The AI renaissance and why open innovation matters

In our previous article, we highlighted various advancements that have been made by the Open Source community around LLMs in recent months. This second part will focus on various AI regulation efforts currently underway and how this will impact open innovation. While some regulation is important to create safeguards around AI, overregulation may in fact negatively impact Open Source developers and startups and consequently inhibit innovation.

Open Source LLMs have the potential to provide many benefits when compared to proprietary solutions, namely, more control, stronger privacy, reduced costs, better results and improved performance. For these reasons, individuals and organizations are considering the alternatives to proprietary solutions.

At the same time, governments around the world are worried about the consequences of making AI technologies generally available to the public and are starting to draft regulations to tackle these concerns.

On June 14th, the European Union voted to move forward with the AI Act, which assigns applications of AI to different risk categories. While many have applauded the groundbreaking work by the European Parliament, others are worried about how overregulation can stifle innovation.

In the US, the White House is drafting a blueprint for an AI Bill of Rights, seeking insights from researchers, technologists, advocates, journalists and policymakers. It’s worth highlighting 3 organizations who have expressed their opinion before the US Congress on this matter.

On May 16th, Stability AI shared a detailed paper emphasizing the importance of open models for a transparent, competitive and resilient digital economy. Stability AI CEO Emad Mostaque commented: “These technologies will be the backbone of our digital economy, and it is essential that the public can scrutinize their development. Open models and open datasets will help to improve safety through transparency, foster competition, and ensure the United States retains strategic leadership in critical AI capabilities. Grassroots innovation is America’s greatest asset, and open models will help to put these tools in the hands of workers and firms across the economy.”

In a more alarming congressional testimony, OpenAI CEO Sam Altman calls for extensive regulation, including a new government agency charged with licensing AI models. If congress decides to move forward in this direction, this will create a regulatory hell for small startups and Open Source developers. One reason why OpenAI is so concerned about AI regulation is that it wants to create a competitive advantage (or moat) that will inhibit innovation from challengers. A leaked internal Google document highlights how Open Source models will outcompete Google and OpenAI.

On June 22nd, another more optimistic congressional testimony was from Hugging Face CEO, Clément Delangue, who defended the importance of Open Source models in advancing innovation, promoting fair competition, and ensuring responsible development. Delangue emphasized that Open Source principles democratize AI and foster a more inclusive and collaborative future in the field.

It’s interesting to note how Open Source models are taking shape, not only from the community of researchers and developers, but also from businesses. A leaked pitch memo from startup Mistral criticized how big players like OpenAI are embracing a “closed technology approach” — not publishing the models they build and only letting people access them through an API. In contrast, Mistral highlights how they are betting on Open Source models and focusing on privacy and security as a way to gain market share. This approach has helped them raise €105m in funding.

Another company that is betting big on Open Source models is Databricks, the company behind Apache Spark, with a valuation of $38 Billion. In their recent Data + AI Summit, which brought in 12,000 attendees to the Moscone Convention Center in San Francisco plus 75,000 online participants, CEO of Databricks Ali Ghodsi reaffirmed his commitment to promoting Open Source models as a path towards democratizing AI. As part of this commitment, Databricks announced the acquisition of MosaicML for $1.3 Billion. MosaicML is known for its state-of-the-art MPT large language models.

Recently, Meta has released Llama 2. While this is not an Open Source model per se because it includes clauses that restrict commercial use, Meta has recognized the importance of community contributions based on their experience with other Open Source projects like PyTorch. By making the source available to a community of developers and researchers from around the world, it allows the project to quickly evolve in ways that it was not foreseen by the original authors.

If you are interested in learning more about Open Source and AI, please join our “Deep Dive: Defining Open Source AI” series. CFPs for the online webinars are open and we are looking for proposals that discuss AI regulation and its impact on open innovation.