Monday, December 23, 2024

Meta releases open source AI model it says rivals OpenAI, Google tech

Must read

SAN FRANCISCO — Meta released a new artificial intelligence model it says rivals technologies from OpenAI and Google — and is making it free for anyone to use.

The new model, called Llama 3.1, extends Meta’s strategy of making its models open source, meaning anyone can use and modify them without paying the company.

If Meta is successful, it could undermine the business models of its Big Tech rivals and make it easier for start-ups to compete against the likes of OpenAI, while potentially giving fraudsters, state-sponsored hackers and other bad actors access to cutting-edge technology.

Meta released the last version of its AI, Llama 3, just three months ago. But the new release includes a model trained on much more data than the previous version, potentially boosting its capabilities and giving a new tool to companies and organizations that want to use a bigger, more powerful AI model in their products.

“Llama 3 is competitive with the most advanced models,” Meta CEO Mark Zuckerberg said in an open letter Tuesday. “Starting next year, we expect future Llama models to become the most advanced in the industry.”

When OpenAI released ChatGPT in late 2022, it kicked off an arms race among Big Tech companies to build new AI products and get people to pay for them. Microsoft did a multibillion dollar deal with OpenAI to access its tech, while Google created its own AI models and integrated them into its products. Facebook has also spent huge amounts of money on AI, but unlike Microsoft and Google, it does not have a large cloud software business to help it sell that AI to other businesses.

Instead, the company has chosen to make its AI open source, hoping to create an ecosystem where companies that don’t have their own AI tech use Meta’s, giving the company influence over huge swaths of the tech world, similar to how Google’s control of the Android operating system gives it influence over the mobile industry.

Meta’s family of Llama AI models have already been downloaded by companies and individuals 300 million times, said Rob Sherman, vice president of policy and deputy chief privacy officer at Meta.

The company’s open source approach has triggered concerns from some politicians, activists and AI researchers that the tech could be used by America’s geopolitical rivals or by criminals and fraudsters. Other open source AI tools have already been used to create child sexual abuse imagery. But the company has for the past year offered a fierce defense of its approach, and on Tuesday, Zuckerberg doubled down in his letter, saying that open tools can more easily be scrutinized by researchers and regulators than the closed systems built by his rivals.

“Open source will ensure that more people around the world have access to the benefits and opportunities of AI, that power isn’t concentrated in the hands of a small number of companies, and that the technology can be deployed more evenly and safely across society,” he wrote. Meta also provides tools that companies can use to test the safety of their AI systems.

Zuckerberg compared closed AI models to Apple’s practice of enacting rules and charging fees to developers who want to distribute their apps on iPhones, something Meta has had to contend with for years.

“Between the way they tax developers, the arbitrary rules they apply, and all the product innovations they block from shipping, it’s clear that Meta and many others companies would be freed up to build much better services for people” without the phone-maker’s rules, he wrote.

Tuesday’s announcement comes as Meta tries to chart a new future for itself by building out a suite of AI products that it says will change the way people shop and communicate online.

Earlier this year, Meta started to integrate Meta AI across its social media apps, allowing the tool to generate images and answer questions from its users in the search boxes on WhatsApp, Instagram, Facebook and Messenger.

Across the tech industry though, it’s still an open question whether consumers will adopt AI tools into their daily lives. Many high-profile AI launches, like Google’s integration of AI into search results, have resulted in gaffes and failures that force the companies to pull back the product.

“We’re in the phase of this where the main goal is getting many hundreds of millions or billions of people to use Meta AI as a core part of what they do,” Zuckerberg told investors in April. “That’s the kind of next goal, building something that is super valuable.”

Latest article