Meta Unveils New Language Model ‘LLaMA,’ Challenges OpenAI's ChatGPT

Meta has joined the race in developing artificial intelligence (AI) chatbots with its very own LLaMA (Large Language Model Meta AI) model, challenging the likes of OpenAI’s ChatGPT, Google’s Bard, Microsoft’s Bing Chat and Baidu’s Ernie Bot.  

Meta CEO Mark Zuckerberg announced its launch in a Facebook post. “LLMs have shown a lot of promise in generating text, having conversations, summarizing written material, and more complicated tasks like solving math theorems or predicting protein structures,” he explained.  

In a Meta AI blog post, the company describes LLaMA as a “smaller, more performant” model, needing less processing power and resources than its competitors. 

“Training smaller foundation models like LLaMA is desirable in the large language model space because it requires far less computing power and resources to test new approaches, validate others’ work, and explore new use cases. Foundation models train on a large set of unlabeled data, which makes them ideal for fine-tuning for a variety of tasks,” the post wrote. 

Meta will make LLaMA available at 7B, 13B, 33B, and 65B parameters, while also sharing a LLaMA model card explaining how the model was built in with their responsible AI practices.  

Full research access with regards to large language models are still limited due to the resources it would require to train them, but smaller models like LLaMA are trained on more tokens, making them easier to fine-tune for more specific cases. “We trained LLaMA 65B and LLaMA 33B on 1.4 trillion tokens. Our smallest model, LLaMA 7B, is trained on one trillion tokens,” the post explained. 

Much like its competitors, Meta has also shared disclaimers with regards to its new AI chatbot, claiming that more research needs to be done to address the risks of bias, toxic comments and hallucination with models like LLaMA. “Like other models, LLaMA shares these challenges,” the post wrote. 

While services such as ChatGPT have been available to all users upon release, it is still unknown when Meta plans to roll out its LLaMA model to the general public.  

According to the company, limited access to LLaMA will “maintain its integrity and prevent misuse.” For now, access to Meta’s new AI chatbot will only be granted to academic researchers, and government, civil society and academic organizations on a case-by-case basis.