Meta is on track to release a commercial version of its open-source AI model Llama, set to be distributed by Microsoft, and is a free-of-charge alternative to models sold by Google and OpenAI.
The new version has been coined Llama 2, and will be distributed by Microsoft through the Azure Cloud, and run on Windows. Meta have referred to Microsoft as “our preferred partner” for the release.
Llama 2 will become available for direct download and through Amazon Web Services, Hugging Face, along with other providers.
Mark Zuckerberg said, “Open source drives innovation because it enables many more developers to build with new technology. I believe it would unlock more progress if the ecosystem were more open.”
Making this model available and free for businesses threatens the early dominance of companies such as OpenAI.
The first Llama competed with models including ChatGPT and Google Bard, while the new version was trained on 40% more data than the first one, with over 1 million annotations by humans, fine-tuning the quality of outputs.
Chief Executive of Replit, Amjad Masad said, “Commercial Llama could change the picture,” claiming over 80% of projects there use OpenAI’s models.
“Any incremental improvement in open-source models is eating into the market share of closed-source models because you can run them cheaply and have less dependency.”
This announcement comes after plans from Alphabet’s Google and Amazon were made to giver business customer a range of AI models.
Amazon is marketing access to Claude from Anthropic, along with additions to its Titan models, whereas Google plans to make Claude and other models available to cloud customers.
Up until now, Microsoft focused on making technology available from OpenAI through Azure.
A Microsoft spokesperson said the company giving developers a choice, helps extend the position as the go-to cloud platform for AI work.
Meta claims the open-source ecosystem plans to earn revenue off the proprietary technology, which would evaporate if developers could use equally powerful open-source systems freely.
Meta also bets it will benefit from the advancements, bug fixes, and products that could arrive through the model becoming a default for AI innovation.
Zuckerberg claims Meta has more to gain through effective crowd-sourcing ways of reducing infrastructure and maximising creation of new consumer-faced tools.
“Unlike some of the other companies in the space, we’re not selling a cloud computing service where we try to keep the different software infrastructure that we’re building proprietary.”
“For us, it’s way better if the industry standardizes on the basic tools that we’re using and therefore we can benefit from the improvements that others make.”
Releasing Llama comes with risks, by supercharging the ease with unscrupulous actors could build products with little regard to safety controls.
April saw the takedown of a chatbot built by Stanford researchers using the first Llama model after generating unsavory text.
Executives at Meta believe public released of technologies reduce safety risks by harnessing wisdom of the crowd, and identifying problems to build resilience into the systems.
They also claim Meta had put in place an “acceptable use” policy, prohibiting “certain use cases” which could include terrorism, violence, child exploitation and other criminal offences.