The Futurum Group's Statement on Israel

Powering Your Future Business with AI Inference – Futurum Tech Webcast

Powering Your Future Business with AI Inference - Futurum Tech Webcast

On this episode of the Futurum Tech Webcast – Interview Series, host Daniel Newman welcomes HPE’s Frances Guida, Director, HPE Compute Workload Solutions for a conversation on how HPE is helping customers navigate their AI journey, AI inferencing, and bringing its expertise to empower AI adoption for businesses.

Their discussion covers:

  • Why AI adoption is so “hot” at present and how HPE’s customers are approaching adoption
  • A deeper dive into AI Inferencing as “the last mile for AI” and the real-world use cases for AI inference
  • The top challenges for customers at this critical junction and how HPE is helping customers navigate this journey
  • How HPE is bringing its own IP and expertise to bear for customers

See our related report: Operationalizing Generative AI: Making It Your Own, Creating Corporate Value.

Watch the video below, and be sure to subscribe to our YouTube channel, so you never miss an episode.

Listen to the audio here:

Or grab the audio on your streaming platform of choice here:

Disclaimer: The Futurum Tech Webcast is for information and entertainment purposes only. Over the course of this webcast, we may talk about companies that are publicly traded and we may even reference that fact and their equity share price, but please do not take anything that we say as a recommendation about what you should do with your investment dollars. We are not investment advisors and we do not ask that you treat us as such.


Daniel Newman: Hey everyone, welcome back to another episode of the Futurum Tech Podcast. I’m Daniel Newman, your host. I am the CEO and founder of The Futurum Group, and very excited today for this interview series here on the Futurum Tech Podcast.

I’m going to be talking to Frances Guida. She joins us from HPE, and we’re going to be talking about something that nobody… That’s not true. We’re going to be talking about something that everybody is talking about AI, and I think we’re going to have some interesting, different, and of course, unique insights like we do with every guest. Frances, welcome. Thanks so much for joining me on the show today.

Frances Guida: Thanks, Daniel. It’s great to be here.

Daniel Newman: All right, first and foremost, name, title, specs. Tell us a little bit about the work you’re doing over at HPE.

Frances Guida: Yeah, so I’m Frances Guida and I lead our workload solutions team for our HPE ProLiant compute platform. Those are the ProLiant servers that are the backbone of a lot of enterprise data centers. We look at workloads from classic virtualization to cloud native workloads, and of course, over the last year, year and a half, we’ve had a very intense focus on what we need to do to help get ready for what we really see as the next wave, which is our all-around AI.

Daniel Newman: So let’s start there with AI.

Frances Guida: Yeah.

Daniel Newman: By the way, I’ve had some great conversations with many of your peers. I have a regular sync with Antonio, so I definitely get the background. But what I love in conversations right now is getting the perspectives of people on this broad AI trend line. For some, it feels like it came out of nowhere. For other people, it’s like this has been going on for a while.

HPE has been doing this for a little while. It isn’t brand new, but the popularity from November 22 with the rise of ChatGPT to now has been palpable. Talk a little bit about what you’re seeing and hearing about why AI is now so hot for all of your clients.

Frances Guida: Yeah, so you are spot-on. In fact, I was doing a little bit of research. The concept of AI goes all the way back to the 1930s, and Alan Turing, the name itself goes back to 1950s timeframe, so it has been around for quite a while. Maybe, let me talk a little bit about why now, at least my point of view on why now. There’s sort of some logical elements, and then of course there’s the emotional elements of the hook with ChatGPT.

But from a logical perspective, we now have… Obviously, compute power is growing Moore’s Law every 18 months or every two years. It’s doubling. So we have more compute power to throw at things, which means that the AI engineers have had more compute power to throw at those neural networks, the deep learning with the billions and billions of connections.

But I think what really has made the difference is that for the last 20, 25 years, the world has become digital. So we have all of this massive digital data, whether it’s digital text or digital video, and what computers need to do AI, is they need to be trained, they need to see a lot of data and identify what the patterns are. So now, unlike 20 years ago or 40 years ago, we have masses of digital data that we can just throw at these AI massive engines and we can find patterns that otherwise couldn’t have been found if we didn’t have all the data to train it.

So I think that’s logically why now. Of course, sort of emotionally why now, it’s all about the killer app, right, ChatGPT, and I’m not even sure I’d call ChatGPT an app. It’s kind of like a great proof of concept about, wow, look what it is that you can do. You can put a prompt in and get a computer to write a bedtime story for your kid. Isn’t that cool?

Daniel Newman: Yeah. I think what it is it’s about hitting this empathic relationship that exists between humans and machines. And interestingly enough, machines can’t actually be empathic, but the feeling is, right?

Frances Guida: Right.

