Techstrong Gang – August 30, 2024

Happy Friday, and welcome to another edition of Textron Gang. I'm John Schwartz in Silicon Valley as we get ready for the holiday weekend. So things never change.

Nvidia earnings knocked outta the park yet again. But with a caveat, the AI chip said market is about to explode, and Amazon and Accenture are doing their best to deliver responsible ai. Those topics in our star study panel is here to explain it all to you.

You're watching Textron Gang. Hi there. Again, it's John Swartz here, intrepid Silicon Valley editor Recorder from the Techron Gang.

September is almost upon us. And where did the summer go? Today we're going to explain some hot topics with Eden Hotter takes from our panel.

And let's introduce everyone. We're gonna start on the West Coast, and we're gonna travel down Highway 1 0 1, or is it two 80 to my hometown of San Jose. And Lisa Martin, CMO advisor at the Ventura Group.

Hey Lisa. Hey, Don and gang, great to be with you guys on this Friday Holiday Eve. We cannot wait.

Yes. Next up we're a little bit down the road from you is in Santa Clara where we say hello to Futur Group analyst, guy Courier. Good to have you guy.

Good to be here. And yeah, three day weekends, uh, just around the corner. You could, you have my permission to go on your three day weekend right after this podcast.

Thank you so much. And in Charlotte, North Carolina, we have Hope Lynch, our Dev DevOps expert who joins us at again, it's good to see you again. Hope.

How are you doing? Good to see you As well, and thank you. And, uh, and I appreciate the kinder timing then, uh, my counterparts have on the West coast today.

Oh, it's, I got two cups of coffee. It's brutal. I know Lisa's working on her second.

Same. Yes. Yep.

I dunno about you guy, but it helps. Um, let's get started with our first topic. You know, what amounted to the biggest tech news of the week?

Nvidia announced its much value results on Wednesday. First and foremost, it reported operating income of nearly $19 billion for the second quarter, which is basically up 174% from the year ago quarter. It's gap operating margin, though, however, narrow to 62%, which weighed down on the shares of some partners such as Dell and Supermicro.

The larger concern, and what I kind of wanna pivot to is the return on investments. Specifically, investors want reassurance that cloud customers and other buyers of NVIDIA equipment are getting sufficient returns on their spending enough to keep pouring money into AI hardware. Which brings me to hope.

And, and, and kind of your, your thoughts, your, uh, overall summary of NVIDIA's results, specifically around AI spending and what we can glean from these results. Mm-Hmm. I, I think it has been anyone who's looking at anything, anything in technology hasn't missed the explosion around AI investment across the board.

But the company that has captured the most attention, the most imagination and and right now looks like the most profit has been in Nvidia, um, unprecedented capital flow into AI infrastructure research development. Uh, their GPUs are now considered the gold standard for AI workloads. So they are the cutting edge for everything from cloud-based AI to research and machine learning, deep learning.

Um, so, so they are broadening out across all aspects of the market, but as you stated, so many concerns. So is it economically viable to continue to invest so heavily, uh, in these areas? Maybe not, because what is the actual ROI on that investment?

Uh, companies that are using technologies like generative AI are saying, yes, we are using it, but we're not quite sure how to quantify our return on that and the benefit that it's providing, uh, to our bottom line. What is the foundation of all of that, this, this work that Nvidia and other companies like it are doing? So there are technical challenges.

Um, they've had some delays in some of the chip sets they wanted to release. Um, environmental impact with, uh, scaling AI infrastructure is a concern. No one really has a solid answer for that, right?

Uh, right now. Um, but once you look at it, can growth continue indefinitely? A lot of people say no, but no one has an answer, uh, as to what that growth point, um, will cap at.

Okay. I think it's something I, I think we had a episode recently where we talked about, uh, I think Nvidia had something like 92% of its market. Maybe that's, maybe that's, uh, number's too high.

I don't know, guy. I'll ask guy about that. But are the, It's looking like 94% right now.

Okay. We're gonna talk about the, uh, that the, uh, recent study from, uh, from my company fu my mine, Lisa's company, Futurum Group. It's around 94% in 2023.

Okay. So there really haven't been, there've been a few tiny fissures in the Nvidia fortress. I was gonna ask you Guy, um, is based on what you said, well, what Lee, what Hope said, and based on the results and also your report, do you see any slowdown soon in NVIDIA's growth?

Or is it just continued to maintain this 90 plus percent dominance? Well, I, I wanna make it clear, it's not my report, actually. My colleagues, uh, mark Beaty and Steven Dickens are, are authors of that report.

