When Meta CEO Mark Zuckerberg announced last year that the company would release an artificial intelligence system, Jeffrey Emanuel was skeptical.
Emanuel, a part-time hacker and full-time AI enthusiast, has been improving “closed” AI models, including those from OpenAI, meaning the system's underlying code cannot be accessed or modified. When Zuckerberg introduced Meta's AI system to a small group of academics by invitation only, Emanuel worried the technology would remain restricted to a select few.
But when Zuckerberg released an updated version of his AI system last summer, he made the code “open source,” meaning anyone could freely copy, modify and reuse it.
Emanuel, founder of blockchain startup Pastel Network, said he appreciates that Meta's AI system is both powerful and easy to use, and most of all, he likes that Zuckerberg supports hacker code to make the technology freely available, which is almost the opposite of what Google, OpenAI and Microsoft have done.
“We have a champion in Zuckerberg,” said Emanuel, 42. “It's good to have someone defending the open-source ethos against other big companies.”
Zuckerberg has become the most prominent tech executive to champion and promote an open-source model for AI, putting the 40-year-old billionaire on one side of a divisive debate about whether the potentially world-changing technology is too dangerous to make available to any programmer who wants it.
Microsoft, OpenAI and Google have adopted more closed AI strategies to protect their own technology out of an abundance of caution, but Zuckerberg has been vocal in saying the technology should be open to everyone.
“This technology is so important and the opportunity is so great that we should open source it and make it as widely available as possible responsibly so that everyone can benefit,” he said in an Instagram video in January.
This stance has made Zuckerberg an unlikely celebrity among many in Silicon Valley's developer community, sparking talk of a “resurrection of fame” and a kind of “Zuckerbergism.” Though the CEO continues to grapple with investigations into misinformation and child safety issues on Meta's platform, many engineers, programmers, technologists and others support his stance on bringing AI to the masses.
Meta says its first fully open-source AI model, LLaMA 2, has been downloaded more than 180 million times since its release in July. A more powerful version of the model, LLaMA 3, released in April, topped the download charts on AI-code community site Hugging Face in record time.
Building on Meta's AI software, developers have created tens of thousands of unique, customized AI programs that can do everything from helping clinicians decipher radiology scans to creating an army of digital chatbot assistants.
“I said to Mark, I think open-sourcing LLaMA is the most popular thing Facebook has ever done in the tech world,” says Patrick Collison, CEO of payments company Stripe. Collison recently joined the Meta Strategic Advisory Group, which aims to help the company make strategic decisions about AI technology. Meta owns Facebook, Instagram and other apps.
Zuckerberg's newfound popularity in the tech world stems from his rocky history with developers. For more than two decades, Meta has had programmers stoop to its limits. In 2013, for example, Zuckerberg bought developer tools company Perth to attract programmers to build apps for Facebook's platform. But three years later, he killed the effort, angering developers who had poured time and energy into the project.
Zuckerberg and a spokesman for Meta declined to comment. (The New York Times filed a lawsuit last year against OpenAI and its partner Microsoft, alleging that news content about its AI systems infringed copyright.)
Open source software has a long history in Silicon Valley, where major technical debates have taken place over open versus proprietary, or closed, systems.
In the early days of the Internet, Microsoft made a big push to provide software that ran the Internet infrastructure but ultimately lost out to open source software projects. More recently, Google open sourced its Android mobile operating system to compete with Apple's closed iPhone operating system. The Internet browser Firefox, the blogging platform WordPress, and the popular animation software tool set Blender were all built using open source technologies.
Zuckerberg, who founded Facebook in 2004, has long been a supporter of open source technology. In 2011, Facebook launched the Open Compute Project, a nonprofit organization that freely shares blueprints for the servers and equipment in its data centers. In 2016, Facebook also developed Pytorch, an open-source software library that is widely used to create AI applications. The company has also published blueprints for the computing chips it develops.
“Mark is a great student of history,” said Spotify Chief Executive Daniel Ek, who considers Mr. Zuckerberg a close friend. “In his many years working in the computer industry, he's always seen that there are closed paths and open paths, and he's always chosen the open path.”
At Meta, the decision to open-source its AI was controversial. In 2022 and 2023, the company's policy and legal teams supported a more conservative approach to releasing software, fearing backlash from regulators in Washington and the European Union. But Meta technologists, including AI research spearhead Yann LeCun and Joël Pinault, pushed for an open model, arguing it would be more beneficial for the company in the long run.
The engineers won: Zuckerberg agreed that the code could be improved and secured more quickly if it was public, he wrote in a post on his Facebook page last year.
Open-sourcing LLaMA means Meta is giving away billions of dollars of computer code for free, with no immediate payback. But Zuckerberg calls it “good business.” The more developers use Meta's software and hardware tools, the more likely they are to invest in the company's tech ecosystem, helping to cement the company's position.
The technology also helps improve Meta's in-house AI systems, aiding in ad targeting and recommending more relevant content on Meta's apps.
“This is perfectly aligned with Zuckerberg's motivations and how he wants to benefit Meta,” said Noor Ahmed, an AI researcher at MIT Sloan. “LLaMA is a win-win for everyone.”
Competitors are taking notice. Google open-sourced the code for two AI models, Gemma 2B and Gemma 7B, in February, a sign that the company is feeling the pressure from Zuckerberg's open-source efforts. Google did not respond to a request for comment. Other companies, including Microsoft, Mistral, Snowflake and Databricks, have also started offering open-source models this year.
For some programmers, Mr. Zuckerberg's work on AI hasn't wiped away all the old clutter. Sam McLeod, a 35-year-old software developer from Melbourne, Australia, deleted his Facebook account a few years ago because he was put off by the company's track record on things like user privacy.
But more recently, he said he recognized that Zuckerberg had released a “cutting edge” open source software model with “permissive licensing terms” — something that can't be said for other major tech companies.
Matt Schumer, a 24-year-old developer in New York, said he used closed AI models from Mistral and OpenAI for his startup Hyperlight's digital assistant. But after Meta released an updated open-source AI model last month, Schumer began relying more heavily on it instead. His concerns about Zuckerberg are now a thing of the past.
“Developers are starting to get over a lot of the issues they had with him and Facebook,” Schumer said. “What he's doing right now is a really good thing for the open source community.”