Red Hat has released Red Hat Enterprise Linux AI (RHEL AI), a platform for simplifying the development, testing, and deployment of enterprise generative AI models. RHEL AI combines IBM Research's open-source Granite language model with InstructLab, a model alignment tool based on the LAB (Large-scale Alignment for Chatbots) methodology. This combination simplifies the complex process of adjusting AI models to meet specific business needs.

IBM states that "open-source AI is broken." And they are right. While open-source in software promotes collaborative improvement, continuous refinement, and collective ownership, this has not effectively happened in AI. When new open-source models are released, there are thousands of forks, but they rarely merge back into the base model, hindering meaningful community improvements.

For example, Llama 3 was released about three weeks ago, and now there are over 6,000 Llama 3 models on Hugging Face. Each forked model has been improved in different ways, but the original Llama 3 model remains unchanged and does not benefit from contributions from the community.

InstructLab addresses this issue by allowing contributors to add specific skills or knowledge to the model. Its model-agnostic technology enables upstream creators to periodically update their open-source models by integrating new skills instead of retraining the entire model, promoting efficient and collaborative development. This approach harnesses the collective efforts of global contributors, leveraging their diverse expertise and perspectives to enhance the model.

InstructLab utilizes LAB technology, which aligns models through taxonomy-guided synthetic data generation and a multi-stage fine-tuning framework. This makes AI model development more approachable by reducing reliance on expensive human annotations and proprietary models. By specifying skills and knowledge associated with taxonomy, InstructLab generates synthetic data at scale from this information and uses the generated data for model training, significantly improving model performance.

RHEL AI combines InstructLab's capabilities with IBM Research's open-source Granite language model, packaged as an optimized, bootable RHEL image that can be deployed in hybrid cloud environments. It is also included in Red Hat's hybrid machine learning operating platform, OpenShift AI, allowing AI workloads to run wherever the data resides, whether in data centers, public clouds, or at the edge.

RHEL AI is now available as a developer preview. IBM Cloud will add support for RHEL AI and OpenShift AI, enabling enterprises to easily deploy generative AI into mission-critical applications. Industry leaders such as AMD, Cisco, Dell, HPE, Intel, Lenovo, and NVIDIA have expressed their support for Red Hat's efforts to drive enterprise AI adoption through RHEL AI.