- The Regression Games Weekly Observation
- Posts
- New Minecraft Agents and Free RL Resources
New Minecraft Agents and Free RL Resources
Learn about Voyager, an agent that uses LLMs to play Minecraft.
Aaron Vontell
May 30, 2023
Welcome to our weekly newsletter and platform update! Every week we’ll bring you the latest AI + gaming news, resources, and updates from the Regression Games team.
Weekly Spotlight - Voyager
Last week, a new approach to building agents in Minecraft came out in the form of a project called Voyager, which iteratively writes code for Minecraft agents, rather than training an explicit model to play. We are excited to try this out on Regression Games, which supports writing code for agents in Minecraft. Check out the thread below for some awesome videos!
What if we set GPT-4 free in Minecraft? ⛏️
I’m excited to announce Voyager, the first lifelong learning agent that plays Minecraft purely in-context. Voyager continuously improves itself by writing, refining, committing, and retrieving *code* from a skill library.
GPT-4 unlocks… twitter.com/i/web/status/1…
— Jim Fan (@DrJimFan)
3:15 PM • May 26, 2023
Fable Simulation
The space of AIs in games for entertainment is heating up! Check out the Fable Simulation project, which aims to build a metaverse world where AI characters interact. “The Simulation is kind of like the world in the Ryan Reynolds movie Free Guy, where video game AI characters lived in their own world.”
You can learn more about them from their website or from this VentureBeat article.
Deep Dive - Stanford CS234: Reinforcement Learning
YouTube is an amazing resource when it comes to learning about reinforcement learning and agents playing games. A great playlist is the Stanford CS234 course, which has lectures covering topics such as Deep Q Learning, Imitation Learning, and other RL foundations.
Technical Paper of the Week
Considering the interesting components of the Voyager system mentioned in our highlight, we figured its paper would be a great recommendation! VOYAGER: An Open-Ended Embodied Agent with Large Language Models covers how their system works, such as generating skills and creating curriculums for exploration.
Members of our community discuss these papers on our Discord #reading-club channel. You can also find all past resources on our GitHub.
Looking for something new? Join the Regression Games team!
Have feedback about our newsletter? Let us know at [email protected]
You are receiving this email because you registered to our mailing list or are a player on our platform.