Over the last two decades, open source has become a significant and sometimes dominant distribution channel and business model for software. Companies like Open AI (at the beginning) and Hugging Face have based their success on this model. Other companies, such as Meta have used open source to weaken a competitor’s position. How to decide when to use open source as a strategic tool?
At its core, open source is a philosophy that promotes the free and transparent distribution of software, allowing anyone to view, modify, and distribute the source code. In the realm of generative AI, this translates to the open sharing of algorithms, models, and training methodologies. Such transparency stands in contrast to proprietary AI systems, which guard their inner workings zealously. Open sourcing in AI not only democratizes access but also fuels collaborative innovation, inviting minds from across the globe to contribute, refine, and advance the state of artificial intelligence.
The past decade has witnessed a tectonic shift in the generative AI landscape. As the demand for sophisticated AI solutions surged, so did the realization of the immense value in open collaboration. Companies and research institutions began to move away from siloed development, instead opting to release groundbreaking models and tools to the public domain. This open-source ethos, underpinned by platforms like GitHub and research forums like arXiv, has cultivated a thriving ecosystem where AI advancements are accelerated, barriers to entry are lowered, and the once-exclusive domain of high-end AI research has become accessible to enthusiasts, startups, and researchers alike.
In Generative AI, open-source models caught up rapidly in terms of performance and are much less costly. A $100 open-source model with 13 billion parameters is competing with a $10 million Google high-end model with 540 billion parameters.
The benefits of open source
Building and distributing software in open source has several advantages.
Accelerated Innovation:
The open-source model acts as a catalyst for accelerated innovation. Unlike traditional software development, where a select group of developers are gatekeepers to progress, open source thrives on the principle of crowdsourced development. This collective intelligence, hailing from diverse backgrounds across the globe, ensures that a vast array of innovative solutions continuously surface. The collaborative spirit of the open-source community allows for rapid prototyping and iteration, enabling solutions that not only meet the current market needs but also anticipate future challenges.
Talent Attraction & Retention:
In an age where skilled developers are highly sought-after commodities, companies that actively participate in and support open source projects have an edge. By engaging with the developer community, businesses can showcase their commitment to transparency, collaboration, and continuous learning. This not only attracts top-tier developer talent but also resonates with the ethos of many modern developers who value the open-source movement. Furthermore, being a part of such projects offers employees unique challenges and growth opportunities, making it not just an avenue for talent attraction but also a strategy for talent retention.
Ecosystem & Network Effects:
Open source is more than just a software development model; it’s a nexus of symbiotic relationships and collaborations. As businesses engage in open source projects, they often find themselves forming partnerships with other enterprises, developers, and even competitors. This collaborative environment fosters a robust ecosystem where businesses can leverage the strengths of multiple entities, leading to products and solutions that are greater than the sum of their parts. Dominant open-source projects can even become industry standards, further amplifying the network effects and ensuring that businesses integrated into this ecosystem benefit from widespread adoption and trust.
In the broader strategic view, while direct revenues from the software might be limited, the cost efficiencies and indirect benefits (like market influence, standard setting, and ecosystem establishment) can be significant, often outweighing the traditional proprietary software model’s benefits.
The tradeoffs made with open source
However, the choice to use open source as a distribution mode and business model has its cost and is a strategic choice for the companies offering digital services. The debate between closed-source and open-source software distribution has been vivid over the last two decades and we have seen successes (and failures) in both models. Interestingly, big tech giants are often using both models and we can expect the same situation for LLMs: some will rely on a closed approach to monetising through API and licences, others will rely on an open approach and will have service-oriented revenues.
Choosing between closed and open source is mainly a tradeoff between diffusion and value capture. The more open, the higher the diffusion and the lower the value capture.
Open source as a strategic tool to change the established competitive positions
Beyond the comparison between the pros and cons of open source as a distribution strategy and business model we could discuss when using open source could serve as a strategic weapon.
Open Source as a Way to Build a New Product Fast and Enter a Market Quickly
The competitive nature of the technology landscape necessitates rapid product development and timely market entry. Open source offers an unparalleled avenue for this, enabling companies to build upon existing frameworks, libraries, and tools. Instead of starting from scratch, businesses can leverage vast repositories of open-source codes, significantly reducing the development cycle.
Hugging Face, a leading name in the natural language processing (NLP) segment of generative AI, adopted this approach. By open-sourcing their Transformers library, they not only built upon the contributions of the global developer community but also positioned themselves as frontrunners in the democratization of NLP tools. This strategic move allowed them rapid market penetration and brand recognition. Similarly, OpenAI started with an open-source model, which helped the organisation benefit from the efforts of other researchers who contributed to the initial development of the model.
Open Source as a Way to Differentiate and Cultivate an Independence Value Proposition
Proprietary models are often criticised for the lock-in effects they impose on their clients and for the unability of the clients to customize the model to their needs.
In a saturated market, differentiation is key. Open sourcing can be that differentiator. By offering open-source solutions, companies can underscore a value proposition rooted in transparency, user empowerment, and independence from proprietary constraints. Users are not just passive consumers but can actively modify, improve, and even commercialize the software, fostering a sense of ownership and trust.
For example, Stability AI’s bold move of open-sourcing Stable Diffusion is a way to differentiate their product against other similar products cultivating a positioning of independence for the users. This had a very positive impact as the product quickly gained incredible traction within the generative AI community.
Open Source to Catch Up with Competitors and Make Their Offer Irrelevant
Companies lagging behind in the competitive race can strategically utilize open source to challenge established market leaders. By open-sourcing tools or platforms that directly compete with proprietary solutions, they can disrupt established revenue streams and force competitors to re-evaluate their offerings. Such a move can democratize access to high-end tools, rendering expensive commercial alternatives less appealing.
Meta (previously known as Facebook), in a bid to position itself prominently in the AI domain, open-sourced their Llama 2 model. Llama 2, designed for large language models, was a significant contribution to the AI research community. By open-sourcing it, Meta not only fostered goodwill within the developer and research community but also placed indirect competitive pressure on entities with similar but proprietary offerings. The strategy aimed at making their competitor’s offerings seem restrictive or less appealing in comparison, while also underscoring Meta’s commitment to a collaborative and transparent future in AI.
Beyond seeing the choice between proprietary and open source models as a once-for-all decision, it’s useful to see this choice dynamically and go from one to the other according to the relative position of the company in its environment and the respective choices of the competitors in that matter.
Photo de Matt Ridley sur Unsplash
