As more businesses adopt new AI and ML technologies, increasingly these capabilities will run on hybrid cloud and hybrid multicloud IT systems, which have become the IT operating model of choice, according to the Enterprise Cloud Index report.
McDowell said many are already using AI for cybersecurity, to create code and other functions, but there’s still a lot to be learned in order to efficiently manage the infrastructure needed to power these applications. That means learning how to build operations that can scale up GPU, CPU and other computing resources.
“I need to figure out how to build the infrastructure to support that, because traditional compute does not account for it,” he said. “So we're seeing hundreds of experiments, thousands of experiments across organizations and IT has to step up and support those often in short order. It's a challenge operationally.”
Related stories:
Transcript:
Ken Kaplan: Any other trends that you see as we're going forward that are going to be momentous, especially the, especially during the economy that we have?
Steve McDowell: The biggest, the thing I'm asking most about right now is the impact of AI on IT operations. And it ranges from, you know, how many GPUs can I put in a rack and how do I heat and cool those? How do I power those? Because GPUs are very power hungry and heat-generating more so than CPUs by far. But on the other end, it's how do I use AI and specifically generative AI to, you know, is it a threat or is it a tool that helps me do my job better? And you know, like any tool it comes down to how you use it. but you know, again, we're still very early stages of how we're going to use this in IT operations. But we're using it for cybersecurity, look for anomaly patterns you know, pattern detection to tell me when there might be trouble. we're using it for compliance to read, you know, specifications across a number of industries and, and consolidate that into, you know, a set of actions that I need to take. you know, we're using it to write code, you know, simple pieces of code. Generative AI's pretty good at writing it if you ask it correctly, right? And that's kind of the double-edged sword, is asking it correctly and, and ensuring that it's trained on, on the right knowledge to really be a benefit.
Ken Kaplan: There's a lot of hype around AI and Chat GPT.
Steve McDowell: There's a lot of hype around AI, and that impacts it in a couple of ways. One is I need to figure out how to build the infrastructure to support that, because traditional compute does not account for it. So we're seeing hundreds of experiments, thousands of experiments across organizations, and it has to step up and support those often in short order. so it's a challenge operationally. The bigger question and where I think a lot of the hype lives is what's really the power of AI to help me do my job better. You know, generative AI and chat GPT and things like that have a lot of promise, you know, as they exist today. I think they're interesting tools, but I don't think anybody's solving real production problems in it, with those tools, right? What chat G P T does very well is, you know, you train it on a set of patterns, whether that's code or language or whatever and then it processes those in a way that it can, it can talk back to you, right? Essentially. So I can say, you know, train it on Python code and say, gimme an algorithm that sorts the problem with as they exist today is the training field is very broad. So we don't know the quality of the data, right? Those of us old in the industry, you know, there used to be a, a garbage in, garbage out saying, and that's really true of generative ai. So technology has a lot of promise, and I think it's going to end up being extremely impactful to it. I think it's just very, very early and, and, you know, I would advise any IT practitioner to well, the IT practitioner already has his expectations set. You have to manage up the chain and set expectations that this is not going to solve all the world's problems. I can't go fire all my tech support guys because I have, you know, Chad GPT answering the call. It's not that simple. So, early days, but let's keep watching it.
