervise multiple agents playing the game simultaneously. Paper and demo: MindAgent: Emerging Gaming Interaction
Human-like adaptive AI agents in gaming environments. Back in July, researchers at Deepmind published a new paper in which they introduced a new Reinforcement Learning adaptive agent (AdA.) AdA can perform on-the-fly hypothesis-driven exploration and efficient exploitation of acquired knowledge as quickly as trained humans, while playing scenarios in a complex game environment called XLand. AdA was trained on a large Transformer model with meta-RL. The paper shows scaling laws in both model size and memory, and demonstrates that AdA improves its performance with zero-shot first-person prompting. Paper: Human-Timescale Adaptation in an Open-Ended Task Space.
A framework for generative AI agents in games. eastworld is an open-source, language-agnostic framework for adding generative gents to your video games. This framework has two goals 1) To elaborate storylines using an easy to use no-code dashboard and 2) To enable a variety of user-agent interactions out of the box beyond just chat, and expose it in a simple small API. Repo: eastworld: A Framework for Generative Agents in Games
Train AI agents to compete in games tournaments. Diambra is a platform where you can submit your AI agents to train them with cutting-edge Reinforcement Learning environments. And then make the agent compete with other agents in live video games tournaments. The platform has a section that comes with loads of books, papers, libraries, and courses on Reinforcement Learning. Checkout: Diambra, Platform for AI agents competing in video-tournaments.
Three AI Gaming startups. This is a hot space now, and investors are happy to throw billions at AI gaming startups. Checkout this blog post from legendary a16z VC: The Generative AI Revolution in Games.
Inworld. A leading-edge platform for generating AI NPCs (non-player characters) with configurable safety, knowledge, memory, narrative controls, multimodality, and more.
Ludo. An AI-powered games ideation, research, and creative platform for gaming studios. Develop high-quality game concept documents, elaborated game plots, character & assets suggestions with AI.
Rosebud. An AI-powered game development platform. Build and customise immersive worlds, AI NPCs, dynamic RPGs, and games assets with AI.
An AI platform for developing real-time game avatars. NVIDIA recently announced Avatar Cloud Engine (ACE.) A bit of a game changer (pun not intended.) This a suite of real-time AI tools for building and deploying intelligent game characters, interactive avatars, and digital humans at-scale. It includes speech AI, animation AI, generative AI, and AI avatar workflows.
Two books on AI and Gaming. Coincidently enough, a friend of a friend of a friend :-) just told me that if your click in the links below, you will be able to download two books on AI and Gaming.
Have a nice week.
10 Link-o-Troned
the ML Pythonista
Deep & Other Learning Bits
AI/ DL ResearchDocs
data v-i-s-i-o-n-s
MLOps Untangled
AI startups -> radar
ML Datasets & Stuff
CulturaX - A Huge Multilingual Dataset, 6.3 trillion tokens, 167 languages
LMSYS-Chat-1M: A Large-Scale Real-World LLM Conversation Dataset
Postscript, etc
Tips? Suggestions? Feedback? email Carlos
Curated by @ds_ldn in the middle of the night.