Meta Announces Llama 2 and Two New Partnerships

Meta announced a new version of Llama (Large Language Model Meta AI), Llama 2, and made it open source. Meta believes an open source approach fosters collaboration, iteration and transparency, which is important for new technologies like large language models. Providing access to tools and computing power can create opportunities for businesses, startups, and researchers. Llama 2 was trained on 40% more data than the original Llama model.

Meta and Microsoft

Meta and Microsoft announced that Microsoft is a preferred partner for providing Llama 2 access through its Azure cloud platform. This allows developers to build generative AI tools using Llama 2. Azure customers can now fine-tune and deploy the 7B, 13B, and 70B Llama 2 models on Azure. This leverages Azure's capabilities for training, deploying, and securing AI models. Windows developers can also use Llama 2 locally on Windows through ONNX Runtime and DirectML. This is an expansion of the existing partnership between Microsoft and Meta on AI. Meta uses Azure for cloud computing and chose Azure as the launch platform for Llama 2. Llama 2 expands Azure's catalog of foundation models for developers. The catalog simplifies discovering, customizing, and deploying models. Windows developers can also fine tune Llama 2 models using Windows Subsystem for Linux. Azure enables testing Llama 2 for specific use cases. Techniques like prompt engineering can optimize Llama 2 for safer, more reliable experiences. Azure AI Content Safety provides another layer of protection for Llama 2 deployments.

Meta and Qualcomm

Qualcomm is partnering with Meta to bring Llama 2 to more devices like smartphones and PCs starting in 2024. This will integrate the AI capabilities directly into devices with Qualcomm chips rather than relying solely on the cloud. On-device AI can help developers save on cloud computing costs and make Llama 2 more reliable by reducing dependency on an internet connection. Qualcomm believes scaling generative AI requires both cloud and on-device support.

Making Llama 2 open source allows wider testing and feedback to improve the AI. The partnerships with Microsoft and Qualcomm aim to make Llama 2 accessible to more developers and researchers. Broad availability of Llama 2 through cloud services and integration into devices should help advance research into generative AI and enable new applications using this technology. Responsible openness is key to developing large language models safely.