On Friday, Facebook co-founder Mark Zuckerberg announced Meta Platforms’ upcoming release to researchers of a new large language model called LLaMA (Large Language Model Meta AI). The model, developed by Meta’s Fundamental AI Research (FAIR) team, is intended to help scientists and engineers explore AI applications and functions, such as question answering and document summarization.
LLaMA’s launch comes as technology companies scramble to advance advances in artificial intelligence techniques and integrate the technology into their commercial products. As CNBC notes, Meta’s release stands apart from competitor models as it will be available in a choice of sizes, from 7 billion parameters to 65 billion parameters. In addition, Zuckerberg said his company’s new LLM technology — which could eventually solve math problems and conduct scientific research — will be available to the research community, and Meta is now accepting applications for access. This is a change from Google’s underlying LaMDA and ChatGPT models, which are not publicly available.
Reuters points out that Meta is engaged in an increasingly intense race to dominate AI technology, which began in earnest in late 2022 with OpenAI’s ChatGPT. As for Meta, LLaMA’s release also represents its commitment to open science—hence the choice to publicly release its state-of-the-art large language fundamental model, while providing researchers with an open resource to advance their work. Meta believes that, unlike more fine-tuned purpose-built models, theirs will prove versatile, with multi-purpose cases.
Another way LLaMA is different, according to Meta: It requires “much less” computing power than previous offerings and is trained in 20 languages, focusing on those based on the Latin and Cyrillic alphabets. With its 13 billion parameters, LLaMA should outperform GPT-3, the model on which ChatGPT is built. Meta also attributed LLaMA’s performance to “cleaner” data and “architectural improvements” to the model that improved training stability.
To maintain the integrity of the model and prevent misuse, Meta will release it under a non-commercial license focused on research use cases. Academic researchers, government, civil society, academic institutions and industry research laboratories will have access on a case-by-case basis.
Meta’s release of LLaMA may mark a major development in AI language models. The social media giant’s commitment to open science and allow researchers to study under a non-commercial license will limit misuse of the model.
LLaMA’s flexibility and problem-solving potential may provide a glimpse of the substantial potential benefits of artificial intelligence to billions of people at scale.