Covering Disruptive Technology Powering Business in The Digital Age

image
Red Hat Launches InstructLab With IBM in Ambitious Bid to Open Source AI
image
May 10, 2024 News AI Open-Source Red hat

Written by: Martin Dale Bolima, Tech Journalist, AOPG.

 

The next frontier of Artificial Intelligence (AI) is open source.

And ushering AI into this open-source era is none other than open-source leader Red Hat, who has time and again shown its unwavering commitment to the open model.

It plans to do the same with AI, having announced at the Red Hat Summit several exciting innovations that will drive AI into open-source territory to ultimately democratise this game-changing technology, elevate it, and expand its already wide, global reach.

“We are now seeing the impact of broadening the open-source ecosystem in the world of AI. And this isn’t going to slow down as models get smaller, as training gets cheaper, [and] as capabilities grow; the reach of this technology will expand across the planet,” said Matt Hicks, CEO at Red Hat, in his Red Hat Summit keynote.

Matt Hicks, CEO at Red Hat

This massive announcement comes exactly one year after Red Hat expanded Red Hat OpenShift AI, which serves as a consistent, scalable AI foundation that makes it easier to create AI/ML models.

Red Hat Taps IBM to Help Open-Source AI

Central to Red Hat’s vision for AI is open-sourcing several models, including a few it co-developed with long-time partner IBM, to pave the way for truly open-source AI—one with permissive licences, transparent training and data, and the ability for others to contribute something that is usable and of value.

“We believe in the power of ‘open’ to drive innovation. That means open licencing, open data, and open contributions,” Hicks added. “And as much progress we’ve made in the ecosystem here, the ability to contribute to a model has yet to be solved. I mean, you can get a model from Hugging Face and fine-tune it today, but your work can’t really be combined with [that of others].”

That is, of course, until Red Hat intervened, with the help of IBM.

Allowing individuals to contribute to AI model enhancement is the ultimate goal of InstructLab, described by Hicks as “a simple technology to make it simple for everyone, not just data scientists, to contribute to and train Large Language Models (LLMs).” Co-created by Red Hat and IBM, InstructLab is an open-source community project that gives more people pathways to enhancing LLMs—even with little experience in machine learning—and tailoring them to specific enterprise use cases.

InstructLab, with LAB standing for large-scale alignment for chatbots, works similarly as Pull Requests in other open-source projects, enabling individuals to adapt pre-trained LLMs—Meta’s Llama models come to mind, as do the Mistral and Granite models of Mistral AI and IBM, respectively—using less data and less computing power in less time. This will allow enterprises to train AI models, an undertaking that is typically expensive and time- and labour-intensive, according to their needs and specifications but without many of the barriers that generally come with actually forking over an LLM from scratch and then training it.

Evolving AI with InstructLab

Perhaps just as important, InstructLab is designed to evolve an already existing LLM with contributions from nearly everyone and anyone. Traditional, proprietary LLMs—think GPT, for example—are dead-ends in that they do not get updated until their next iterations. In contrast, InstructLab-trained, open-source LLMs are the gift that keeps on giving, continuously being updated by people worldwide who can add knowledge and skills to these models. These contributions can, in turn, be incorporated into future releases, effectively evolving the models and making them better and more powerful.

“Having watched open-source projects for a long time, you reach a point of critical mass where its innovation will exceed everything else,” Hicks noted, as he emphasised why open-sourcing AI is the way to move the technology forward. “I am not trying to predict the future of technology, but I think this is a safe prediction: AI won’t be built by a single vendor. It isn’t going to revolve around a single, monolithic model. Your choice of where to run AI will be everywhere, and it’s going to be based on open source.”

Red Hat is giving organisations that choice, enabling even those with limited budgets, expertise, and resources to not only contribute to developing AI but also to use it for their own business needs.

 

(0)(0)

Archive