During an event held in London, Meta unveiled its plans to launch Llama 3, the upcoming iteration of its advanced large language model designed to power generative AI assistants, within the next month.
The confirmation of this release aligns with recent reports indicating Meta’s imminent launch, as reported by The Information just a day earlier. Nick Clegg, Meta’s president of global affairs, disclosed that the company aims to introduce Llama 3 in various iterations, each offering distinct capabilities and versatility throughout the year.
Chris Cox, Meta’s Chief Product Officer, elaborated on the strategy to integrate Llama 3 across multiple Meta products, signaling a comprehensive deployment strategy within the Meta ecosystem.
Meta’s pursuit to catch up with industry competitors like OpenAI, which made significant strides with the viral success of ChatGPT, underscores its urgency in advancing AI technology. Previous iterations of Meta’s Llama model faced criticism for their limited scope, prompting the development of Llama 3 with enhanced capabilities to address a broader range of inquiries, including more contentious topics.
Joelle Pineau, Meta’s Vice President of AI Research, expressed the company’s ambition to establish a Llama-powered Meta AI as the foremost assistant globally, acknowledging the work required to achieve this goal. However, Meta refrained from divulging specific details regarding Llama 3’s parameters or providing demonstrations of its functionality, although it is anticipated to boast approximately 140 billion parameters, doubling its predecessor’s capacity.
Notably, Meta’s decision to develop Llama as an open-source project reflects its commitment to fostering collaboration within the AI community, distinguishing it from more proprietary approaches adopted by competitors.
While Meta progresses with the launch of Llama 3, it remains cautious in its approach to other generative AI technologies. Despite ongoing development efforts, Meta has yet to release Emu, its image generation tool, citing considerations related to latency, safety, and usability.
Interestingly, Meta’s stance on generative AI has attracted skepticism from within its ranks, notably from Yann LeCun, Meta’s chief AI scientist. LeCun advocates for alternative approaches like joint embedding predicting architecture (JEPA), which Meta has been exploring for more accurate predictive AI in image generation, signaling potential shifts in the company’s AI strategy.
In light of these developments, Meta’s foray into the next generation of AI technology promises to shape the future landscape of digital assistants and generative AI applications.