What is open source AI? New definition shows that Meta’s version is not what it claims to be

Meta and some other tech companies rolling out so-called open source generative artificial intelligence (AI) models are “disenfranchising the public from innovation cycles” and profiting from them, according to the group that pioneered the open source term in software for the past 25 years.

Open source is yet another buzzword in AI circles at Big Tech companies like Meta and Elon Musk’s Grok AI model, stating that open source is “good for the world.” said Facebook founder Mark Zuckerberg.

But no one can agree on what open source AI means.

That could change if the Open Source Initiative (OSI), the organization that is the self-appointed steward of the term, states a definitive definition for open source AI on Mondayand it is not the same as Meta’s version of the term.

“They are failing, especially Meta, because their terms of use and distribution conditions are incompatible with the open source definition and open source principles,” Stefano Maffulli, head of the OSI, told Euronews Next.

“They’re basically the Microsoft, the Oracle and the Adobe of this space, where they say, ‘build on top of my platform, don’t worry about it and I’ll continue to get subsidies from you if you use our platforms.’ But they also say, ‘it’s open, so anyone can use it,'” he added.

What is the OSI’s AI definition?

It took a few years to create the OSI definition, and the organization consulted a group of seventy researchers, lawyers, policymakers and activists, as well as representatives from major technology companies such as Microsoft, Meta and Google.

It says that an open source AI can be used for any reason without the company’s permission, and that researchers should be able to freely see how the system works.

It also says the AI ​​system can be customized for any purpose, including changing its output and sharing it so others can use it, with or without modifications for any reason.

Related

Meta’s Llama 3.1 model is partially open source, by the OSI definition, in that developers and researchers can download and modify it for free.

But Meta doesn’t specify where it gets the data to train Llama 3.1, which could be problematic for users as it could lead to copyright issues or biased data.

Maffulli said when tech companies say where the data comes from, they are often vague and will say the Internet. But he said the “real innovation” and how AI models perform better lies in the way the data sets are passed through the training machines.

“When you talk to companies, they don’t want to release that code,” Maffulli said, adding that “that’s where the innovation happens.”

Related

What are the consequences?

By confusing which AI models are truly open source, Meta and other companies could hinder the long-term development of AI models that are controlled by the user rather than by different technology companies, Maffulli said.

“I fear that society as a whole would be worse off if we just let a handful of companies get on with it and be the only ones who have the edge and access to innovation in this way,” he added.

Euronews Next has contacted us Meta for a response, but did not receive a response at the time of publication.

However, Zuckerberg said in a blog post “we are taking the next steps to help open source AI become the industry standard” and said Llama “has been a foundation of AI innovation worldwide.”

Maffulli said other companies such as Microsoft and Google had pushed back by using the open source term for their models that were not fully open by definition. But according to him, the conversations with Meta did not yield any results.

Related

What’s in it for Meta?

The open source label can have a positive connotation for a technology company’s image, as it is free to use.

But confusion around the term can lead to ‘openwashing’, experts previously told Euronews Next thismeaning they promote open models without contributing to the commons, which can impact innovation and public understanding of AI.

Using the open source term can also impact the bottom line as other companies can use the open source technology which then integrates new innovations into their products.

In a February earnings call, Zuckerberg said: “Open source software often becomes an industry standard, and when companies standardize when building with our stack, it becomes easier to integrate new innovations into our products.”

Related

The open source future

Unlike the 2000s, when social media and Big Tech companies were booming and largely unregulated, Maffulli believes it will be a different story with AI as “regulators are now watching and already regulating.”

Although the OSI is the custodian of the open source AI definition, it does not have strong power to enforce the definition. However, judges and courts around the world are beginning to recognize that the open source definition is important, especially when it comes to mergers but also regulation.

“We expect the definition to have an impact on regulators,” Maffulli said.

‘They’re watching us. We have become credible interlocutors.”