Red Hat’s former CEO Paul Cormier has warned that it could take today’s AI pioneers longer than they think to work out both how enterprises can “consume” AI and how they can actually turn a profit on the technology.
Cormier joined Boomi CEO Steve Lucas on stage at the recent Boomi World conference in Denver, just weeks after officially retiring from the Linux trailblazer.
The Red Hat veteran, and chairperson emeritus, is also a board member of the API and data integration specialist, which is pitching its technology as critical to enterprises’ ability to marshal their data to supply the AI systems they’re so desperate to develop.
AI Predicated on Open Source
For his part, Cormier said the current AI wave was predicated on the work of Red Hat, and others in the open source world. “Linux wouldn’t have happened without open source, right? And cloud wouldn’t have happened without Linux being accepted in the enterprise,” he said.
At the same time, enterprises had realized they couldn’t move everything to the cloud, but this could only be enabled with the rise of Linux-based technologies like containers. “How would we have done containers with a Microsoft operating system that you couldn’t get at?”
Now, he said, “AI is the killer app for hybrid. So all of this is going to drive together.”
But, he continued, “Everybody thinks this stuff happens overnight. “But the open source timeline stretched out 23 years. (Or more, depending on your perspective.)
“Some of the things in AI are going to have to be invented, just like we did in the Linux world, in order for the enterprises to be able to consume it.”
Moreover, “The biggest problem with it is how do mere mortals consume it, especially in the enterprise, where things like security and reliability, all those big things make a difference?”
Massive Effort Required
All that grunt work takes a massive amount of resources, he pointed out. And this doesn’t decline over time.
Of the 2000-odd Red Hat engineers working on RHEL, he explained, “lots” are focused on backporting and other lifecycle management issues, rather than innovation. This is because of enterprise customers’ — entirely reasonable — demands for extended lifecycles of 10, even 15 years.
The AI era would be no different, he said. But as it stands, there are tens of thousands of AI models in the wild.
“So how do you consume it as an as an enterprise customer? You’re not an AI expert. That’s why I say the problems are so parallel,” he said.
It was down to the industry to make AI more consumable. But that comes on top of the complexity that CIOs face managing their infrastructure as a whole.
“Hybrid cloud has brought a huge amount of innovation and capability to the world. But from a CIO’s perspective, it’s just made their problem exponentially more difficult.”
To build automation on the backend doesn’t make sense “So all the products, including the building products, are going to are going to be built with that automation built inside. It’s going to be driven by AI,” he said.
But that also means that customers will have to train the AI models in question for the specifics of their situation, Cormier said.
Big 3 Issues
And those specifics include the three big issues for every CIO: security; efficiency; and resiliency. It was “too complex and too costly” for humans to do all of this, he said.
All of this represents a major challenge for vendors racing ahead with AI. OpenShift took almost 10 years to get profit, Cormier said, while RHEL took six years.
“Probably at least half of those years was working on the consumption of those technologies from the enterprise perspective. So, there’s still a lot of work to be done on the consumption of AI, laid on top of this very complex hybrid model,” he said.
Set a Strategy
It was likely that the “first pass of that will get thrown away.” In Cormier’s role at a private equity firm, “I see so many software companies that don’t have a strategy,” Cormier noted.
It was critical for both vendors and customers to understand what their overall strategy is and what their overall architecture is. Only then were they in a position to experiment.
“And if you don’t understand that, you’re going to inevitably start in the wrong place.”
While open source drives innovation because a “wide swathe” of people can contribute to a technical problem, that technical problem is only in the lab, he continued. “It’s when you start to deploy this out, that you really find the big problems.”
“I think we all, including our customers, underestimate how long it takes to get these technologies into the real world. The enterprise is a very unforgiving place,” Cormier stated.
Indeed, this pointed to one of the problems of open source, “When you do the next release, you forget about the previous release. That’s not how commercial software works,” Cormier said.
The post Ex-Red Hat CEO: AI’s Bumpy Road to the Enterprise appeared first on The New Stack.
The enterprise is an unforgiving place where new innovation like AI can take longer than expected for large organizations to consume, according to former Red Hat CEO Paul Cormier.