Why is the future of AI open-source? A discussion of how open-source continues to drive AI forward

The most powerful technological force shaping our world today is artificial intelligence (AI). From speeding up the drug delivery process to optimizing global supply chains, AI has immense potential to improve lives. However, when not deployed with prudence, AI holds the power to perpetuate bias, mess up industry, and even reduce the qualities that make us human: creativity and connection.

In recent years, the AI world has made significant progress, and now, our daily lives are inextricably intertwined with AI. The driving force behind this evolution is the open-source movement. Open-source enables researchers and developers to share ideas and resources to collaboratively work on path-breaking AI projects.

The open-source revolution brewing at the heart of AI development seems like its taken straight from science fiction. Open-source promises a future where AI is not trapped in corporate dungeons but built outside in the open, code by code, by a community of innovators from across the world. Open-source provides a platform where it’s not competition but collaboration that drives advancements and where both raw power and ethical considerations hold equal weight.

Big Tech has its own agenda, masking closed-source models as open-source while seeking to reap the benefits of a truly open community. This article explores how open-source is one of the primary forces driving future innovation and advancements in the continual development of AI.

Open-source: The Genesis of AI

In 1971, most people were more familiar with Issac Asimov’s Three Laws of Robotics than with AI. But, when Richard M. Stallman (RMS) joined 美国麻省理工学院 's Artificial Intelligence Lab that year, AI was already a real subject.

As the years passed and proprietary software began popping up, RMS developed the revolutionary idea of Free Software. This concept took several decades to transform into open-source, which would become AI’s birthplace.

Open-source powers the AI of today and tomorrow

In an article in The Atlantic, Matteo Wong described the growing trend in the software community and academia for a truly open-source AI. The idea was to create relatively transparent AI models that are easier and cheaper for the public to study, use, and reproduce. Democratizing an incredibly concentrated technology that holds the potential to transform law enforcement, work, leisure, and maybe even religion is open-source’s goal for AI.

The same article states that Big Tech firms such as Meta are trying to fill market roles via ‘open-washing’ their products. Despite not truly open-sourcing their product, they are claiming the open-source community’s qualities and positive reputation. Nonetheless, there is no alternative to the real thing. This is because real open-source software fosters collaboration and innovation—two qualities AI desperately requires moving forward responsibly.

Even Bill Gates , who is not a fan of open-source, admits that the biggest thing since his introduction to the graphical user interface (GUI) idea in 1980 is open-source-based AI.

Though they cannot be considered open-source, today’s incredibly popular generative AI models like ChatGPT and Meta ’s LLaMA 2 originate from open-source origins.

An early investor in OpenAI , tech mogul Elon Musk said, "OpenAI was created as an open-source —which is why I named it "Open" AI — non-profit company to serve as a counterweight to Google, but now it has become a closed source, maximum-profit company effectively controlled by Microsoft. Not what I intended at all."

Despite that, all generative AI programs, including OpenAI, are built on open-source foundations. Today, Hugging Face ’s Transformer, in particular, is the top open-source library for constructing machine learning (ML) models. Don’t be fooled by the funny name; Hugging Face provides pre-trained models, tools, and architectures for natural language processing tasks. Using the company’s open-source libraries, developers can build upon existing AI models and fine-tune them for specific tasks. ChatGPT, in particular, depends on Hugging Face for its GPT LLMs. So it’s safe to say that without Hugging Face’s Transformer, ChatGPT wouldn’t exist.

Furthermore, even the mighty ChatGPT is fueled by 谷歌 and Facebook’s TensorFlow and PyTorch , respectively. Deep learning models are built and trained on the essential tools and libraries provided by these open-source Python frameworks.

If it weren’t for the open-source community’s efforts to tirelessly innovate, improve, and build the extensive open-source LLM libraries that currently exist, the development of AI would have truly suffered for the worst. Even private tech companies, with their endless coffers, would have found it impossible to gather the enormous amounts of data required to power the current AI models.

AGI and how the future of AI is undoubtedly open-source

Without any doubt, the next phase of AI will be artificial general intelligence (AGI), a field where computers will meet and exceed human intelligence. According to industry experts, AGI will most certainly be open-source.

There are several reasons for AGI to choose open-source, such as the usual reasons: community, security, customization, quality, and cost. However, three main factors to AGI differentiate it from other open-source software:

1. There are extreme risks and ethical concerns. Developers and tech companies must make these public and set a system for compliance and verification with whatever standards arise.
2. Since the algorithm is unknown, open-source encourages experimentation.
3. There are chances that AGI will arrive faster than people think. Therefore, it is important to get serious about the conversation.

Vitally, open-source can facilitate a rational conversation. We cannot ban AGI outright. This would only shift the development of AGI to countries, companies, and organizations that would not recognize the ban. Undoubtedly, we also cannot accept an AGI free-for-all because the existence of miscreants willing to harness AGI for nefarious purposes is a real thing.

Therefore, we should seek out open-source strategies for AGI that include numerous perspectives and strive to make the development of AGI public while getting the AI and AGI community on the same page about limiting the risks of AGI.

Additionally, having a reasoned discussion and building safeguards into the code is vital because an open development process is our only chance of achieving these goals.

What are the drawbacks of open-source AI?

According to the The Linux Foundation ’s 2023 survey, 41% of the organizations noted that the preference for open-source generative AI is juxtaposed with lingering security concerns, highlighting the requirement for more robust security measures within open-source AI models.

While favored for its integrative and collaborative potential, there is a gap in translating open-source AI’s potential into successful operational implementations, bringing into view the inherent disconnect between practical execution and collaborative intent. Issues such as the long-term sustainability of open-source AI, short-term challenges, and the difference in the performance quality and user experience of open-source AI models compared to proprietary AI software show that all is not fine even within the open-source AI development community.

From the cradle to the grave and everywhere in between: Open-source is the future of AI

Without a doubt, modern-day proprietary generative AI models stand upon the shoulders of the open-source giants, which have been in development for several decades at least. The open-source community has played and continues to play a significant role in the advancement of AI. The community’s collaborative spirit, a penchant for innovation, and the freedom to experiment provide the jet fuel AI needs to propel itself to the forefront of technology.

While there are significant issues with relying upon the open-source community for the continued development of AI, it cannot be counted out. Open-source is resolutely poised as one of the most prominent forces pushing AI into the coming future. The future of AI is undoubtedly open-source.