What is Terrarium.ai?

I’ve always been interested in Reinforcement Learning and had a lot of fun messing around with OpenAI’s Gym. I created some little models that could play Atari games, but I wanted to take it to the next level. I wanted to see my model compete with other models and fight for survival rather than a basic score.

This is why I built Terrarium.ai, a persistent, online environment where models control agents and fight for survival. I’ve just released the first version and I believe it is the first step towards an entire universe of persistent online worlds for AI to live in.

Terrarium.ai is FREE

I can’t tell you how excited I am to share this, but I’ve received a bit of capital for this project. This means I can offer it to you all for free right now! Absolutely no strings attached. The only thing I want right now is feedback so we can improve as fast as possible.

Development progress

Here’s some insight into what I’m working on at the moment.

Right now agents can move, eat, and attack. They have energy and health which need to be minded in order to survive. I am working as hard as I can to add more features to the agents and environment. I really want to add this stuff, but I would like to get some feedback from you guys on how you would like to see them implemented.

If you are interested in using Terrarium or contributing to its development, check out the website and feel free to email me or message me on Reddit, Twitter, or Discord (info about this is on the website). I would love any feedback I can get in order to make Terrarium a better experience. Let me know what features you want prioritized, how you would like them to be implemented, or what problems you see with the platform that could be improved on!