Daniel Newman: Historically speaking like that, human machine, it was always a human figuring on how to speak to a machine in the language that the machine understood, so code, right?

Frances Guida: Right.

Daniel Newman: And so even other things like web development or SEO, we learned to optimize. So the early era of Google, because Google’s been doing AI a long time, the early era of Google was all about humans learning to prompt Google in a way that Google best understood it, so it could give you the best return. So you wouldn’t say like, “I want to go out to dinner at a restaurant that serves French food that’s less than 20 miles.” You didn’t talk to it like that.

Frances Guida: Right.

Daniel Newman: You’d be like, “French restaurant, 20 miles, Santa Clara,” and you would hope that it understood that. And over time now, the example I gave is what ChatGPT or Bard or all these different solutions are starting to allow is we start to have these much more humanistic interactions with the machine. I think that’s super powerful, but I think it goes so far beyond that.

And your point really comes down to the data, the exponential nature of data, enterprise data, customer data, experiential data. You’ve got structured and unstructured, you’ve got observability and edge, you’ve got this whole gamut of data that companies are trying to extract value from. And so while I think this is the opportunity, it also creates a ton of confusion. It creates a ton of challenges.

So talk a little bit about what you’re seeing in those customer interactions about how they’re adopting this because it’s not just start using ChatGPT, and we certainly aren’t just ingesting all of our files. We know in fact that we absolutely shouldn’t and can’t do that if we want to protect ourselves. So what are you seeing your customers do in terms of trying to move down the adoption curve?

Frances Guida: Yeah, well, let me just, first of all, lay out a couple of the basics of how does AI go from a big model to something that’s actually usable from a business perspective. So you have to start with those models and ChatGPT is a great example of the model, Llama 2. Those are all the LLMs, but there’s many different kinds of models. And so in order for an enterprise to be able to use AI, you first have to start with someone. Somewhere has to have created that model.

Once the basic models are created, and that’s where again, ChatGPT proved, here’s a great model that can be used, you want to tune them and customize them. You want to make them smarter to do a specific purpose. Maybe, it’s to distinguish between a positive and a negative movie review. And you got a lot of sarcasm often in movie reviews. So how can the computer identify sarcasm? It’s not necessarily evident unless you’re teaching it how to do that.

And then once you have something tuned, maybe what you’ve got is an embedded model which has private data that’s specific to your enterprise. The real magic of AI comes when you can actually use that for inference. So what I think is going on right now in the industry is, all of a sudden, we’ve got this great public sort of proof that this works.

And now, there’s a massive shift within organizations to figure out how in the world can I put this to use for my business to deliver a better business outcome, because at the end of the day, enterprises are profit-making operations and they’re all about, “How can I do something faster, better, cheaper,” right?

Daniel Newman: Yeah, you make a lot of good points there. Definitely, I’m going to ask you here momentarily about inferencing because it definitely is the killer app. We’re hearing a lot about training because, well, we know the market always leads the discussion and all the volume of super powerful, A100 and H100 chips from NVIDIA, and this is being delivered to sell crazy amounts of conceptual AI inferencing in the future, right?

Frances Guida: Right.

Daniel Newman: We got to train all these large language models. We got to get them all prepared, and then the apps will come. We’ve heard that about 5G, too, so we do hope that it actually happens in time, but I do think it will. I do think it will. And I do think what we are starting to see now is we’re seeing we’re kind of past the, I think, we’re past the peak of hype.

I feel like the first quarter of this year was peak hype, like, “Oh my gosh, we’re going to use these apps, like you said, these proof of concepts, and I’m never going to write anything again. I’m never going to have to build another PowerPoint deck. I’m just going to type in what I want and done.”

Frances Guida: Set up

Daniel Newman: There’s a lot of responsibility. There’s ethical. There’s data governance. There’s, of course, data cleansing. There’s sovereignty. There’s all the same problems, by the way, we had in the Big Data era. But you did mention that inferencing is kind of the killer workload. I do agree with this.

I think we’re going to move from training being the story and all the volume and it’s going to move to inferencing, and that’s also going to change the whole economics because who’s making money or which companies make money changes, because it goes from very hardware-centric to more software-centric. So talk about, at least in your mind, why inferencing is so important, why it’s kind of the last mile for AI.

Frances Guida: Yeah. So I think the best way to illustrate why inferencing is to just think about a use case. And I’m actually going to take a computer vision use case. I know a lot of the buzz is all about LLMs, but there’s a lot that we can do just based on video feeds, right? Cameras these days are everywhere. You go to a gas station and there’s a camera aimed at you as you’re pumping your gas, and then there’s a camera aimed at you as you go into the gas station, there may be one on the outside of the restroom to see who’s going into the restroom. They’re everywhere.

