Meta announces Purple Llama initiative to promote responsible and safe generative AI development

Meta Platforms Inc. has launched a new initiative, Purple Llama, to promote the responsible and safe development of generative AI. The initiative will provide tools and evaluations for the AI development community to build open generative AI models safely. The project will start with the release of a free and open set of cybersecurity evaluation benchmarks for Large Language Models (LLMs), called CyberSec Eval, and Llama Guard, a safety classifier for input/output filtering.

As part of the Purple Llama initiative, Meta will collaborate with partners including AI Alliance, Advanced Micro Devices Inc., Amazon Web Services Inc., and many others. The components of the Purple Llama project, including CyberSec Eval and Llama Guard, will be licensed on a permissive basis for both research and commercial use. The first of these components will be showcased at the NeurIPs 2023 event.

Key takeaways:

Meta Platforms Inc. has announced a new initiative called Purple Llama to promote the responsible development of artificial intelligence systems.
The initiative will provide tools and evaluations for the AI development community to build open generative AI models in a safe and responsible way.
As part of the initiative, Meta is releasing a set of cybersecurity evaluation benchmarks for Large Language Models (LLMs), called CyberSec Eval, and a safety classifier for input/output filtering, called Llama Guard.
Purple Llama is a collaborative effort with partners including the AI Alliance, Advanced Micro Devices Inc., Amazon Web Services Inc., Google Cloud, IBM Corp., Microsoft Corp., and many others.

Meta announces Purple Llama initiative to promote responsible and safe generative AI development - SiliconANGLE

Key takeaways:

Comments (0)

Newsletter