a New Open Source License: Protecting Code from AI Exploitation
In the rapidly evolving landscape of artificial intelligence, we've seen open-source source code used to train powerful machine learning models. While open-source software (OSS) has long thrived on principles of transparency and collaborative growth, there's growing concern about their exploitation by closed-source proprietary AI systems.
There's an urgent need for a new kind of open source license. A license specifically tailored to the age of AI. This license would directly address the issue of code being silently harvested for large language models and machine learning algorithms without giving anything back to the community.
The solution is a license that explicitly discourages unauthorized usage of open-source code for machine learning training unless clear conditions are met:
1. A Transparency Requirement: Entities that crawl, copy, or utilize open source code repositories for ML training must publish the resulting machine learning model, ensuring it's accessible to the public.
2. Reciprocal Openness: Those entities must also share the full source code used to train and deploy their models, reinforcing a cycle of openness and contribution.
3. Supports Open-Source This new license doesn't limit or change an existing license's spirit. The idea is just to add limits to any existing OSS license for machine learning.
I've made a start at such a license. Feel free to use it in whichever way you like. Here is version 0.1 of the Rock Machine Learning License.