Meta has confirmed the upcoming release of Llama 3, the next generation of its large language model, at an event in London. The company plans to roll out different iterations of Llama 3 with various capabilities over the course of the year. This move is aimed at catching up to OpenAI, which launched ChatGPT over a year ago. Llama 3 is expected to address previous limitations by answering questions more accurately and handling a wider range of topics. With about 140 billion parameters, it is twice the size of Llama 2. Meta’s open source approach with the Llama family is intended to appeal to developers. However, the company is proceeding cautiously, as it has not yet released Emu, its image generation tool. Yann LeCun, Metas chief AI scientist, expressed skepticism about generative AI and emphasized the potential of joint embedding predicting architecture (JEPA) as the future of AI.
