Introduction

On June 22, 2023, the US Congress heard testimony from Hugging Face CEO, Clément Delangue. It was a clear milestone for the new era of Artificial Intelligence (AI) innovation that we're currently living. Clément Delangue's insights underscored the importance and imperative nature of Open-Science & Open-Source in AI.

In this article, we explain the key takeaways from this testimony, shedding light on the need to democratize AI for universal benefit, and the challenges and solutions that lie ahead.

🎬 Full video - Artificial Intelligence: Advancing innovation towards the National interest

Hugging Face's CEO testimony starts at 26:03.

Alternatively, you can read the official transcript here.

👐 Open-Source AI: Democratizing technology

Open-source AI, as Clément Delangue points out, is a requirement to democratizing AI. Thousands of open-source AI models exist. They span domains like finance, biology, and more. Making these models accessible boosts innovation for hundreds of thousand of non-profit organizations, startups and businesses.

However, to leverage AI's transformational potential, akin to that of the Internet, democratization must be widespread. The conduit to achieving this is open-science & open-source. The open approach allows free and easy access to AI models, enabling anyone with the skill and initiative to build upon them. It fosters a culture of knowledge-sharing and collaboration, thereby breaking down technological barriers and promoting a more level playing field.

Here's a quote from Clément Delangue that I particularly resonate with:

Open-science and open-source foster innovation and fair competition between all thanks to ethical openness. It creates a safer path for development of the technology by giving civil society, nonprofits, academia and policy makers the capabilities they need to counterbalance the power of big private companies.

📈 Open-Source and Open-Science's role in future innovation

Clément Delangue emphasizes the key role of open-source in the history of AI evolution. Today's AI progress owes a lot to open-source, open-science. Papers like "Attention is All You Need" on transformers and the latest Diffusion paper exemplify this.

PyTorch, Scikit-learn, Tensorflow, Hugging Face - all open-source products. They show open-source as an enabler of innovation. As we look forward, open-source remains key to distribute these gains. It's not just a technological tool. Open-source is a socio-economic catalyst, promoting fair competition and safe, ethical AI development.

Here's a quote from Clément Delangue that I particularly resonate with:

Open-science and open-source prevent black box systems, make companies more accountable and help solving today’s challenges like mitigating biases, reducing misinformation, promoting copyrights, and rewarding all stakeholders

🤗 Hugging Face's open and ethical AI approach

Clément Delangue explained Hugging Face's approach to ethical openness:

• Open documentation, with model cards
• Safeguards, with staged releases
• Community moderation
• Opt-in/opt-out datasets to respect copyright

Hugging Face's involvement in the creation of BLOOM, an open-source AI model, is a case in point. BLOOM has been assessed by Stanford as the most compliant with the EU AI Act. Last but not least, all the research advancement on copyright protection via watermarking of AI-generated content can only be done with open models and open datasets

Conclusion

Clément Delangue's testimony is a call-to-arms. The world is on the precipice of the AI era. Open-science and open-source AI is no longer a choice but a necessity. By embracing this ethical openness, we not only enable innovation but also ensure that AI progress benefits all.