https://opensource.org/deepdive/drafts/the-open-source-ai-definition-draft-v-0-0-8
Open Source has demonstrated that massive benefits accrue to everyone when you remove the barriers to learning, using, sharing and improving software systems. These benefits are the result of using licenses that adhere to the Open Source Definition. The benefits can be summarized as autonomy, transparency, frictionless reuse, and collaborative improvement.
Everyone needs these benefits in AI. We need essential freedoms to enable users to build and deploy AI systems that are reliable and transparent.
David Wiley notes that, at the time of writing, only three models qualify as open source under the current definition:
- Mixtral (a Mixture of Experts model): https://arxiv.org/abs/2401.04088
- IBM Granite (Granite models are code LLMs aimed at supporting software development): https://arxiv.org/abs/2405.04324.
- OLMo: https://arxiv.org/abs/2402.00838. I’ve briefly mentioned OLMo on this blog before.
None of the models above are frontier models in the same capability class as GPT-4, Gemini 1.5, or Claude 3. However, they all exceed GPT-3.5-level capability at significantly lower cost. Open source AI models will probably never be at the cutting edge of LLM development but I have no doubt they will be the models running on most devices.
If the influence of open source software on the broader technology ecosystem is any indicator, open source AI will have a huge influence on how the world is shaped in the decades to come.
David Wiley. Toward a Definition of Open Source AI.