The author also discussed the growing open source ecosystem in China, with many popular AI repositories on GitHub targeting Chinese audiences. The article concluded with the author's personal favorite ideas being developed by the community, including batch inference optimization and model merging. The author found the analysis helpful in understanding the AI ecosystem and encouraged readers to suggest any missing repositories.
Key takeaways:
- The author conducted an analysis of the open source AI ecosystem, focusing on foundation models, and found 896 repositories, 845 of which were software repositories.
- The AI stack consists of four layers: infrastructure, model development, application development, and applications. The most growth in 2023 was seen in the applications and application development layers.
- Open source software follows a long tail distribution, with a handful of accounts controlling a large portion of the repositories. The rise of one-person companies is notable, with individuals often gaining more stars than organizations.
- China's open source ecosystem is growing, with many popular AI repositories on GitHub targeting Chinese audiences. Six of the top 20 accounts on GitHub originated in China.