Ep 1. AI Insights and Innovation with David Linthicum

[David Linthicum]

Introduction

Welcome to AI Insights and Innovation, your go-to podcast for the latest news, trends, and insights in artificial intelligence, including generative AI. Join us as we explore groundbreaking developments, interview leading experts, and provide in-depth commentary and advice on how you can find AI success within your organization. I’m your host, David Linthicum, author, speaker, Beatles geek, longtime AI systems architect, and analyst with theCUBE Research.

AMD Processors Compete with NVIDIA GPUs

Let’s get into the topic. So this topic came to me from some news reports I saw last week around the release of some AMD processors to compete with the NVIDIA GPU processors and Microsoft’s adoption of those processors into their cloud. And the article is linked in the description below.

[00:01:04]

But Microsoft said on Thursday it plans to offer its cloud computing customers and platform of AMD’s artificial intelligence chips that will compete with components made by NVIDIA. So long and short of this is AMD is coming out with chips to compete with the GPUs that NVIDIA is putting forth. And I’m sure they’re going to work perfectly well, and they’re there to compete in a market that’s going to be relatively fairly competitive, I think, as time goes on, because of the demand for generative AI systems and the ability for the processor manufacturers to create processors to live up to what the expectations are.

GPU Shortage and Processor Market Heating Up

Right now, there’s a shortage of GPUs because everybody’s rushing and buying them from the market because they’re looking for this big run on GPU processors, mainly the NVIDIA stuff, and looking at AI as being really the generative AI being the primary driver of that.

[00:02:04]

So what’s interesting here is that we’re going to see this occur over and over again over the next few years. And that’s because the processor market is heating up. Therefore, the investments are being made in creating technology that’s going to provide better and faster processors, high-performance computing, GPUs, and offering these chipsets in such a way we’re going to have one faster than the other and the other faster than the other and have many of them enter the market that I think it’s going to commoditize pretty soon. Even though NVIDIA has a leg up, they certainly have their platform, their architecture, and leveraging GPU-based systems, I think that AMD, Intel, and a number of chip manufacturers are going to have some fairly compelling offerings in the system.

Chip Manufacturers Offering Competitive Alternatives

They’re going to be in some instances more cost-effective than the NVIDIA stuff and therefore are going to provide just as good if not better services in some instances and people are going to migrate to those chipsets.

[00:03:08]

So if you’re looking for a detailed description and what are the technical differences between the processors, I did write an article out on LinkedIn. It’s posted in the description here where we get into the details behind it and the reasons why you may use one chipset over the other. That’s really kind of not the story here.

Focus on Business Use Cases for Generative AI

I think the story here is that we may be overly focused on processors as the enablers for generative AI and other AI systems. So I see in the marketplace just really kind of the tech focus there in terms of where that is becoming the place for innovation. The NVIDIA, Intel, AMD, all these sorts of different technology providers are out there competing in the space. The larger issue is that we may have some trouble in finding some of the use cases out there and there’s also a CIO article that’s linked in the description where it really kind of talks about what’s on everybody’s mind.

[00:04:08]

The fact of the matter is that if we’re moving to generative AI that we’re going to have to have a business purpose for doing so.

Business Purpose for Generative AI

So while we’re arguing about the processors and the different API sets and the different chip architectures and all these sorts of things that are enabling the generative AI systems to operate at speed, at the end of the day, I don’t think we have identified many of the business use cases that we’re going to need to justify spending the money that it’s going to take to build these generative AI systems. And keep in mind, specialized processors, HPC systems, whether you use it within public cloud providers or on-premise, it really kind of doesn’t matter. It’s still expensive. They’re going to be about three to four times the cost of traditional systems. So if we’re building stuff using traditional development technology and traditional CPUs, that’s going to be fairly commoditized now and that’s going to be the best bang for the buck if you’re able to get away with it.

[00:05:03]

And we’re going to find that not all these generative AI systems are going to require higher-end chipsets and that CPUs are perfectly fine for those. And I addressed this in an article, I think, last month on the reasons why we should not always consider GPUs as being slam dunk in creating generative AI systems. But back to the core purpose of why generative AI exists.

