Meta is releasing a commercial version of its open-source artificial intelligence model Llama, the company said Tuesday, giving startups and other businesses a powerful free alternative to expensive proprietary models sold by OpenAI and Google.
The new version of the model, called Llama 2, will be distributed by Microsoft through its Azure cloud service and will run on the Windows operating system, Meta said in a blog post, referring to Microsoft as “our preferred partner” for the release.
The model, which Meta previously provided only to select academics for research purposes, will also be available via direct download and through Amazon Web Services, Hugging Face and other providers, according to the blog post and a separate Facebook post by Meta CEO Mark Zuckerberg. .
“Open source drives innovation because it allows many more developers to build with new technology,” Zuckerberg wrote. “I think it would unlock more progress if the ecosystem was more open.”
Making a model as sophisticated as Llama widely available and free for companies to build threatens to upend the initial dominance established in the nascent market for generative AI software by players like OpenAI, which Microsoft supports and whose models already offer to commercial customers via Azure. .
The first Llama was already competitive with models powering OpenAI’s ChatGPT and Google’s Bard chatbot, while the new Llama has been trained on 40 percent more data than its predecessor, with more than 1 million annotations from humans to fine-tune the quality of its results, Zuckerberg said.
“Commercial Llama could be a game changer,” said Amjad Masad, chief executive of software development platform Replit, who said more than 80 percent of projects there use OpenAI models.
“Any incremental improvements to open source models are eating up the market share of closed source models because you can run them cheaply and have less dependency,” Masad said.
The announcement follows plans by Microsoft’s biggest cloud rivals Google and Alphabet’s Amazon to give business customers a range of AI models to choose from.
Amazon, for example, is marketing access to Claude, an AI from high-profile startup Anthropic, as well as its own family of Titan models. Google, likewise, has said it plans to make Claude and other models available to its cloud customers.
Until now, Microsoft has focused on making the technology available from OpenAI on Azure.
When asked why Microsoft would support an offering that could degrade the value of OpenAI, a Microsoft spokesperson said that giving developers choice in the types of models they use would help extend its position as the cloud platform for AI work.
For Meta, a burgeoning open-source ecosystem of AI technology built with its models could hamper rivals’ plans to earn revenue from its proprietary technology, the value of which would evaporate if developers could use equally powerful open-source systems for free.
A leaked internal Google memo titled “We don’t have a moat, and neither does OpenAI” lit up the tech world in May after forecasting such a scenario.
Meta is also betting that it will benefit from advancements, bug fixes, and products that can grow from its model and become the default for AI innovation, as it has done in recent years with its widely adopted open source AI framework. PyTorch.
As a social media company, Zuckerberg told investors in April that Meta has more to gain by crowdsourcing effective ways to reduce infrastructure costs and maximize the creation of new consumer-facing tools that could drive people to its ad-supported services. charging for access to their models.
“Unlike other companies in the space, we’re not selling a cloud computing service where we’re trying to maintain ownership of the different software infrastructures we’re building,” Zuckerberg said.
“For us, it’s much better if the industry standardizes the basic tools we’re using and therefore we can benefit from the improvements that others make.”
However, releasing Llama into the wild also carries risks, as it increases the ease with which unscrupulous actors can build products without regard to security checks.
In April, Stanford researchers removed a chatbot they had built for $600 using a version of the early Llama model after it generated nasty text.
Meta executives say they believe that public releases of technologies actually reduce security risks by leveraging the wisdom of the crowd to identify problems and build resilience in systems.
The company also says it has implemented an “acceptable use” policy for Commercial Llama that prohibits “certain use cases,” including violence, terrorism, child exploitation, and other criminal activity.
© Thomson Reuters 2023
(This story has not been edited by NDTV staff and is automatically generated from a syndicated feed.)