At Red Hat Summit 2024 In Denver, Colorado, the leading open source software provider announced major latest initiatives to harness the facility of generative AI in enterprises.
The principal actors are Red Hat Enterprise Linux AI (RHEL AI), a base model platform for developing and running open source language models, and InstructLab, a community project that allows subject material experts to enhance AI models with their knowledge.
How Red Hat differs from other corporations integrating and offering open source AI
According to Matt Hicks, CEO of Red Hat, RHEL AI differs from the competition in a number of key ways.
First, Red Hat focuses on open source and a hybrid approach. “We consider that AI isn’t really different from applications. That you could have to coach them in some places and operate them somewhere else. And we’re neutral about this hardware infrastructure. We need to run in every single place,” Hicks said.
Second, Red Hat has a proven track record of optimizing performance across different hardware stacks. “We have long shown that we will get the very best out of the hardware stacks beneath us. We don't produce GPUs. I can run Nvidia as fast as possible. I can run AMD as fast as possible. I can do the identical thing with Intel and Gaudi,” explained Hicks.
This ability to maximise performance across different hardware options while providing location and hardware options is sort of unique out there.
Finally, Red Hat's open source approach means customers retain ownership of their mental property. “It’s still your IP. We provide this service and subscription business, and also you don’t quit your mental property to work on it with us,” Hicks said.
In the fast-moving AI market, Red Hat believes this mixture of open source, hybrid flexibility, hardware optimization and customer IP ownership will prove to be a key differentiator for RHEL AI.
“We're expanding the flexibility to deploy and run these models at scale,” said Ashesh Badani, senior vp and chief product officer at Red Hat, during a Q&A with reporters and analysts after the keynote in Denver: “Whether they arrive from ours .” Partnership with IBM Research or, for instance, something customers could do with their very own proprietary models.”
A brand new platform is emerging: RHEL AI
RHEL AI combines open source language models comparable to the Granite family of models developed by IBM Research with tools from the InstructLab project to enable customization and improvement of the models.
It offers an optimized RHEL operating system image with hardware acceleration support and enterprise technical support from Red Hat.
“We are in search of to increase the investments our customers have already made in infrastructure-supporting applications to this latest critical workload support through enterprise AI, predictive analytics and generative AI,” said Chris Wright, chief technology officer and senior vp, global engineering at Red Hat.
Red Hat's goal is to deliver the identical reliability and trust customers expect on a single, unified platform. You'll give attention to improving today's hybrid cloud infrastructure while advancing the present state of app development and deployment in cloud-native environments.
“It's really exciting because we're taking numerous what our customers already know and expanding on it so that you don't need to learn all the pieces, you simply need to learn all the pieces latest,” Wright added.
InstructLab enhances LLMs with synthetic training data generated from your organization's examples
The InstructLab project, also presented on the summit, goals to enable subject material experts without data science backgrounds to enhance language models by contributing their knowledge. It uses a novel method called LAB (Large-scale Alignment for chatBots), developed by IBM Research, to generate high-quality synthetic training data from a small variety of examples.
The LAB method consists of 4 easy steps. First, experts give examples of their knowledge and skills. Next, a “teacher” AI model looks at these examples to create numerous similar training data.
This synthetic data is then checked for quality. Finally, the language model learns from the approved synthetic data. This allows the community to repeatedly improve the models by sharing their knowledge. This is a cheap approach to make AI significantly smarter using a small variety of human examples.
This allows models to be continually improved and refined through community contributions in a cheap manner. IBM has already used the LAB method to create prolonged versions of open source models comparable to Metas Llama and the Mistral family of models.
“The way it really works is analogous to what number of open source developers were used to working,” Badani said. “It's like you may submit pull requests if you could have certain skills that you would like to bring to the table and if you could have some skills that you would like to make certain that the model can do this work.”
“(InstructLAB users have) the flexibility to contribute this to a community or a particular group of experts… after which bring the facility of synthetic data generation to make sure it becomes much more powerful.”
Developers can start with InstructLab on their laptops at no cost using the open source InstructLab CLI. They can then move to RHEL AI on servers for higher fidelity models and expand training on Red Hat's OpenShift AI platform for Kubernetes.
OpenShift AI 2.9
OpenShift AI also receives an upgrade to version 2.9 with latest capabilities for deploying each predictive and generative models and an expanded partner ecosystem. Red Hat emphasized its commitment to providing customers with flexibility and selection in deploying AI.
Red Hat is phasing out its AI offerings to bring open source innovation to enterprises.
Developers can start immediately with the InstructLab community project, now available, to increase open source models with domain knowledge. RHEL AI can also be launching in
Developer preview to supply a streamlined foundation for these enterprise-supported models. The latest updates to OpenShift AI at the moment are generally available and supply MLOps capabilities to deploy each predictive and generative AI models at scale. Moving forward, latest Ansible Lightspeed offerings for automating AI workflows are planned for later this yr.
With RHEL AI and InstructLab, Red Hat goals to do for AI what it did for Linux and Kubernetes: make powerful technologies available to a broad community through open source. If successful, it could speed up the adoption of generative AI within the enterprise by enabling subject material experts to enhance models with their knowledge and deploy them to production environments with confidence and support.
“It can also be a very important call. It reflects our tradition of investing in the facility of openness and the facility of community,” said Badani. “And then we would like to make certain we will advance that in AI.”
“We are very happy that the state-of-the-art is now at a degree where we will take into consideration how we will expand the meaning of ‘open’ on this context,” added Wright.
