Spawning Autonomous Worlds with #traintolearn

home | .game | .club | .dao | .world

1025-1280-30.mp4

Summary

We demonstrate how simple coordination tools can help us accomplish large, complex tasks: finetuning generative AI, spreading memes at scale, and spawning new **autonomous worlds.

Our #traintolearn journey takes foundational concepts from #BUTMAKEITKAWAII and builds upon them with new crypto + AI tools fit for the coming age of decentralized worldbuilding.

Highlights

🌎 1,900+ original posts, 3,600+ replies, 5,500+ likes across Lens, Farcaster, Bluesky, Threads, and X

🌍 27 players generating with Stability AI

🌏 2 datasets curated using Jokerace + Hats Protocol + Collab.Land

🌏 2 generative AI finetuned with Mochi!

Purpose

The purpose of this journey was to define, test, and evaluate a model for an autonomous world machine.

The autonomous world machine does three things:

1. Incorporates advanced on-chain technologies for decentralized data curation + training
2. Enhances generative skills through open-source AI (Stable Diffusion)
3. Replicates itself across web3 social platforms (EVM)

Building from #butmakeitkawaii, we designed an autonomous world focused on collectively training generative AI. We utilize the generative AI to envision and share images of future worlds, aiding its growth and replication.

We tested the model in a four-week experiment, #traintolearn, from September to October 2023.