Digital Economy Dispatch #214 -- AI's Open Questions

Digital Economy Dispatch #214 -- AI’s Open Questions
15th December 2024

As AI continues to reshape the technology landscape, fundamental questions have emerged at the intersection of innovation, value creation, and access:

How important is open source to the future of AI? What does it really mean for AI to be open source? Where should open source fit in an organization’s AI-at-Scale strategy?

The recent release of the Open Source Initiative's (OSI) first formal definition of open source AI marks a pivotal moment in addressing such questions, but it also highlights the complex challenges facing the industry. Following extensive collaboration with academic and industry partners, the OSI's AI definition establishes a benchmark for assessing the open source nature of AI technologies. But, does it matter?

The drive toward open source AI represents more than just a sideshow or a philosophical stance—it's becoming a strategic imperative that could determine the future development and democratization of AI. By observing BigTech companies as they drive advances in AI technology, we see important tensions arising between their aspirations for ensuring openness and transparency, and the commercial realities of ensuring significant recurring revenue from their substantial AI investments. The Economist sees this result of this struggle to be core to each BigTech company’s pursuit of AI dominance.

The Stakes of Openness

The implications of how we define and implement open source AI extend far beyond technical considerations. In an era where AI capabilities are increasingly concentrated among a handful of large technology companies, the open source movement represents a potential counterbalance to this consolidation of power. Open source AI holds a key role in fostering innovation, transparency, and accessibility, ensuring that AI benefits society as a whole rather than being controlled by a few powerful entities.

However, interpretations of the term “open source AI” have been broad and diffuse. Hence, the importance of the OSI's new definition. It establishes clear criteria – to be considered open source, an AI model must be:

• Transparent: The model's design and training data should be accessible and understandable.

• Free to Use: Users should have the freedom to use the model for any purpose, without restrictions.

• Free to Modify: Users should be able to modify and improve the model.

• Free to Share: Users should be able to share their modifications with others.

While many companies have labelled their AI models as "open source", some have faced criticism for not fully adhering to these principles. Meta, for instance, has been criticized as “polluting open source” for imposing restrictions on the use of its Llama models, despite claiming them to be open source.

However, access to the AI models is only one concern. The use of proprietary training data presents another challenge. Companies often rely on vast amounts of data to train their models, and sharing this data can be a competitive disadvantage. Additionally, copyright issues and potential legal ramifications further complicate the landscape of open source AI.

Lessons on Open Approaches to Software

To understand these issues more deeply, it is worth recalling that open source has a long history in computing. For instance, more than 20 years ago I co-authored a well-received paper with Grady Booch on the broader implication of taking an open approach to software delivery. In that paper we argued that open source isn't merely a licensing model—it's a comprehensive approach encompassing open collaboration, transparent processes, and frequent releases.

The core of our argument was that commercial software and solution providers must approach the decision to adopt open source approaches with care by considering several critical factors. These include the software's potential as a platform, impact on revenue models, identified market needs, and emphasis on a healthy ecosystem via server-focused services. Success in open source initiatives demands proper licensing choices, early momentum building, and deployment of appropriate tools for managing distributed development. The business case must be clearly defined, as giving away intellectual property only makes sense when it provides access to a substantially larger market or user base.

As a consequence of these concerns, we highlighted that complete transition to open source may not be economically viable for most commercial vendors. Hence, selective adoption of open source practices and technologies is inevitable, and can provide significant competitive advantages. Success requires careful strategic planning, clear understanding of both opportunities and risks, and thoughtful integration of open source elements into existing business models.

This message remains critical in the age of AI. With the arrival and rapid evolution of AI technologies, the ability to effectively leverage open source while maintaining commercial viability is increasingly important for business success. The success of open approaches in software development suggests a promising path for AI, but the stakes and challenges are arguably even higher.

Commercial Realities and Strategic Choices

The key lesson from those observations, therefore, is that the journey toward open source AI is ideologically important, but complicated by commercial realities. In reality, companies must carefully balance the benefits of openness against competitive advantages.

Meta's evolution provides a telling example: their Llama model, while promoted as open source, has significant usage restrictions. Moves such as Meta's decision to allow their AI technology to be used for defence applications are on the one hand important to overcome some of these barriers. But, on the other hand, can equally be viewed as a demonstration of how open source strategies are inevitably tempered by broader business and geopolitical considerations.

Of course, promoting open source can also be an important commercial advantage. For example, in the case of Red Hat, their stated vision for AI offers different perspective, emphasizing how open source can democratize access while maintaining commercial viability. Unsurprisingly, their approach aligns with the core values of Red Hat, which has long championed open source solutions like Linux, KVM, OpenStack, and Kubernetes. Their AI focus on smaller, specialized models highlights a practical approach to balancing accessibility with business requirements. By enabling organizations to fine-tune AI models for specific needs, they're demonstrating how open source can create business value without requiring companies to completely abandon proprietary advantages.

The Security and Innovation Paradox

As critics will argue, use of open source is not without its issues. One of the most pressing challenges in open source AI involves security. The transparency that makes open source valuable for innovation also raises concerns about potential misuse. However, as the software industry has demonstrated, community scrutiny often enhances security rather than compromising it. The key lies in fostering responsible development practices while maintaining the collaborative benefits of open source.

Innovation presents another apparent paradox with open approaches. While proprietary development can protect intellectual property and maintain competitive advantages, open source approaches often accelerate progress through collective effort. The success of Linux, Kubernetes, and other open source projects suggests that carefully managed openness can drive innovation more effectively than closed development.

Toward an Open Path For AI-at-Scale

Addressing these issues raises a significant challenge in defining an appropriate AI-at-Scale strategy. For digital leaders navigating these waters, several strategic considerations emerge:

1. Selective Adoption: Organizations don't need to choose between completely open or closed approaches. As with traditional software, companies can strategically incorporate open source elements while maintaining proprietary advantages where they matter most.

2. Platform Thinking: Success in open source AI may depend less on individual models and more on creating platforms and ecosystems that enable broader innovation. This approach allows organizations to benefit from community contributions while maintaining valuable positioning.

3. Market Access: As demonstrated in traditional software, giving away certain intellectual property can make strategic sense when it provides access to substantially larger markets or user bases. The same principle applies to AI, particularly as the technology becomes more fundamental to business operations.

4. Infrastructure Integration: The ability to deploy AI workloads across various environments – from edge to cloud – becomes crucial. Organizations need to consider how open source AI fits into their broader technology infrastructure.

Opening Up AI

The future of AI will likely mirror the software industry's evolution, where open source and proprietary solutions coexist and complement each other. The OSI's definition provides a framework for this future, but the real work lies in implementation. Success will require careful strategic planning, clear understanding of both opportunities and risks, and thoughtful integration of open source elements into existing business models.

For digital leaders, the key question isn't whether to embrace open source AI, but how to leverage it effectively while maintaining competitive advantage. As AI becomes increasingly central to business operations, the ability to navigate this balance will become a critical determinant of success. The open question facing AI isn't just about technology—it's about finding the right model for innovation that serves both commercial interests and the broader good.

To deliver AI-at-Scale requires balancing openness with strategic value and commercial realities. The path forward requires careful consideration of how open source principles can be applied to AI in ways that promote innovation, ensure security, and create sustainable business value.