The battle for openness in AI

Severine Faurax -

A fundamental debate is shaping the future of AI: the battle between open source and closed source AI.

🎯What’s at stake?

The core of the debate lies in how AI technologies should be developed, shared, and controlled. Open-source AI advocates argue for transparency, collaboration, and widespread access, while proponents of closed systems emphasize security, control, and commercial viability. The outcome will influence innovation, economic dynamics, and ethical standards in AI development.

🔓🔒 Open Source vs. Proprietary

  • Open Source AI refers to models whose code, architecture, and sometimes training data are publicly available. This transparency allows developers worldwide to study, modify, and improve these models, democratizing AI and making it accessible to a broader community.

  • Proprietary AI Systems keep their code, architecture, and data private, restricting access to APIs or licensed use. This maintains control over how the technology is used and ensures competitive advantage.

🌐 Who's who in Open Source AI

  • Mark Zuckerberg is a vocal advocate for open-source AI. Meta has released models like Llama-3 as open source to foster innovation and collaboration. “I don't want AI to be in the control of a few companies. There's a risk that a handful of companies could run these closed models and control the APIs, dictating what developers can build,” says Zuckerberg.

  • Mistral CEO emphasizes the empowerment of developers through open-source models, fostering innovation, transparency, and competition. “We want to allow developers to deploy and modify our technology as they wish, enabling the freedom to innovate outside the constraints of major American cloud providers”.

  • Hugging Face Founders support open-source AI by providing a platform for sharing and building on pre-trained models, promoting robust advancements through a diverse community.

  • Dario Gil (IBM) advocates for open-source AI, emphasizing transparency and trust. IBM’s Granite series of open-source models highlight the benefits of collective innovation. “The future of AI is open,” says Gil.

  • Microsoft’s Phi-3 initiative under the SLM framework is a notable example of their commitment to open-source AI, aiming to stimulate innovation and collaboration.

🔐 The closed system perspective

  • OpenAI has shifted towards a closed system for its most advanced models, like GPT-4, due to concerns over safety, security, and misuse. "I think open source is good and important, but I also believe that precaution is really good,” says Sam Altman.

  • Google favors a closed approach for its advanced AI models, such as Gemini, to ensure security, reliability, and competitive edge, maintaining control over AI development and deployment.

  • Anthropic focuses on creating safe and aligned AI systems with its Claude model. By maintaining a closed approach, Anthropic aims to prevent misuse and ensure ethical deployment of AI.

🚀 The road ahead

The future of AI might not be a stark choice between open and closed. A hybrid approach could leverage the benefits of both worlds: open source for faster development and proprietary for highly sensitive tasks.

As AI continues to evolve, regulations might also be necessary to ensure ethical development.

What do you think?

📢 Want to experience how AI can transform your website?

Enter your domain on https://kayros.ai and test your AI chat assistant today!