top of page

Meta to Introduce the New Generative AI Model Lama 3 in May

Meta, the parent company of Facebook, is gearing up to introduce a revamped iteration of its robust language model, Llama 3, designed to empower AI assistants. As reported by The Information, Meta's forthcoming model aims to enhance user interaction through advanced question responses and task execution capabilities.

Nick Clegg, Meta's president of global affairs, unveiled the plan during an event in London, highlighting the company's intention to unveil multiple iterations of Llama 3 with varying functionalities throughout the year.

Acknowledging the strides made by OpenAI's ChatGPT AI, which captivated users with its conversational prowess over a year ago, Meta seeks to address past criticisms of its cautious approach to AI development. Previous iterations of Llama fell short of expectations due to their limited functionality.

With Llama 3, Meta endeavors to deliver a vastly improved experience, boasting enhanced accuracy in question responses and broader topical coverage, including contentious subjects, in a bid to enhance user engagement.

While aspiring for Llama-powered AI to achieve global preeminence, Meta concedes that considerable refinement is still required. However, specific details regarding Llama 3's capabilities remain scant, with no demonstrations offered thus far.

Diverging from conventional practices, Meta adopts an open-source approach with Llama, granting developers access to and the ability to modify its underlying code. This decision could potentially boost Llama's appeal among developers compared to proprietary models.

Despite its commitment to advancing conversational AI, Meta exercises prudence in other AI domains, such as image generation. The company opts to withhold its image generation tool, Emu, until it meets stringent safety and usability standards.

Interestingly, internal skepticism toward generative AI exists within Meta, with Yann LeCun, the company's chief AI scientist, advocating for an alternative approach dubbed JEPA. This methodology prioritizes outcome prediction over response generation, reflecting Meta's diversified strategy in AI development.


bottom of page