Retail stores can take all of those video feeds. And all of a sudden, they can use inferencing to produce a tremendous amount of real-time insight as well as operational insight, so real-time insight. They can do things like loss prevention, “Oh, somebody just snuck a can of soda underneath their coat. I’m going to decide if it’s worth it to send my manager out after them, or if I’m just going to let them go because they’re a teenager,” but at least you can make those decisions. It’s not something that’s just going to happen to you. It’s much more under your control.

Operational decisions, I’ve talked to… Actually, this is a real example. I’ve talked to somebody who runs gas stations and convenience stores around at a thousand locations in the United States, and they’re like, “Yeah, what we want to know is when people pump their gas and then they come into the store exactly what are they buying, so we can maximize the workflow.

Now, we’ve had analytics that have done that, but they don’t do that to the specific individual. With video, you follow that individual you know exactly what it is that they’ve pulled off, so you can make better decisions about how to maximize the profitability of your convenience store based on what it is you place where, and that’s again, based on a much more granular level of data than what it is we’ve been able to do in the past with a lot of analytical tools.

So I think those are real-world use cases that you can do with AI inference. Now, I haven’t even gotten to some of the things you can do with LLMs around chatbots, around voice detection and fraud detection around customer service, around recommendation engines and all of those, once you’ve trained them on data, you can then use them on an ongoing basis, and then maybe you want to fine-tune them and continuously train them so that they get better and better and better over time.

Daniel Newman: Yeah, we know reinforcement learning is going to be very important. In fact, probably one of the more prophetic things I’ve thought about is you almost have to have things go wrong in order to make AI work correctly, right?

Frances Guida: Right.

Daniel Newman: Because it has to keep, it continues to learn and get better, so it’s basically, it does what all of the people that we’ve ever probably been frustrated in managing and stuff don’t do. It’s like, “You make a mistake, great. Did you learn from it? Or are you going to keep doing the same thing?” AI will keep learning and then it gets better. That’s self-driving cars, a great example of the power of reinforcement learning and inference.

But look at your Amazon Netflix experiences, how much better it is now in terms of identifying or Spotify the next song you want to hear, it just keeps getting better and having more data, more opportunity to learn, and just more continuous improvement in that experience. So Frances, it’s really powerful, and these are everyday examples. And that’s also to your point in, I think we made the same point here, is that this isn’t all new. I mean, it isn’t new. We’ve been doing this for a while.

Frances Guida: We have been doing a lot of this for a while. It’s just a perspective shift to think about it in a different way and to think about it in kind of like a way where instead of your improvements being incremental, you can actually make some greater leaps.

Daniel Newman: We’ll call it exponential, right?

Frances Guida: Exponential. We’ll call it exponential.

Daniel Newman: Exponential.

Frances Guida: So there we go. We’re marketing.

Daniel Newman: So let’s pivot a little bit to the challenges here.

Frances Guida: Yeah.

Daniel Newman: I speak to a lot of CIOs and CEOs, and this has been really a frontline-to-board conversation with AI. It changes every part of a business, whether it’s a knowledge business, whether it’s a very industrial business, lots of labor. We’re seeing it make some substantial changes, but the change management is significant. So talk a little bit about what you’re seeing and hearing as the top challenges for your customers in trying to really navigate this path to successfully implementing and utilizing AI.

Frances Guida: Yeah, so I’m seeing a couple of challenges. So first of all, I’m seeing the challenge of… Where can I start with this? Okay, I see that this has got a possibility. What’s the lowest-hanging fruit both from an implementation perspective, from a risk management point of view, from a learning perspective? Is there something that I can do now that will get me a quick win and that I can learn from and prove? This is not any different than some of the last waves of technology. People need to know where it is that they start.

Certainly from an enterprise perspective, there’s a lot of concern about privacy. I know a number of corporations who are actually restricting use of tools like ChatGPT, because yes, AI learns and gets better, and the way it learns and gets better is by taking the data that you put into that, to put into ChatGPT to use it, to then use that to chunk through the next set of data. So if you put company private data out into that ChatGPT, you’ve just taken company private data and you’ve put it into the general public.

So a lot of organizations are very, very rightly concerned about how do they make sure that they keep the things that are private to them within their organizations and they don’t… And there’s lots of solutions for that, right? So this doesn’t prevent you from using AI. It just means you have to use things a little bit differently than what you might give to a general purpose public.

Daniel Newman: Yeah, that’s a great point. Companies have to really navigate everything from, “Is the material accurate? Can it be trusted? If you did use an AI or an LLM, is it your material? What’s the source attribution-

Frances Guida: Right.

Daniel Newman: Right, license. If that’s being used, is all the data private in different global jurisdictions, you have much more severe penalties for private data leaking out, then perhaps in the US. So you’ve got a lot of risk there, and you’re seeing companies stepping up and actually saying like, “If you use our generative tools, we’ll go to court with you.”