Um, but, um, I, I, first of all, I mean, it's fair to say, I think it's fair to say, um, there has been no single brand as dominant in an area, uh, in a wave as important, um, as AI is now, or cloud was before, or PCs or mainframes there, there really is no single company that has ever been this dominant. Even Intel in its heyday for X 86, um, was not really this dominant. And, um, economic, uh, economics would tell you that that's not sustainable, not in an open market.

Um, but Nvidia is, I think a prime example. It's somewhat Apple-like in the sense of all the pieces of it working together, uh, uh, at a high level of execution. It's not just, uh, engineering and product development, it's software, um, it's solutions.

It's, uh, partnerships. Um, and, uh, you know, that they can sustain, they have the capital, um, they have the pipeline. I mean, you know, for every dollar that, uh, Nvidia can recognize, there's probably a, I don't read the financial reports as much as I should, but there's probably a dollar or two, um, just waiting to be recognized once they can ship their biggest problem is the same big problem the whole industry has, which is, uh, manufacturing and supply.

Um, as more of these fabs go online, um, that will alleviate, um, but the inevitable, um, point that I think hope is talking about is going to come, it's not weather, but when, which is the point at which everybody looks around and realizes that they have these sleds of GPUs, um, that they haven't been utilizing and maybe they don't even know how to utilize and they're gonna stop a little bit. And wait, one of the things the study shows is how dominant in terms of, um, in terms of on the buy side, uh, the cloud providers are the various cloud providers, uh, that they are consuming. What's something like 43% of 43 or 44% of all the GPUs produced?

So I think that, um, it's, it's, that's one of the reasons probably for the market's reaction, because Nvidia saw one of its biggest, uh, drops in stock, uh, price after the announcement despite beating estimates. Um, I think that there's a general sense that at some point the shoe is gonna drop. I like to say the truth always wins.

Um, it's not whether it wins, it's when it wins. That's the whole question. And so that's really the question right now with Nvidia in particular, not for the market as a whole, which is showing robust signs of growth.

Yeah, those are, I guess even by the route size expectations in terms of what was expected, they felt just a little bit short. So maybe hence that and ROI and maybe some small nitpicks are what dropped the stock a little bit, but I'm sorry to interrupt. Yeah.

Um, oh, I was just gonna say, guy made a great point about the, it's, it's not if it's when, you know, being out here in Silicon Valley, um, wall Street, Silicon Valley this summer have really been questioning can generative AI produce enough measurable benefits for organizations? But we know that the trend is the CapEx investments aren't slowing down. Guy mentioned in the future intelligence report, the majority of, uh, the those buying are hyperscalers.

And we also know that from NVIDIA's customer perspective, it's the mega cap companies, mag Sevens, it's meta, Microsoft alphabet, amazon do com, Tesla, that's, that are the biggest consumers of this. So Hope brings up a great point about the ROII was on Schwab network last month talking about it because investors want to see proof in the pudding. I also get to talk with a lot of, um, companies who are using gener J value from a marketing perspective, what they're seeing to hope's point isn't necessarily bottom line impact a lot of productivity improvements, but I think it's a conundrum because we're, we're kind of in early innings.

So the sustainability factor guy brings up a great point. How sustainable is it? Will the shoe drop eventually?

But I think what investors wanna see out here in Silicon Valley and Wall Street is where are the measurable ROI impact? Where, where are the benefits to justify this continued trajectory that is up until the right in terms of spend? Yeah, I know, I know Daniel Newman, it's a group, has written quite a bit of extensively.

I know I've helped him or ghost written a couple things for him where he talks about the early age, early era of generative AI and the early winners and how we're gonna go through three network effects till we finally, finally reach consumers in a certain way. So in a sense, Nvidia is going to benefit from that for the foreseeable future. Uh, is, I'll throw this out to the entire panel.

Is there anyone or anything, and we're gonna talk about this in the second segment, is there anything along the way, coming along the way that could make even a, a free d in what Nvidia is doing in terms of dominating the news in this area? Dominating the news is one Thing. Dominating the news, the markets, the mindset, everything.

Well, I'll, I'll start us off briefly. Um, so I think that, that there are, um, sort of three or two and a half, uh, um, areas where, uh, NVIDIA's, um, at least technical relevance and need are being somewhat muted. The first one is the obvious one, which is competitors.

