When Meta CEO Mark Zuckerberg announced last year that his company would be releasing an artificial intelligence system, Jeffrey Emanuel had reservations.
Mr. Emanuel, a part-time hacker and full-time AI enthusiast, had been tinkering with “closed” AI models, including OpenAIs, meaning the systems’ underlying code could not be accessed or changed. When Mr. Zuckerberg introduced Meta’s AI system to just a handful of academics at the invitation of just a handful of academics, Mr. Emanuel worried that the technology would remain reserved for only a small group of people.
But in a release last summer of an updated AI system, Mr. Zuckerberg made the code “open source,” allowing it to be freely copied, modified and reused by anyone.
Mr. Emanuel, the founder of blockchain startup Pastel Network, has been sold. He said he appreciated that Meta’s AI system was powerful and easy to use. Above all, he liked how Mr. Zuckerberg advocated making the technology freely available to hack code – largely the opposite of what Google, OpenAI and Microsoft have done.
“We have that champion in Zuckerberg,” said Mr. Emanuel, 42. “Thank God we have someone who is protecting the open source ethos from these other big companies.”
Mr. Zuckerberg has become the most high-profile technology executive supporting and promoting the open source model for AI. That’s put the 40-year-old billionaire right at the end of a contentious debate about whether the potentially world-changing technology is dangerous in making it available to any programmer who wants it.
Microsoft, OpenAI and Google are adopting more of a closed AI strategy to protect their technology, which they say is out of caution. But Mr. Zuckerberg has been vocal about wanting the technology to be accessible to everyone.
“This technology is so important and the possibilities are so great that we should open source it and make it as widely available as possible, as responsibly as possible, so that everyone can benefit from it,” he said in an Instagram video in January .
That attitude has made Mr. Zuckerberg the unlikely man of the moment in many of Silicon Valley’s developer communities, prompting talk of a “light-up” and a kind of “Zuckaissance.” Even as the CEO continues to address misinformation and child safety issues on Meta’s platforms, many engineers, programmers, technologists and others have embraced his position on bringing AI to the masses.
Since releasing Meta’s first fully open-source AI model, called LLaMA 2, in July, the software has been downloaded more than 180 million times, the company said. A more powerful version of the model, LLaMA 3, released in April, reached the top of the download charts on Hugging Face, an AI code community site, at record speed.
Developers have created tens of thousands of their own custom AI programs based on Meta’s AI software to do everything from helping doctors read radiology scans to creating numerous digital chatbot assistants.
“I told Mark that I think open-sourcing LLaMA is the most popular thing Facebook has ever done in the tech community,” said Patrick Collison, CEO of payments company Stripe, who recently joined a Meta strategic advisory group The aim is to support the company in strategic decisions about its AI technology. Meta owns Facebook, Instagram and other apps.
Mr. Zuckerberg’s new popularity in technology circles is notable because of his checkered past with developers. Over the course of two decades, Meta has sometimes pulled the rug out from under programmers’ feet. In 2013, for example, Mr. Zuckerberg bought Parse, a company that built developer tools to attract programmers to develop apps for the Facebook platform. Three years later, he canceled the project, angering the developers who had invested their time and energy into the project.
A spokeswoman for Mr. Zuckerberg and Meta declined to comment. (The New York Times sued OpenAI and its partner Microsoft last year for copyright infringement of news content related to AI systems.)
Open source software has a long and storied history in Silicon Valley, with major technology battles revolving around open versus proprietary – or closed – systems.
In the early days of the Internet, Microsoft struggled to provide the software that ran the Internet infrastructure, but ultimately lost out to open source software projects. More recently, Google has open-sourced its Android mobile operating system to rival Apple’s closed iPhone operating system. Firefox, the internet browser, WordPress, a blogging platform, and Blender, a popular set of animation software tools, were all built using open source technologies.
Mr. Zuckerberg, who founded Facebook in 2004, has long supported open source technology. In 2011, Facebook launched the Open Compute Project, a nonprofit organization that freely shares designs of servers and devices in data centers. In 2016, Facebook also developed Pytorch, an open source software library commonly used to build AI applications. The company also shares blueprints of computer chips it has developed.
“Mark is a great student of history,” said Spotify CEO Daniel Ek, who considers Mr. Zuckerberg a confidant. “Over time, he realized in the computer industry that there are always closed and open paths you can take. And it’s always open by default.”
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 the software, fearing a backlash from regulators in Washington and the European Union. But meta-technologists like Yann LeCun and Joelle Pineau, who are at the forefront of AI research, pushed for the open model, which they said would better serve the company in the long run.
The engineers won. Mr. Zuckerberg agreed that the code could be improved and protected more quickly if it was open, he said in a post on his Facebook page last year.
While open-sourcing LLaMA means giving away computer code that Meta spent billions of dollars to create without an immediate return on the investment, Mr. Zuckerberg calls it “good business.” The more developers use Meta’s software and hardware tools, the more likely they are to invest in the technology ecosystem, helping to strengthen the company.
The technology has also helped Meta improve its own internal AI systems to support ad targeting and recommendations for more relevant content across Meta’s apps.
“It’s 100 percent consistent with Zuckerberg’s incentives and how Meta can benefit from them,” said Nur Ahmed, a researcher at MIT Sloan who studies AI. “LLaMA is a win-win situation for everyone.”
Competitors are taking note. In February, Google open sourced the code for two AI models, Gemma 2B and Gemma 7B, a sign that the company was feeling the pressure of Mr. Zuckerberg’s open source approach. Google did not respond to requests for comment. Other companies, including Microsoft, Mistral, Snowflake and Databricks, have also started offering open source models this year.
For some programmers, Zuckerberg’s AI approach hasn’t erased all the baggage of the past. Sam McLeod, 35, a software engineer in Melbourne, Australia, deleted his Facebook accounts years ago after he became uncomfortable with the company’s track record on user privacy and other factors.
But more recently, he said, he realized that Mr. Zuckerberg had released “cutting-edge” open-source software models with “permissive licensing terms” that couldn’t be said of other major technology companies.
Matt Shumer, 24, a developer in New York, said he used closed-loop AI models from Mistral and OpenAI to power digital assistants for his startup HyperWrite. But after Meta released its updated open-source AI model last month, Mr. Shumer began relying heavily on it instead. Any reservations he had about Mr. Zuckerberg are a thing of the past.
“The developers have started to move past a lot of the problems they had with him and Facebook,” Mr. Shumer said. “Right now, what he’s doing is really good for the open source community.”
Source link
2024-06-01 02:04:24
www.nytimes.com
