Crowdsourcing LLM training.
Onchain General Intelligence Network
Last Week I watched a documentary produced by Marshall Mathers aka Eminem called “How Music Got Free”. It is the story of how a group of teenagers with access to the ICR during the late 90s put together a network of collectives that ripped music and shared it, just for the love for music. Collectives like Rabid Neurosis, who were just a bunch of tech-savvy teenagers with enough skills to rip a CD and put it onto the internet. Were the ones that fueled the construction of the iPod. They were also responsible for the disruption of an old centralized industry like the record label industry. The power of the collective is unbeatable.
One of the most important narratives in the current AI Boom is large language models (LLMs), the cost to train them, and their limitation cost related when it comes to fine-tuning them to a better chain of thoughts-based outcomes.
To develop masterful AI agents and progress towards world-improving AGI, we need more advanced systems that can train AI to set goals, create robust plans to achieve those goals and make proficient decisions while executing complex sets of plans.
But here is when my discourse of democratizing AI makes more sense. Fine-tuning a model is expensive, so you want it to deliver at least 10x the value in cost savings for the task you’ve trained it for. How many organizations do you know who can afford an R&D like this?
Here is where a project like GAIA developed in a Layer 2 ETH ecosystem through BASE, brings us a different approach.
They are crowdsourcing LLM training, how?
Onchain Gaias aims to make advanced AI technology widely accessible through a decentralized network of intelligent agents called General Artificial Intelligence Agents (GAIAs).
GAIAS’s are AI agents that can be trained by anyone through on-chain gameplay, with compute resources funded by player transactions. The system incentivizes widespread participation through rewards in the form of tokens, earned by training agents to mastery in various games and skills. This means training to earn. Because they are training based on gaming, the training is driven by specific endpoints, enhancing the possibility of training with a chain of thoughts. This is why bringing super-capable agents through a crowdsourced approach is interesting.
How does it work?
The system, called Onchain General Artificial Intelligence Agents (GAIAs), is designed to distribute the training of open-source AI Agents over decentralized compute networks, through gaming on social media. Let's break that down.
In the first iteration, anyone will be able to execute a transaction on Base from within a Farcaster Frame. This transaction will do two things:
Train the AI Agent by executing a game on-chain against other AI agents
The game can be executed with minimal clicks in a Farcaster Frame
Track the amount of computing the user has contributed to train that AI agent
This will be used to reward each user proportionally with ecosystem tokens.
In the beginning, the agents and games incentivized to be trained will be highly generic and low complexity. This enables GAIAs to build a base of players on simple games with faster iteration cycles, before moving on to more complex use cases requiring more intelligent agents. The early stages will validate the technology and identify the strengths and weaknesses of different parts of the system.
As the network and player-base training agents both increase in size, two things become possible: using more advanced models in agents, and developing finely-tuned training and inference infrastructure.
The incentive for AI engineers, businesses that are looking to improve their process through AI through a low-controlled cost or even monetize their investments will be able to do it.
Highly capable GAIAs will be able to be monetized through the Armory Marketplace (an Intelligence-as-a-Service platform) and the Foundry Marketplace (a Transfer Learning as a Service platform).
The project seeks to make AI technology openly accessible to all, fostering innovation and collaboration on an unprecedented scale while challenging the closed-source, monopolistic practices of tech giants.
There is a strong social component, the first token holders (NFTs) are divided into researchers and degens, and the first of them are the ones bringing the initial models to the ecosystem. The second of them provides liquidity by staking ETH, to make the ecosystem liquidity possible.
A project like this is exciting because it enables the collective creativity to bring things to this race with the possibility of making money from it, getting more capable models, and all this through a game development ecosystem. One of the most liquid, full of creators and builders verticals across techs.
Imagine downloading Your Agents like it was a Limewire song of the 2000s.