A MD has its instinct line. Intel's come out with gouty, um, not to mention, uh, a fair amount of a pretty intense development by the hyperscalers themselves to provide alternatives. Um, this is natural when supply is a problem and, and nobody wants to be as, as excellent.

Um, in many ways as the NVIDIA experiences, nobody wants to be stuck to, to a single vendor. So that's the obvious one. I think a less obvious one, um, is the challenge from alternative ways to at least infer do the inference part of ai.

Um, which would is in other words saying, uh, you can use CPUs for this. Um, you can use X 86 CPUs, you can use, uh, high core count X 86 CPUs. You can use the national high core count to arm, um, CPUs, uh, being developed like, you know, graviton or, or Ampu and so forth.

So CCP, an all CPU, uh, model is not impossible. So you don't necessarily have to stick to, uh, GPUs, um, in order to, um, in order to, uh, deliver either training or, or, or inference of ai. And those are both potential, um, ways to, uh, to dent NVIDIA's share and dominance while it will still grow nonetheless.

Yeah, and to add to that, there is, um, an area of research that, that some are, some companies are looking at, like Google, um, with TPUs, Amazon with graviton chips. So they are trying to understand how they can design to optimize very specific AI workloads so that you can have an advantage, maybe performance or energy efficiency or cost. This specialization, if it were to catch on, could make the market a bit more fragmented and you could have a more of a difference in what the AI solutions look like in the hardware configurations they require.

Now with that being said, I don't think Nvidia is slow by eating meat. And as they see that, I think they are going to be probably better positioned than anyone to take advantage of it. Well, that's a good way to end it because we're gonna segue into our next block, and we're gonna talk about some of the contenders or pretenders, however you wanna look at them, in, in kind of this overall view of the AI processor wars.

And we'll be back with that segment in a moment after these messages. Welcome back to Techron Gang. It wasn't officially AI Chip Accelerator week, but financial results from NVIDIA to recent conferences and a new research report would suggest otherwise.

So first and foremost, Nvidia dominated the news light always does. But while it was doing that on Wednesday at nearby Stanford University Intel, others were touting their competing AI accelerators. Intel's gearing up for its September product launch around GTI three.

We can talk more about that later. Meanwhile, at the VMware Explorer Conference in Las Vegas, our Mike Ard was very busy reporting on benchmarks that showed a 40% gain in Nvidia Blackwell, GPU performance, Broadcom at a support for GTI two to the VMware platform. And then even IBM reviewed the next generation of AA processors for the mainframe.

On top of that, came this intriguing report from the FUTURUM group's research arm, Futurum Intelligence. Last week it released an AI chip set market analysis and study providing a five year market forecast. That's what brings us to Guy.

And maybe we could talk a little bit about the reports, where the market is headed and what this means in the larger narrative for Nvidia and everyone else. Well, intriguing is right, um, uh, you know, some of the results just put numbers on what we already know, like 30% annual growth of the market. But to say that in 2023, um, term intelligence measures, the, the, the AI related, uh, uh, chip set market at $38 billion.

So that would've been last year, but by twenty twenty eight, a hundred and thirty $8 billion, that's what 30% growth means. Um, so, uh, you know, we could break that down a little bit. Three quarters of that is GPUs.

So not all of it is GPUs. Not all of the chip sets being bought for AI are GPUs. Around 20% are cpu and then 5% or so specialty.

Um, I like to think of it also just in terms of the data center market as a whole, 38 billion. Put it in context, that's about 10% of data center spending. I mean, that's a lot.

That was 2023. Um, with that strong growth growing faster than the data center market by 2028 instead of being a 10th of the market, um, a fifth of spending on data centers will be for, uh, AI related chip sets. I mean, that's kind of remarkable.

Um, one of the things that we've talked about on this show before is that AI itself is not really, you know, an application. You don't really have AI applications. ai, AI is a service provided to an application.

Even a chat that is, you know, you with an ai that's a chat application that happens to have AI as its foundational service. What's the implication of that, that AI can go into everything and anything. And so when you look at it that way, you're seeing that a lot of what you might call the chip set, uh, resource, um, the processing resource of a data center increasingly is going towards AI related functionality wherever the application is or whatever it is.

And this report highlights that pretty well. Um, Lisa, I know you dug into this. Uh, is there anything, anything that jumped out at you, um, in particular, uh, or kind of you see as a trend or something that's gonna lift to the surface soon?

