In a recent blog post, Meta has unveiled its new large language model Llama 2, a rival for OpenAI’s ChatGPT and Google’s Bard.
Llama 2 has been created in partnership with Microsoft and will be free to use for research and commercial purposes, something its competitors do not currently offer.
The news confirms that Meta is well and truly throwing its weight into researching and developing AI technologies. Last month the company shared a preview of its first generative AI tools for Facebook, Instagram and WhatsApp.
Meta's New AI Model Will Be Open Source
Llama 1 was first announced back in February and received more than 100,000 access requests from researchers. Since then, Meta and Microsoft have worked on developing “an open ecosystem for interchangeable AI frameworks.”
Llama 2 has been trained on 40% more data than its predecessor, with more than 1 million annotations by humans to fine-tune its output. More than this, Meta seems excited to announce that Llama 2 is open source.
According to Meta’s blog post unveiling the AI technology, they consider an open approach to be the “right one for the development of today’s AI models, especially those in the generative space where the technology is rapidly advancing.”
As well as keeping up with the pace of development, Meta are hoping an open source approach will foster more software security too.
“It also improves safety and security because when software is open, more people can scrutinize it to identify and fix potential issues. I believe it would unlock more progress if the ecosystem were more open, which is why we’re open sourcing Llama 2.” – Meta CEO, Mark Zuckerberg
In addition to its Microsoft partnership, Llama 2 will also be available through Amazon Web Services, Hugging Face and other providers.
Transparency Considered A Priority
Meta has made it clear that “building responsibly” is a key priority when it comes to its AI systems. As well as acknowledging the risks of such technology, Meta set out several key areas of focus to ensure its development remains as responsible and transparent as possible.
These include red-teaming exercises using both internal and external efforts, creating and publishing a responsible use guide, and putting a policy in place to ensure the models are being used fairly.
The company has also released a research paper that shares all known challenges and issues in an effort to provide insight into the mitigations and resolutions it will make.
Meta confirmed that all models have been tested internally and are also due to undergo external adversarial testing by third parties.
More Chatbot Choice Could Lead To Better Future Developments
Although ChatGPT has dominated the AI-powered chatbot scene, with over 100 million people reported to have conversed with it, the rise of its rivals has really kicked up a gear in recent weeks.
From the release of “safe and steerable” Claude 2 to Amazon’s cloud-based Bedrock, the choice of generative language learning models has widened for both consumers and developers.
This current introduction of open-source models is particularly significant however, as they’re cheaper to run and require less dependency. It could lead to them encroaching into the market share of closed-source options, forcing companies like OpenAI with no option than to respond with its own open-source alternative. Now's the time to watch this space.