Meta’s LLaMA AI aims to be more honest than ChatGPT

Meta logo on screen of mobile phone on Facebook word background. Facebook after rebranding and changing name to Meta.
(Image credit: Viacheslav Lopatin | Shutterstock)

Facebook's parent company Meta has thrown its hat into the AI ring with the initial release of its own AI tool, designed specifically with research in mind. 

In a blog post, Meta explained more about its Large Language Model Meta AI (LLaMA) which aims to “help researchers advance their work in this subfield of AI.” LLaMA will not challenge the likes of Bing with ChatGPT for everyday use but will instead be used under a non-commercial license. 

Rather than being used to create new seasons of canceled Netflix shows, LLaMA will be targeted at the likes of universities, governments, and laboratories for research purposes. Mark Zuckerberg even posted on FaceBook referencing LLaMA's potential to solve math theorems and predict protein structures. 

With the research focus of LLaMA, Meta is determined not to repeat the mistakes of Google’sBard AI and Bing with ChatGPT, which have made some high-profile mishaps. Meta is offering access to LLaMA in several different sizes, with the smallest using 7 billion parameters when generating responses, compared to the largest’s 65 billion parameters. The biggest version has also been trained on 1.4 trillion ‘tokens’ (described as pieces of a word), so has an impressive vocabulary.

Meta has dabbled in the world of AI before with its August 2022 release of BlenderBot a chatbot more in the style of ChatGPT. It also created Galactica, which was designed to write scientific papers, but that was short-lived, lasting only three days before it was decommissioned for inaccuracy. 

One of the most dangerous aspects of AI-driven search is the potential to generate incorrect information that sounds plausible that users then take for gospel. Meta is concerned about this and has put its commitment to using the right methods in writing with its Responsible AI practices. Certainly, by having an approval-only walled garden, LLaMA could avoid much of the corruption that other tools have faced. 

Considering the research focus of LLaMA it’s important that its information is correct. This commitment to reliable information is to be commended, we just didn’t expect it to come from the creators of FaceBook.

Andy Sansom
Trainee Writer

Andy is Tom’s Guide’s Trainee Writer, which means that he currently writes about pretty much everything we cover. He has previously worked in copywriting and content writing both freelance and for a leading business magazine. His interests include gaming, music and sports- particularly Formula One, football and badminton. Andy’s degree is in Creative Writing and he enjoys writing his own screenplays and submitting them to competitions in an attempt to justify three years of studying.