A couple things stuck me from this feature of intelligence. Um, report one was, uh, and guy mentioned this in the last segment, that the over 40, I think it's 43% spend is coming from the hyperscalers themselves. So we know that the demand has got to be there, but there's some interesting competitors in this space, grok with a Q being one of them, cerebrals as well, that gr both of those were mentioned by Nvidia in earnings yesterday.

Yes. And gr it was funny. They, they flew under the radar.

I, I sat down with their CEO last fall about eight months or so ago, and one of the kind of biggest newsworthy stories at the time was how we sent a, a kind of a cease and desist letter to Elon Musk, uh, because of Exiss grew up with a k, fast forward this summer, they've just received 640 million in funding. 8 billion. So the competition is heating up.

Uh, Nvidia is aware of it, and, and I, so I, I'm interested in some of those smaller players as to what they're doing, um, in that TPU space as well. But guy's point, it's not, you know, the, the reports demonstrating the investment in GPU isn't necessary to run these workloads. CPUs are sufficient.

I, I suppose it will depend on the use case, but looking at, you know, 18 different chip makers, um, it's an interesting market. Nvidia clearly has a moat, but, um, there's competition, which I think is good. I think, I think society will benefit from that competition.

What, what about you hope? Any, any thoughts on, on this? Also, I want to ask you too, if you could assess, you know, we referenced them earlier, the, the, the prospects of, of like a, a GTI three or a, uh, uh, any, what IBM is up to or even, uh, Broadcom, I mean, is there, are, are there, is there any traction there?

I guess I'm trying to be diplomatic about it, so that's, yeah, It's actually boring. I, I think, um, I, I think traction is relative, uh, with the outsized numbers we're seeing from NVIDIA right now. Um, but, um, companies like Broadcom, yes.

So, um, they're, they're moving forward. They're trying to figure out how they can, um, have AI accelerators in VMware's cloud infrastructure so they can integrate those AI processing capabilities into much broader IT environments that then enables, uh, other industries even because we think about, um, automobile manufacturers, um, AI and healthcare, uh, there will be AI touching probably every device and every capability we can think of in even more ways, uh, than we have now. So those integrations, um, let them leverage this advanced AI hardware and not hopefully, eventually have to have completely new systems.

So the way VMware has been looking at it, um, upgrade your existing VMware based infrastructure with the AI capabilities, AI adoption is more accessible and cost effective. Now you have a huge multiplier for those industries and companies that want to take advantage of it. If VMware is able to, uh, make something like that work, it'll be revolutionary for them.

And, and not knowing what, what their results might look alike, but it will give them a buried good foothold. I think the, the, oh, sorry. Go.

It's worse when you think about, um, uh, NVIDIA's relevance, um, I think a lot of it hinges on qda, the software development platform, qda and solutions is what I would say are, are, are the, the strengths of Nvidia qda is well known. It's well enabled in the developer community and the data science community. Um, everyone, uh, tries to emulate or fit in with it.

It's a de facto standard, even though it's proprietary. Um, Intel, uh, is famously, uh, uh, helpful and, uh, sup uh, supportive of adoption through software, but I'm not sure that Gouty has really shown, um, the gouty software suite has, has shown any particular strength there yet. Um, a MD has a rock, uh, q to compatible, so hope I, you know, I turned to you, uh, with, with wondering your perspective on, on these things.

Like the idea is that acute a developer could use rocker, could use the gouty suite without a whole lot of, um, uh, need to, uh, ship thinking. And even port code like that sort of thing has similar frameworks. Um, I think that's a key enabler though, and there's a lot of stick on qda and that keeps people on Nvidia.

What do you think? I, I think you're right. Having, uh, let's say when, when everything's really started heating up around gener, uh, generative ai, uh, I was curious.

So I signed up for the developer program at Nvidia to see, you know, what do they offer, what tools do they have? And let me tell you, they roll out the red carpet. It's beautiful.

Yeah. But I think the point we are at right now does remind me of stages we've been at with software development and other technologies. In the beginning, it is so siloed and so specialized by the manufacturer, it is so proprietary that it is harder to shift from one to the other because you are learning so many things over again.

I think that's the point where we are now. So if you really understand, um, how develop to develop with Nvidia software, if you really understand their GPUs fantastic, but now someone wants you to shift to a different technology, you have a steep learning curve, I think that will take some time to smooth out. But each of those vendors is providing, I would say, a pretty good experience, good tool sets, um, and a good community.

