Risk Category- Cyber Risk
Meta has launched Llama 3.1, its largest and most advanced open-source Artificial Intelligence (AI) model to date, boasting 405 billion parameters. This model aims to rival the closed-source models from industry leaders such as OpenAI and Anthropic. Mark Zuckerberg emphasised the importance of open-source AI, drawing a parallel to the transformative impact of Linux in operating systems and cloud computing. Meta’s Llama 3.1 is designed to allow developers to use and train their models independently, fostering innovation and reducing reliance on proprietary systems. The model is also cost-effective, with inference costs estimated to be about half of those for closed-source models like OpenAI’s GPT-4.
Analyst Comments
- By making Llama 3.1 open-source, Meta is lowering the barrier for researchers and developers worldwide, facilitating broader experimentation and innovation in AI.
- The model’s design focuses on cost efficiency, potentially offering significant savings compared to closed models. This could make advanced AI technologies accessible to smaller organisations and independent developers.
- Open-source AI fosters transparency, allowing the global AI community to scrutinise, improve, and build upon the model, accelerating advancements and ensuring more robust and reliable AI systems.
- The large size of Llama 3.1 brings challenges, such as high computational resource requirements and increased energy consumption. This could pose sustainability issues and increase operational costs.
- Meta’s open-source strategy could position it as a leader in AI innovation, promoting a more collaborative ecosystem and potentially outpacing competitors reliant on proprietary models. Currently, Google and ChatGPT, with its AI models, are dominating the industry. Llama 3.1 poses serious competition for them. They may feel pressured to adopt open-source strategies to keep up with Meta’s transparency and cost advantages, which could disrupt their current business models.
- The open nature of Llama 3.1 also brings risks, such as misuse of technology and difficulties in controlling the distribution and application of powerful AI tools.