Frances Guida: Yep. I have seen some stories about that just today actually.

Daniel Newman: Yeah. So there’s some of that out there. And I think companies like HPE are going to be thinking about this as how it can make sure that customers can adopt and can feel comfortable and know that while technology is constantly improving, it can be mighty imperfect at times. So Frances, to wrap things up, I guess, let’s kind of end with HPE.

You’ve given some great perspective on your lens and through the customers and conversations, but what is HPE’s kind of unique? What is its IP and its expertise that it’s bringing to this particular conversation? There’s so many companies, every company this year, I’ve been a hundred events this year, and every company’s saying, “We’re the AI company.” How is HPE’s making that claim and differentiating itself?

Frances Guida: Well, I can’t speak about all of the other companies, but I think we are entitled to that label. So we have the largest supercomputers on the planet that are built with HPE, the HPE Cray Supercomputers. So in terms of building the machines that can do that AI training and modeling, sort of HPE is hands down, I would say the leader in building those things, but it goes beyond that, because AI is not just about those computers, it’s about having the right software to help organizations navigate through taking advantage of that, managing all of the data it is that they have, so we’ve got a whole set of tools there.

And then finally, and I think this is where we haven’t… We’re really at the very, very tip of the iceberg. The opportunity for AI, not now, but in two years, in three years, is all around that inference. And I think HPE, what we’ve already done with our HPE ProLiant Gen11 to really create systems that are optimized for AI inference, we have that out of the gate, we’re ready to go, and that’s something that’s only going to get better from here on in.

Daniel Newman: Well, congratulations, Frances, on all the success. It is an exciting time from speaking with you, speaking many of your colleagues, partners, and other executives across the HPE team. I think there’s a lot of very interesting work and exciting work that’s being done. We know it’s a ecosystem, a lot of partnership, a lot of collaboration. No one company, despite how much some companies want to believe they are the entire solution, will provide everything consumers and enterprises need for AI, but HPE continues on the compute side through software, through hybrid cloud offerings.

Edge has impressed me as an analyst with what it’s doing, how it’s positioning, and how it will play an important role for many companies in their approach to deliver and take advantage of the promise of AI. So Frances, hope to have you back sometime soon here on the Futurum Tech Podcast. Thanks so much for joining me.

Frances Guida: Well, thank you for having me. I appreciate it.

Daniel Newman: All right, everybody hit that Subscribe button. Join us here for all of our episodes of the Futurum Tech Podcast and our interview series with myself, all of our other analysts, and so many thoughtful leaders from across the technology industry. But for this episode, I’ve got to say goodbye. See you all really soon.

Author Information

Daniel is the CEO of The Futurum Group. Living his life at the intersection of people and technology, Daniel works with the world’s largest technology brands exploring Digital Transformation and how it is influencing the enterprise.

From the leading edge of AI to global technology policy, Daniel makes the connections between business, people and tech that are required for companies to benefit most from their technology investments. Daniel is a top 5 globally ranked industry analyst and his ideas are regularly cited or shared in television appearances by CNBC, Bloomberg, Wall Street Journal and hundreds of other sites around the world.

A 7x Best-Selling Author including his most recent book “Human/Machine.” Daniel is also a Forbes and MarketWatch (Dow Jones) contributor.

An MBA and Former Graduate Adjunct Faculty, Daniel is an Austin Texas transplant after 40 years in Chicago. His speaking takes him around the world each year as he shares his vision of the role technology will play in our future.


Latest Insights:

Shopping Muse Uses Generative AI to Help Shoppers Find Just What They Need—Even Without the Exact Words to Describe It
Sherril Hanson, Senior Analyst at The Futurum Group, breaks down Dynamic Yield by Mastercard’s new personal shopping assistant solution, Shopping Muse, that can provide a personalized digital shopping experience.
On this episode of The Six Five – On The Road, hosts Daniel Newman and Patrick Moorhead welcome Chetan Kapoor, Director at AWS EC2 for a conversation on AWS Generative AI Infrastructure announced at AWS re:Invent.
A Deep Dive into Databricks’ RAG Tool Suite
Steven Dickens, VP and Practice Leader at The Futurum Group, shares his insights on Databricks' RAG suite and Vector Search service, which are reshaping AI application development.
Marvell Industry Analyst Day 2023 Sharpened Its Vision and Strategy to Drive Infrastructure Silicon Innovation Key to Advancing Accelerated Computing
The Futurum Group’s Ron Westfall believes Marvell is solidly positioned to drive infrastructure silicon innovation for accelerated computing throughout 2024 and beyond, especially as the advanced computing opportunity expands during AI’s ascent.