Initial thoughts on Mozilla’s public API paper are that they are a new voice that is necessary and welcome in the world of AI. They correctly point out the “ladder pull” of private companies hiding their research, and also the need for a focus on small language models (SLM).
What are key takeaways?
- Smaller, efficient models are better for a variety of reasons
- SLM require less compute which reduces carbon footprint of AI development
- Open source (actually full open source) enables a broader access to AI technology globally
- Public funding could shift towards customizable and energy efficient AI
- Privacy, a human right, is enhanced with local deployments of SLM.
- Truly open, non-commercial SLM are a great era to invest time and energy into as they may be both the future of AI jobs, but also help humans at scale vs hurting them at scale.