Um, but again, Nvidia is definitely leading the way. Hey, can I just jump back to something that you just said, hope is interesting, and it kind of parallels what guy I believe mentioned earlier when he discussed Apple and Nvidia and, and Gaia. Maybe I'm, I'm kind of going far astray 'cause I used to cover Apple quite a bit and I see the same type of market reaction and kind of consumer, or not enterprise reaction to Nvidia that I saw to Apple, and that has a lot of staying power.

I'm just wondering, is that kind of, if you can maybe elaborate on what you said earlier about Apple and Nvidia and how they kind of are in the same spot and hence their success. Well, I don't have any insider information there. It's just apparent from the outside, um, that that's what I meant, right?

Yeah. Apple's, Apple's famously, um, famously successful by being run and really controlled by a product manager. Steve Jobs, everything flowed out from a very tight product management discipline.

They're also famous for being functionally organized instead of product line organized. And that, that I think is almost, it's very difficult to pull off. NVIDIA's not organized that way.

Um, but what it does is it allows for this kind of spirit and practice of product management in one area. Um, sometimes called silo, but product management in one area, marketing in another area, sales in another area, solutions and a third and so forth. Um, what NVIDIA's doing without that kind of structure though, is very similar, which is, um, really, uh, executing at a high level, uh, not just in each one of those functional areas, but in terms of, um, all of them working together so that there is no, there's no like c***k or hitch in a, an adopter and user's journey where they suddenly realize, oh, I gotta do this other thing that I don't wanna do.

Everything gets enabled. Hopes, description of the developer experience is a perfect example of it. NVIDIA's a chip maker, they're a chip maker, okay?

But your experience of them is through the software that you develop and use. They understand that. I don't know if that is, um, well enough appreciated around the industry.

It's a secretive Nvidia success, and that was the secret of Apple's success as well. Yeah, that was interesting. When, when Steve Jobs came back, it was 1996.

Technically they bought next in 1996, next or Apple did, Apple bought next. Yeah. Does that right?

That was December, 1996. It happened late after the B acquisition filter. This is like ancient history, but one of the first things when jobs eventually became the CEO and that was fairly quick, was he streamlined and stripped out.

They had like 80 variations of a Macintosh. He had a performa, I dunno if you remember those, all these different vari you said, we're gonna go down to a quadrants, we're gonna go down to like four basic products, but that's it. Yeah.

We're gonna simplify everything. So, you know, that's, that, that's a different time. But anyway, I just thought I'd throw it out.

Um, I do wanna mention one thing from the, from the, just to get back to the, the research report. Um, I hope you talked about generated that generative AI a few times that has gotten so much attention. Um, but one of the really fascinating results that was just like, I, I read it and I was like, oh my God, that is so true, and I've been forgetting it all along.

Um, the, the top, um, the number one, uh, uh, AI application in 2023, um, not generative ai, visual and audio analytics. Number two, simulation and modeling generative AI is one of the fa not the fastest actually, but one of the fastest growing. But despite all the attention it gets, it is not what predominantly this hardware is being used for right now.

It's being used simulation and modeling, but above that, uh, a visual, visual or audio analytics is the top use of all these chip sets in AI today. On that note, we're gonna go talk about something completely well, some another AI episode, uh, topic. I mean, what else do we have?

ai, ai, ai. We're gonna talk about responsible AI when we get back in a few moments. Discover Techron Group, the epicenter of tech innovation.

We are your go-to for reaching IT leaders and practitioners worldwide. Our secret impactful content that sparks awareness, engagement, and top quality leads with us. You'll access editorial websites, streaming videos, virtual events, custom content analyst research, and more.

Join our satisfied clients. Let's revolutionize your tech journey. Contact us today and tell your story to the world in the most powerful way with Textron Group.

Welcome back to Textron Gang. Accenture and Amazon Web Services just launched something called the Accenture Responsible AI platform powered by AWS. That is a mouth.

What a great name. Oh my God, it rolls off the tongue, doesn't it? Jesus.

Um, anyway, its mission is to help organizations automate and operationalize responsible AI practices. Uh, I know next to nothing about these types of things. That's why I'm so happy that Lisa is here as well as hope and guide to explain to us what this means and why it is so important.

And that's something I need to learn about. So, Lisa, I hand the baton to you. Sure.

So I get asked all the time, I post about this on X and LinkedIn and Instagram, explainable ai, responsible ai, and people say, come to me as a marketing person. Is that a marketing term? It's really not.

