Meta announces the launch of Llama 2, the biggest AI release since ChatGPT

Meta made a significant announcement at the Microsoft Inspire conference, débuting its latest AI model, LLaMA 2 (Large Language Model Meta AI). Unlike the previous LLaMA, this new large language model is open-source and commercially available. In other words, LLaMA is entirely free for enterprises and developers - a move that has the potential to reshape the landscape of generative AI, providing businesses with more options to embrace AI technologies without cost restrictions.

The first version of LLaMA was initially available only for academics and researchers under a research license. However, model weights were leaked, leading to controversy and a government inquiry. With LLaMA 2, Meta is addressing these concerns and introducing a more robust model that is usable on a broader scale and may have a far-reaching impact on the field of large language models.

LLaMA 2 is now available on Microsoft Azure, a significant development since Azure is also the primary platform for OpenAI and its GPT-3/GPT-4 models. Microsoft's investment in Meta and OpenAI underscores its commitment to hedge its bets in the AI market.

Meta's founder and CEO, Mark Zuckerberg, expressed enthusiasm about making LLaMA open-source, highlighting that open-source initiatives drive innovation and allow more developers to build with new technology. Yann LeCun, VP and chief AI scientist at Meta, also praised the open-source release, calling it a game-changer for the large language model market.

LLaMA 2 is a transformer-based auto-regressive language model with multiple model sizes, including 7, 13, and 70 billion parameters. It has been trained on a massive dataset 40% larger than the one used for LLaMA 1, with an expanded context length of two trillion tokens, double that of LLaMA 1. The benchmarks provided by Meta suggest that LLaMA 2 outperforms its predecessor.

Safety is a critical aspect of LLaMA 2, with a multi-step process to ensure responsible AI development. The model undergoes supervised fine-tuning and reinforcement learning from human feedback to enhance safety and limit potential bias. Meta's research paper on LLaMA 2 provides in-depth details about the safety measures implemented to increase transparency and address potential downstream issues.

The availability of LLaMA 2 on Microsoft Azure and other platforms is expected to drive innovation in the AI market and encourage responsible AI development practices. Businesses now have the option to leverage this powerful language model for various AI tasks without facing financial constraints. The move towards open-source AI models also fosters collaboration and knowledge-sharing among developers, leading to faster advancements in AI.

As the AI landscape continues to evolve, introducing LLaMA 2 gives enterprises more choices when adopting AI technologies. It signals a shift towards greater accessibility and democratisation of AI, enabling more businesses to harness the potential of large language models in their applications and services.