AI Gaming in Screeps and Kaggle

Learn about AI gaming experiences from Screeps and Kaggle, as well as the power of LLMs in games.

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.

Join our community on Discord and start playing!

Regression Games is hiring! Join our team to work on exciting projects at the intersection of AI and gaming. We have great benefits, competitive pay, and an amazing team. 👉 https://regression.gg/careers

 

Weekly Spotlight - Screeps Season 5 Begins!

Screeps is a resource-gathering AI game competition that has been running for many years. They just announced that Season 5 kicked off last week - if you’d like to try it out, you can play it online or on Steam. An interesting twist on the AI gaming genre with Screeps is that your bots run 24/7 in an open environment with many other players.

Deep Dive - Kaggle Game AI and RL Course

If you’re looking for a quick way to become familiar with game AI, check out this course from Kaggle - Intro to Game AI and Reinforcement Learning. This course is a great starting point since it covers the basics of tools like OpenAI Gym in a digestible and fast manner.

Fun Video - AI Learns to Walk using Deep RL

I saw this great video from the AI Warehouse channel where a bot is trained to walk through a series of challenges. What’s really awesome about this video is the usage of progressively harder obstacles to encourage the bot to begin walking like a human, versus flailing around like you may see in other RL videos.

Technical Paper of the Week

This newsletter often covers topics in both RL and LLMs. This week’s paper looks at comparing the two; more specifically, how LLMs can sometimes outperform RL algorithms in playing games. The “SPRING: GPT-4 Out-performs RL Algorithms by Studying Papers and Reasoning” paper shows that in some cases, an LLM reading the paper and description of the game can sometimes outperform an RL algorithm trained to play that game. With the Voyager paper from a few weeks ago, it’s very interesting to see how powerful these LLMs can be if prompted correctly.

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.