Facebook will make its latest Llama AI model free to use

The announcement shows Facebook’s increasing focus on making its AI openly accessible, in contrast to companies like OpenAI and Google, which have chosen to keep the details of how their technology works private and charge for access to it. It also highlights that Facebook doesn’t have a large cloud computing business through which it can sell software directly to business customers like Microsoft and Google do.

Open access to a powerful language model could allow all kinds of startups and established companies to build their own AI tools instead of having to rely on contracts with Microsoft, Google and OpenAI.

“Llama 2” could be a “game changer,” Matt Bornstein, a partner at venture capital firm Andreessen Horowitz, said on Twitter. The model’s capabilities rival newer versions of OpenAI tools, he added.

As Meta tries to keep up with its competitors and develop new services based on artificial intelligence, the company is increasingly promoting the idea that this technology should be open source – meaning that the underlying code is available to anyone. Meta argues that by making the technology freely available, developers will be better able to create new and exciting AI products, as well as new tools to protect the public.

“Open source drives innovation because it allows many more developers to build with new technology,” Meta CEO Mark Zuckerberg said in a Facebook post on Tuesday. “It also improves security because when software is open, more people can examine it to identify and fix potential problems. I believe there would be more progress if the ecosystem was more open, which is why we are making Llama 2 open source.”

But critics say open-source AI models could lead to misuse of the technology. Earlier this year, Meta shared Llama with a select group of researchers, only for the model to be leaked and later used for applications ranging from drug research to sexually explicit chatbots. Last month, Senator Richard Blumenthal (D-Conn.) and Josh Hawley (R-Mo.) wrote to Zuckerberg in June, arguing that in the short time that they have become more widely available, applications of generative artificial intelligence are already serving problematic purposes content has been misused, from pornographic deep fakes by real people to malware and phishing campaigns.

“Meta’s decision to distribute LLaMA in such an unrestricted and permissive manner raises important and complicated questions about when and how it is appropriate to openly release advanced AI models,” the senators wrote.

Meta said Tuesday that its latest AI model went through “red teaming” exercises, where human testers try to get it to make mistakes or produce objectionable content, and then train it to avoid such responses . The company also asks potential users to promise not to use it to promote terrorism, create child sexual abuse material, or discriminate against people.

Microsoft CEO Satya Nadella also mentioned Facebook’s partnership to spread AI through its cloud business at a company event on Tuesday. Nadella announced a version of its Bing chatbot that would allow business customers to ask the bot questions about their company’s internal data and use it more fluidly at work.

Meta, which has fallen out of the rankings of the world’s most valuable tech companies in recent years, is pushing to show it can lead the generative AI boom surrounding the new generation of chatbots and image generators. In recent months, Zuckerberg and other executives have touted the company’s investments in AI research and computing infrastructure, as well as new products like an in-house productivity assistant, a generative AI-based advertising product, and a new photo generation tool.

The AI ​​announcements follow months of sluggish financial results and a variety of challenges facing Meta’s business. Apple’s new privacy regulations, rising inflation and a slump in post-pandemic e-commerce market growth impacted the company’s digital advertising business. In the last six months, Meta has laid off more than 20,000 employees in order to downsize and make it more efficient. Still, stock prices have risen sharply this year as the company seeks to tighten its belt.

Meta has also been vocal in defending itself against scenarios being proposed by a growing number of prominent AI leaders, including Elon Musk and Google’s chief AI researcher Demis Hassabis, who say technology is advancing so rapidly that it’s putting human intelligence within of 10 years could surpass.

Nick Clegg, President of Meta Global Affairs, urged regulators not to fear doomsday scenarios and to act against AI models as soon as possible. He argued that some of the potential “existential threats” raised by critics are only hypothetical and still a long way off. Instead, Clegg has argued that AI should be regulated with an emphasis on keeping the technology open and available.

“Nobody thinks about the kind of models we see [with] Llama one or llama version two even knock on the door of this kind of high performance models [AI models] That could require special government licensing,” Clegg told The Washington Post earlier this month.