The AI wars might have an armistice deal sooner than expected

A few months back, everyone wondered who would win the AI arms race. Microsoft aligned itself with OpenAI. Google launched Bard. Meta began working on its own large language model, LLaMA. Other companies began thinking of launching AI platforms, and curious users pitted the models against each other. 

But a recent deal suggests we may also see a growing number of partnerships, not just head-to-head competition. Earlier this week, Meta offered its LLaMA large language model for free under an open license and brought it to Microsoft’s Azure platform. The decision highlighted the benefits of interoperability in AI — and as more companies join the field, it probably won’t be the last of its kind. 

Well-known LLMs to date have been relatively siloed and offered in a more controlled environment where users need permission to build with the model or use the data. OpenAI continues to train GPT, releasing GPT-4 in March and providing developers with paid API access to the latest version of its model. Apple is developing its own LLM, called Ajax, though details are scarce; it is not yet publicly available, and its open-source status is unknown. Bard, Google’s LLM, is not open source at all. 

LLaMA was initially not publicly available and was accessible only through Meta, and Meta has yet to reveal its training data. But LLaMA was always intended to be open source and built to “further democratize access” to AI. This week, Meta at least partly delivered on that promise. Users of closed systems must pay a licensing fee for accessing the model where it is housed and distributing applications using that same model. The way Meta opened LLaMA, by making it available to Azure users and unlicensed to a certain degree, removes that inconvenience.

Meta opening up LLaMa and bringing it to Azure makes business sense, especially if Meta believes in openly developing AI. It’s a first step toward letting people access more LLM models on platforms and compare the results. A larger variety of LLM frameworks to choose from also puts into focus the question of how each model can work together. And LLM developers want people to use their models, so having them available on a wide array of platforms brings them to more users.

Even the most competitive Big Tech companies do business with each other. Meta is no stranger to working with Microsoft — Meta brought Microsoft’s Teams product to Workplace by Meta, which already runs the Office 365 suite. 

Openness has its risks. Ilya Sutskever, co-founder and chief scientist of OpenAI, a more open organization when it was founded in 2015, told The Verge he regrets sharing research over fear of competition and safety. Opening up datasets makes it easier to sue for copyright infringement, for example, because people can see which sources were scraped for data to train models. 

But having more LLM frameworks to choose from could be good news for advocates of AI interoperability. 

Since LLMs are, by default, distinct from each other, developers often have to choose which model to build apps with. There is no good way for the systems to talk. 

Walled gardens are no shock to most modern tech users, but AI interoperability advocates argue the only way AI can grow and evolve is not through closed silos but through open structures that can speak to each other. Even Microsoft believes in an interoperable AI; it joined other tech companies in joining the Open Neural Network Exchange, a group that wants to promote an industry standard for AI interoperability so developers can “find the right combinations of tools.” 

Letting AI systems work in tandem could lead to better results for things like search queries. Companies that can train models on different datasets could provide a better, fuller service — and, if one model is wrong, potentially avoid a catastrophic overreliance on one source of information. And being able to develop for both LLaMa and OpenAI’s GPT models in one place could cut development costs and timelines.

For now, LLaMa being available on Azure does not mean apps made with LLaMa can suddenly talk to those running on OpenAI’s GPT models. No one has created that bridge yet. Also, not everyone agrees that LLaMa checks all the boxes for open-source software, especially since it doesn’t use a license approved by the Open Source Initiative. It also limits who can commercially use LLaMa without a fee. Per its community license agreement, developers who have more than 700 million monthly active users “must request a license from Meta.”

But this is a good step in the right direction for open source and interoperability, if only to allow developers easier access between models. There is room for healthy competition, but if the companies truly want AI to evolve, working together is the best option.