When the Commons Disappears

2025-05-26 · Source: Intentional Arrangement · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Government & Public Sector, Research Methodology & Innovation · Depth: Intermediate, extended

Summary

The United States is falling behind China in the AI race due to a fundamental difference in knowledge-sharing cultures, with Chinese labs embracing open-source practices and extensive disclosure while US labs increasingly adopt closed, proprietary approaches. For instance, DeepSeek's API charges significantly less than Claude Opus 4.6, and its DeepSeek-R1 model underwent independent peer review and disclosed its reinforcement learning costs of $294,000 in a September 2025 *Nature* paper. By February 2026, Chinese models constituted 41% of Hugging Face downloads, surpassing the US at 36.5%. This shift is mirrored by OpenAI's move from open disclosure in GPT-2 and GPT-3 to redacting architectural details in GPT-4 and subsequent models, alongside internal turmoil and departures of key safety personnel. Concurrently, US federal knowledge infrastructure, including IMLS, NIH, and NSF, faced significant defunding and dismantling, impacting public data availability and research. This trend of knowledge hoarding in the US contrasts sharply with historical periods of enlightenment driven by open knowledge circulation, creating a competitive disadvantage as AI systems rely heavily on shared human knowledge.

Key takeaway

For AI Engineers and Machine Learning Engineers developing frontier models, your approach to knowledge sharing directly impacts long-term innovation and competitive standing. Embrace open-source practices, detailed technical reports, and peer review for your models and methodologies, as this fosters community contributions and accelerates progress, contrasting with the diminishing returns of proprietary, closed-source development and the risk of data scarcity.

Key insights

Open knowledge sharing fosters innovation and competitive advantage in AI, while proprietary approaches lead to stagnation and decline.

Principles

• Knowledge compounds when shared.
• Secrecy is the antithesis of science.
• Open communication drives innovation.

Method

Chinese AI labs consistently publish full architectural details, training methodologies, and cost disclosures, often under open licenses, facilitating reproducibility and community-driven improvements, exemplified by DeepSeek-V2's Multi-head Latent Attention and DeepSeek-V3's FP8 mixed-precision training documentation.

In practice

• Publish model architectures and training methods.
• Release weights under open licenses like MIT.
• Engage in peer review for AI models.

Topics

• AI Race
• Knowledge Commons
• Open-Source AI
• Proprietary AI Models
• Federal Data Infrastructure

Best for: AI Engineer, Machine Learning Engineer, NLP Engineer, AI Scientist, Policy Maker, Director of AI/ML

Related on AIssential

• Can open-source beat OpenAI? — Open-source AI models foster global collaboration and efficiency, challenging closed-source dominance through accessible innovation and diverse monetization.
• OpenAI Closes Biggest AI Deal Ever — Current AI research prioritizes exploitation over exploration, hindering fundamental breakthroughs.
• The AI Revolution Will Not Be Monopolized: How open-source beats economies of scale, even for LLMs — Open-source development is posited as a critical force to prevent monopolization of the AI revolution, particularly for LLMs.
• The Future Belongs to AI Products, Not AI Wrappers — Sustainable AI competitive advantage requires building proprietary products, not just thin wrappers over foundation models.
• Open Source Ecosystems — Open protocols are vulnerable to "complement capture" by private actors, shifting competitive advantage to developer experience.
• AI News: Everyone's Mad At Anthropic Now — Open-source AI models are rapidly closing the performance gap with proprietary models while offering significant cost and privacy advantages.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Intentional Arrangement.