Meta Platforms (META 3.43%) recently announced that it will make its new large language model (LLM) for artificial intelligence (AI) freely available as open-source software. It's a move that contrasts with strategic shifts that generative-AI leader OpenAI has made for its own hugely popular ChatGPT service.
While OpenAI previously made its LLM software available on an open-source basis, the company ultimately decided that the approach wasn't a good idea. Meta's move to offer Llama 2 as freely available and customizable software might close off some sales opportunities and create some additional downsides, but CEO Mark Zuckerberg believes it's ultimately the right move for the company. Here's why.
Meta aims to shake up the AI space
In a post published on Meta's Facebook platform, Zuckerberg outlined details about the launch of his company's new large language model and the decision to make the software freely available for most commercial users. Here's what he said on the significance of making Llama 2 open source:
Open source drives innovation because it enables many more developers to build with new technology. It also improves safety and security because when software is open, more people can scrutinize it to identify and fix potential issues. I believe it would unlock more progress if the ecosystem were more open, which is why we're open-sourcing Llama 2.
While the core service won't generate any direct revenue for Meta, the company may be able to sell additional support and add-on services. Making the software open source will also potentially curb growth opportunities for its biggest rivals in the AI space.
Rather than pay for commercial LLM software, many potential users may opt to use Meta's free offering. This could lay foundations for the company to generate valuable data that can be used to train increasingly advanced artificial-intelligence models.
The launch of Llama 2 also represents an expansion of Meta's existing partnership with Microsoft. The social media giant has chosen Microsoft's Azure cloud infrastructure service as the preferred partner for its new, open-source large language model. The software has already been made available on Microsoft's Azure cloud infrastructure service, and it will also be available on Amazon's AWS, Hugging Face, and other providers.
While Meta competes with a wide range of players in the tech space, its focus on the digital advertising market means that Alphabet is one of its biggest rivals. Apple is another company of concerned. Because Alphabet and Apple own the leading operating systems for mobile platforms, decisions they make can have a large impact on Meta's business.
By offering Llama 2 as open-source software, Meta may be able to head off some of Alphabet's and Apple's moves in the AI space. But OpenAI believes that making advanced LLMs freely available for use and customization also opens doors for potential misuse.
Is OpenAI right about the dangers of open source?
OpenAI actually started as a company focused on providing open-source AI software. While the company's name hasn't changed, its approach to software access and customization has. The artificial intelligence specialist has pivoted away from making code and information about data training available, and it believes that trying the open-source approach was a mistake.
While OpenAI has cited competitive reasons as the primary rationale for ditching the open-source approach, top executives at the company have also cited safety concerns as a motivating factor for shifting to a closed model.
In particular, Open AI's chief scientist and co-founder Ilya Sutskever has warned about the potential for abuse as AI models become more advanced. Sutskever believes that it will eventually be very easy for bad actors to use artificial intelligence applications to cause harm -- and that making access to the inner workings of AI software less readily available could help avert damaging outcomes.
While it remains to be seen whether Llama 2 will be abused by bad actors, it's clear why Zuckerberg and Meta have a different take on the competitive aspects of open source. As a massive social networking technology company that also has business and potential growth drivers in a wide range of other categories, Meta isn't as dependent on directly monetizing its LLMs.
By bringing more users into its own AI ecosystem, Meta strengthens its ability to compete with ChatGPT, Bard, and other leading offerings in the space. Additionally, Llama 2's open-source status comes with a bit of a catch -- at least for some of Meta's biggest competitors. The company has stipulated that potential licensees with more than 700 million monthly active users must receive special permission to use the software.
Therefore, while remaining open source clearly wasn't the right move for OpenAI, it could wind up paying off for Meta Platforms.
