Meta to release commercial AI model in effort to catch rivals

Meta is ready to release a commercial version of its artificial intelligence model, which allows start-ups and businesses to build custom software on top of the technology.

The move will allow Meta to compete with Microsoft-backed OpenAI and Google, which are leading the race to develop generative AI. The software, which generates text, images and code, is powered by large language models (LLMs) that are trained on large amounts of data and require large amounts of computing power.

Meta released its own language model, known as LLaMA, to researchers and academics earlier this year, but the new version will be more widely available and customizable by companies, three people familiar with the matter said. plans said. A release is expected soon, one of the people said.

Meta says its LLMs are “open-source”, which means the details of the new model will be released to the public. This is in contrast to the approach of competitors such as OpenAI, whose latest model GPT-4 is a so-called black box where the data and code used to build the model cannot be used by third parties.

“The competitive AI landscape will change completely in the coming months, in the coming weeks maybe, when there are open source platforms that are actually as good as those that are not,” vice-president and Meta’s chief AI scientist, Yann LeCun, said at a conference in Aix-en-Provence on Saturday.

Meta’s upcoming release comes as a race among tech groups in Silicon Valley to establish themselves as hot participants in AI.

Writing in the Financial Times this week, Meta’s head of global operations Nick Clegg extolled the virtues of an open source approach, saying that “openness is the best antidote to the fears that revolves around AI”. But the move will also help Meta in its attempts to catch up with rivals, as an open model will allow companies of all sizes to improve the technology and build applications on it.

Meta has been working on AI research and development for more than a decade but appears to be lagging behind after OpenAI’s ChatGPT, a conversational chatbot, was released in November, prompting other Big Tech to launch similar products.

“The goal is to reduce the current dominance of OpenAI,” said a person with knowledge of Meta’s high-level strategy.

Meta declined to comment.

While Meta’s technology is open source and currently free, two people familiar with the matter said the company is exploring charging business customers for the ability to tailor the model to their needs through using their own proprietary data. One person said that there are no current plans to charge and Meta will not do so in future releases.

Joelle Pineau, Meta’s vice-president of AI research, declined to comment on the development of a new AI model and how it will fare but said: “At the end of the day, because you release one thing. [open source]you cannot completely give up the intellectual property of that work.”

“We are not shy about the fact that we want to use these models [in our] product,” he added.

In 2021, chief executive Mark Zuckerberg announced a pivot to build an avatar-filled digital world known as a metaverse and spent more than $10bn a year on the project. That expensive ambition has proven unpopular with investors and Meta has recently raced to increase investment in AI.

Earlier this year, the social networking giant established a generative AI unit led by chief product officer Chris Cox. Pineau said that Cox’s team is mainly in the field of AI research but also product development, because it “creates completely new businesses”.

Zuckerberg and other executives proposed to push to create more AI chatbots for individuals, advertisers and businesses on Meta platforms Instagram, WhatsApp and Facebook, which are run by its LLMs.

The benefit of open source models includes a higher uptake of users who then input more data to be processed by AI. The more data LLM has, the more powerful its capabilities become.

In addition, open source models allow researchers and developers to find and fix bugs, simultaneously improving technology and security – at a time when technology companies like Meta are facing years of investigating various privacy scandals and misinformation.

While providing software for free may seem antithetical to making money, experts believe that corporations can also use this strategy to capture new markets.

“Meta realizes they’re behind the current AI hype cycle, and this gives them a way to open up the ecosystem and look like they’re doing the right thing, being generous and giving back to the community,” said one person. who is familiar with the company’s thinking.

However, there are clear risks with open source AI, which can be shaped and abused by bad actors. Child safety groups report an increase in AI-generated child sexual abuse images online, for example.

The researchers also found that a previous Meta AI model, BlenderBot 2, released in 2021, spread false information. Meta says it has made BlenderBot 3 more resistant to this content, although users still find it generates false information.

There are also regulatory and legal risks regarding intellectual property and copyright. On Monday, comedian and actor Sarah Silverman filed a lawsuit against Meta and OpenAI for claiming her work was used to train models without her consent.

Meta released the open source LLaMA model to researchers in February. A month later, it leaked more widely through the online forum 4chan, prompting developers to build on top of it in violation of Meta’s licensing rules, which specify that it cannot be used in commercial products.

“This model is out there in ways we wish it wasn’t,” Pineau said.

Other AI companies, such as French start-up Mistral, are also exploring the potential of releasing open source versions of their technology. OpenAI, which has released open source AI models for speech and image recognition in the past, said its team is looking at developing an open source LLM, as long as they can reduce the risk of misuse below the minimum threshold.

“We have a choice between deciding that artificial intelligence is too dangerous a technology to keep open and keeping it under lock and key and in the hands of a small number of companies that will control it,” Meta’s AI chief LeCun said. “Or, on the contrary, the open source platforms that ask for contributions . . . from all over the world.”

Additional reporting by Tim Bradshaw