Meta Announces Imminent Release of Llama 3, Next-Gen Language Model

During an event held in London on Tuesday, Meta, the parent company of Facebook, revealed its plans to unveil Llama 3, the next iteration of its large language model designed to power generative AI assistants, within the next month.

This confirmation follows a report by The Information published on Monday, indicating that Meta was nearing the launch phase of Llama 3.

Nick Clegg, Meta’s president of global affairs, provided insights into the imminent release, stating, “Within the next month, actually less, hopefully in a very short period of time, we hope to start rolling out our new suite of next-generation foundation models, Llama 3.” He hinted at the introduction of various iterations or versions of the product, each with distinct capabilities and versatility, set to be released throughout the year.

Chris Cox, Meta’s Chief Product Officer, emphasized the strategic integration of Llama 3 across multiple Meta products.

Meta has been striving to keep pace with OpenAI, which caught the tech industry off guard with the launch of ChatGPT over a year ago. The widespread popularity of ChatGPT underscored the mainstream appeal of generative AI, prompting Meta to intensify its efforts in this domain.

While Meta has adopted a cautious approach towards AI development, its previous Llama iterations faced criticism for their perceived limitations. Llama 3 aims to address these concerns by expanding its scope, enabling more accurate responses to queries and accommodating a broader spectrum of topics, including potentially controversial subjects.

Joelle Pineau, Vice President of AI Research at Meta, articulated the company’s ambition for Llama-powered Meta AI to become the most useful assistant globally. However, Meta refrained from disclosing the parameters of Llama 3 or providing demonstrations of its functionality. It is anticipated that Llama 3 will boast around 140 billion parameters, doubling the size of its predecessor, Llama 2.

Significantly, Meta’s approach with the Llama series, built as open-source products, diverges from the proprietary model favoured by some competitors. By fostering developer engagement, Meta aims to garner broader support for its AI initiatives.

Despite Meta’s strides in text generation, its cautious stance extends to other domains of generative AI. Meta has opted not to release Emu, its image generation tool, citing concerns regarding latency, safety, and user experience.

Interestingly, despite Meta’s push for Llama 3, scepticism persists among some prominent figures within the company. Yann LeCun, Meta’s Chief AI Scientist, expressed reservations about the limitations of generative AI and advocated for alternative approaches, such as joint embedding predicting architecture (JEPA). LeCun’s remarks underscore the ongoing debate within Meta regarding the future trajectory of AI development.

As Meta prepares for the rollout of Llama 3, the company navigates a landscape shaped by both anticipation and scepticism, highlighting the complexities inherent in advancing generative AI technology.