What role does Red Hat play in the red-hot AI market right now?

It’s really exciting to be able to train a model — and there are these big foundation models that are trained by arguably just a few companies in the world. But then the ability to refine those with your data — that’s the exciting part to us.

I don’t think every enterprise on the planet will train their own core models just from data to start, but they’ll take a model — whether it’s the foundation code model from IBM or GPT-3 — and they can incrementally train it with their unique data. That’s another really powerful aspect of our specialization to it.

In Red Hat’s world, we live in the, ‘How do we enable training?’

We have the hardware attach from RHEL to Nvidia. We have the distributed computing attached with OpenShift, which you need to do training. This isn’t like a one GPU thing. We work pretty closely with IBM — that is a supercomputer environment to train core foundation models on.

Although enterprises do that on a lower scale, I’m like, ‘Well, if I can make OpenShift work and OpenShift AI for IBM, we know it’ll work for those refinement layers.’ So we work on the training side to, given your data, get you a new model. And then it’s, ‘How do we actually deploy that so your app developers can use it?’