𝗪𝗵𝘆 "𝗼𝗽𝗲𝗻 𝘀𝗼𝘂𝗿𝗰𝗲" 𝗶𝗻 𝗔𝗜 𝗶𝘀 𝘀𝗼 𝗶𝗺𝗽𝗼𝗿𝘁𝗮𝗻𝘁?

👉 As Artificial Intelligence becomes more pervasive, it is essential to diversify beyond closed proprietary AI systems for ensuring safety and security, unlocking the technology’s benefits for all, and promoting diversity of visions and representation.

With Mozilla & Columbia Institute of Global Politics, we brought together 40 leading AI scholars and practitioners in New York to create a framework for understanding and advancing openness in AI.

I’m pleased to share our collaborative paper: https://arxiv.org/pdf/2405.15802

𝗡𝘂𝗺𝗲𝗿𝗼𝘂𝘀 𝗸𝗲𝘆 𝗶𝗻𝘀𝗶𝗴𝗵𝘁𝘀 𝗲𝗺𝗲𝗿𝗴𝗲, 𝗶𝗻𝗰𝗹𝘂𝗱𝗶𝗻𝗴:

1️⃣ AI models are not just code; they are trained on massive datasets, deployed on intricate computing infrastructure, and accessed through diverse interfaces. Thus defining openness in AI is even more complex than for software. 

2️⃣ There will be different “forms” of openness in AI. Beyond “open washing” and “open bashing”, this work seeks to highlight the diverse benefits of openness, such as enhancing reproducible research, fostering competition dynamism, and catalyzing unexpected innovations.

3️⃣ The paper introduces a descriptive framework that lays out the key tradeoffs and nuances for leveraging openness to promote transparency, accountability and safety in generative AI systems.

4️⃣ As AI regulations emerge, policymakers need enhanced frameworks to navigate the evolving landscape effectively. Civil society, independent researchers, and regulators need tools to assess the risks of large-scale AI models and systems. Entrepreneurs, builders, and organizations seek guidelines for developing and releasing their models responsibly. 

5️⃣ Safety cannot be thoroughly addressed by merely scrutinizing AI models. The context and deployment environment are also essential to ensuring AI is safe, making it vital to differentiate between open models and open AI systems. The latter can incorporate additional safeguards and introduce additional risks that may not be present in AI models alone.

🙌 This has been an incredible effort powered by a brilliant group of interdisciplinary authors (full list in comment) united by the belief in the importance of openness. 

👀The upcoming international AI Summit in Paris, next February will emphasize the critical nature of this topic.