Can Open Source AI End OpenAI’s Dominance?

Gatekeeping AI is so passé; unleash the innovation. 

Big tech companies like OpenAI are ruling the roost when it comes to artificial intelligence (AI) technology development. As of 2023, this AI research giant has secured investments worth over US$12 billion, backed by major names like Microsoft and Amazon. Impressively, the company made up about 80% of the global generative AI market in the same year. 

Well-known tools like ChatGPT, DALL-E and Llama 2, though originating from open-source software, are no longer open-source themselves. This strategy safeguards their dominance but raises the question of whether this approach is bending the playing field too much. 

So, can broader access to open-source AI disrupt this dominance, or are smaller companies forever destined to play catch-up?

But first, what is “open source”?

The open-source approach is grounded in the free distribution and sharing of technology for the public to use, share and modify. This philosophy has not only helped shape the modern Internet and cloud computing but also facilitated the rise of many billion-dollar companies. In the context of AI, open-source would mean openly distributing the technology that powers AI software.

Open-source AI models are appealing to businesses because they allow the use of large language models without the need to pay or share data with major vendors like Microsoft. Moreover, open-source AI promotes greater innovation, safety and transparency as development occurs in full view of the public. 

Reflecting this growing interest, global venture capital funding for open-source AI startups rose from US$900 million in 2022 to an impressive US$2.9 billion in 2023. According to the Wall Street Journal, some of the biggest names in tech believe open-source AI could be their ticket to challenging OpenAI’s market stronghold.

How “open” is open-source AI, really?

Traditional open-source software could be easily packaged in a “.zip” file and distributed without restriction. However, AI presents a more complex challenge and is less accessible, despite widespread claims of being open-source. Udbhav Tiwari, Head of Global Product Policy at Mozilla, points out that many so-called “open” AI projects are, in fact, “not open-source at all”.

Some critics deem such nominally accessible releases as “open washing”— a strategy where companies gain a reputation boost and free research contributions without truly providing the information needed for others to study, recreate or compete with their models.

Still, at the end of the day, open source is “open” source, to whatever extent it may be. A 2022 study noted that 80% of IT leaders plan on increasing their use of enterprise open-source software to build tools.

Is access to open-source AI enough?

Sadly, no. Open access means you and I can create AI software, even on a mobile phone. That said, there is little guarantee that these creations will match the quality of AI tools produced by big tech companies. And that boils down to money, as with most things (Meryl Streep in Mamma Mia sang it best: “All the things I could do, if I had a little money”).

Significant resources—both time and money—are required to audit even “open” models. For instance, it took nearly two years to identify images of child sexual abuse in the largest open-source dataset used for training generative AI. This highlights a stark reality: there’s a significant difference between democratizing access to AI and making it equitable.

Companies like Microsoft and Alphabet can afford the expensive Nvidia GPUs and the billions needed to build cutting-edge AI, but you and I probably cannot (thus, the search for a man in finance, trust fund, 6’5”, blue eyes, continues). 

If you’re a small startup wanting to make the most of open-source AI, the situation might seem bleak. Fret not. National agencies have taken it upon themselves to create more equitable conditions. 

The US federal government plans to establish a National AI Research Resource, and several universities have teamed up to create a high-performance computing center in Boston for advanced AI research. Apple, too, has recently launched OpenELM, which allows users to build AI models on their iPhones or Macbooks. Researchers are also designing smaller open models that are sufficiently powerful for many commercial uses and cheaper to train and run. 

Once open-source AI is accessible to all, what will happen to competition?

With access to high-quality open-source AI, businesses will shape the technology to suit their unique needs. Businesses and the average person will be able to create their own AI chatbots and apps without having to rely on major tools like ChatGPT or Bard. That’s where competition will thrive—not on who has access to AI tools but on who can best adapt and refine open-source AI to create innovative and effective solutions. 

Also read:

Header Image from Freepik


Share on facebook
Share on twitter
Share on linkedin
Share on email