Generative AI Solving Business Problems

In other words, it’s there to solve a particular business problem. They’re there to deal with use cases. And as the CIO article points out, there doesn’t seem to be as many use cases. There doesn’t seem to be as much ROI on leveraging generative AI systems as people thought. And that’s causing a bit of a dilemma. In other words, if we’re moving toward generative AI systems or we’re looking for them to revolutionize the business, there should be lots of use cases that are fairly apparent to us that we’re looking to build these systems and to address problems that we currently have in the industry.

[00:06:03]

CIOs Evaluating Generative AI’s Value

Sometimes those are going to be there. Sometimes those aren’t. And I think that the CIOs taking a look at where the current enterprise is, the business that they’re in, the industry that they’re in, the use cases that are there, and the ways in which that generative AI can add value and ways in which it’s probably not a good fit. And so we’re going through this iteration right now, we’re adopting this technology. So the demand and the hype seems to exist in the tech press and certainly in the cloud providers. All the cloud conferences become generative AI conferences and enterprises have reacted to that. And so in other words, that’s why we’re hoarding GPUs and the NVIDIA chips. And we’re talking about different processors that are out there. And as the micro clouds are launching, which are GPU-based clouds that are out there as well, that seems to be another industry focus.

Addressing the Core Issues of Generative AI

And all this stuff is going on when we’re not addressing the core issues as to why the generative AI systems need to exist, the ability to find use cases and the ability to solve business problems.

[00:07:08]

And that really kind of needs to be the focus. And that’s my conclusion.

Homework for Enterprises and Consultants

So if you’re going through this issue, you need to do your own homework. You’re working for an enterprise, you may have an enterprise as a client if you’re working as a consultant. So what are the current problems and use cases that you see where generative AI could be a potential fit? And if it is a potential fit, then what kind of ROI can it generate if you leverage generative AI systems for those particular use cases? And the ability to kind of look at the business in an objective way, that if this generative AI system is put in place, it’s going to return so many millions back to the business over the next several years. Therefore, it should justify its development. You’re going to find that there’s not a ton of those that exist in the enterprises out there.

[00:08:00]

And in some cases, people are force fitting generative AI for use cases where they’re not a fit. Transaction processing, sales entry systems, inventory control systems are some of the things that I’m seeing out there. Well, certainly generative AI is going to add some value, but it’s not necessarily going to be enough value to cover the development costs and the risk associated with building systems using the specialized technology. And of course, the higher power chips. So again, my advice is this needs to be about the use cases.

Technology Solving Business Problems

That’s why technology exists. As architects, we’re there to solve business problems that are able to bring value back to the business. And our ability to be successful is going to be directly measured with our ability to return value to the business. And using that metric, we should look at any kind of new technology, whether cloud computing, cloud native systems, serverless technology, all the things that emerged in the last 10 years, and now generative AI in terms of its real business value as it’s applied to our particular use case for our particular enterprise in the business problems we’re looking to solve.

[00:09:07]

Avoiding Force Fitting Generative AI

And if we can’t find use cases there, then we shouldn’t force fit generative AI or any other technology for that matter into those problems, because we’re just making things worse. We’re making them more costly to develop, more costly to operate, and that’s going to leave the business in a bad state. And that’s what we’re not looking to do. We’re there to add value back to the business. So a couple of other words for advice.

Dave Vellante’s Research on Generative AI

Dave Vellante’s articles out on theqbresearch.com, and I linked a few of them in the description below. He does a great job in breaking apart the whole generative AI market emerging and certainly around the chip manufacturers and how they’re competing in the market right now and how it’s affecting networking and data and all the other moving parts that are part of this industry. Take a look at that because it’s good research because he does back it up with some data that he’s able to get from ETR and other sources.

[00:10:05]

So it’s not just an opinion. He’s not reproducing press releases, but getting out there and understanding what the enterprises are doing out in the space.

Follow Dave Vellante’s Research

And I think it’s very important. So make sure you’re following Dave and his research out there. He seems to post most Sundays, but it’s excellent and certainly it’s going to be innovative unto itself and its ability to shed some more light on where this market is going and what we should pay attention to and what we shouldn’t. So that’s all the time we have for this week.

Conclusion and Call to Action

Remember to check out our analysis and available services at theqbresearch.com. All one word. You can find me there or reach out to me at david.lenthicum, L-I-N-T-H-I-C-U-M, at siliconangle.com.

Sign Off

So until next time, keep the intelligence in your head. You guys take care now. Cheers.