We're seeing this now platform approach from Accenture and AWS, but really it's, it's been for a while a framework or an approach that organizations can take to kind of help guide the design, the development, the deployment of ai aiming to build trust in AI solutions. Trust is currency these days. It doesn't matter what industry you're in, it absolutely is, and it needs to invoke responsible AI transparency, uh, accountability, explainability, fairness.

And so while it's not a marketing term, it's a, it's an approach or a process that's kind of, I think, emerging within the context of ai. But I think there's a lot of areas in which marketing can help. Um, on the transparency side, the accountability side, the explainability side, some of that comes back to the basic piece of marketing, and that's positioning.

But it's interesting this platform approach that's being taken. We know that both Accenture and AWS have thousands, if not more generative AI projects within their customer base. And I applaud what they're doing to kind of take more of a platform approach to the responsible ai, uh, framework.

Okay. Well, it's like, in a sense, does the platform help an organization so in assess AI risks and develop mitigation strategies to comply with regulations? Because over and over again, I keep coming across these types of themes, compliance, uh, the cost, the, the unintended consequences.

And this all kind of plays into this idea that it's gonna start in California, and it's probably gonna eventually go to other states. This, this idea of holding these companies accountable to how they use AI and the con and the unintended consequences of, of, of the use. I, I think, um, I think it's critical for companies to especially invest in transparency and explainability.

So if, let's say a, a finance company is sued because someone says, you know, we've been analyzing the patterns, and you are denying funding more for certain type of companies and others, um, we, we think there's bias at play. Now, if they have not invested in explainability and some way to make, uh, those decision making processes and models transparent, they may not have much of a leg to stand on. But if they have invested in that, now they know that, uh, they can produce accurate outcomes, they can provide the explanations for those outcomes.

Um, then how the AI arrived at the decision is a lot clearer, and they can have some way to push back or at least, um, maybe tune those models. Right. So they are fair.

Yeah, I mean, it's just, it always comes back to me in this kind of haste or rush to adopt AI and, and to, uh, spread it throughout an organization that what ends up happening is a lot of these companies, uh, there's a lot, a lot of, uh, angst, I guess is, is the word or, or, or miscommunication or disconnects on the intended use versus the actual use. It always kind of comes back to this idea that at the c-suite, they want a plan, they want something within a company, but they'll be, well, we'll attend to the mess when it happens later. Right.

Good. Oh, thanks. Sorry guys.

The stakeholder engagement just kind of pinged in my mind with what you just said, John, and I think that's where I look at this from a marketing lens. And what, what can CMOs do to facilitate the transparency, the explainability, the accountability? I think marketing teams can help engage with stakeholders, and we're talking the board investors, analysts press customers and kind of the broader community to ensure that the AI products and services that these companies are developing are done so, and developed appropriately and marketed as such.

Marketing can also come in and help from a responsible communication perspective, ensuring that what, how they're marketing these AI products and services are, um, transparent, avoiding exaggerated claims or miscommunication. So I think that communication piece around this is key, and it has to be responsible. I think that's, that's right.

I think that not only is it key, I think it, it may even be primary. I mean, I think I'm gonna put my Mike Ard, uh, cynic hat on right now since he's not, uh, leading this discussion because, um, I, I think that that may even be the primary and most common goal in general, uh, across, especially, you know, the, the, the, the largest companies who are really investing in what responsible AI initiatives and so forth. I don't mean it's pure lipstick on a pig or anything like that, but, um, when you think about, or, you know, when I think about responsible ai, and I'd like to know you're hopes thoughts about this.

I think it's really two things. Um, uh, it's, uh, uh, sort of representation slash protection on the one hand and transparency on the other. Mostly we talk about explainability accountability.

Um, but also, and a key part of this Accenture platform with, with AWS is, um, using guardrails to be able to ensure that, um, uh, uh, protected data gets stage protected, um, that there aren't inappropriate uses, that sources are appropriately a representative and like all that sort of thing. And this is all terrific, and I think it's, it's, it's absolutely what needs to be done. But the customers I talk to, they, um, they, they are halfway between fear and ignorance.

They just want to keep moving. That's the ignorance side. They see the benefits.

It's just, it's a whole lot like security and cybersecurity. Until the lightning strikes, they just, la la la la, la, everything's going great. But they have this fear in the back that lightning might strike.

They might get accused, they might, uh, get their brand damaged and all that other sort of stuff. So, so explainability, you know, responsible AI is a way out of this. The, the, the catch in my mind is that, um, there really is no auditability here in the legal sense.

These models just don't work that way. In the end. You cannot true.

You can, you can explain, and the explanations are great, but I think the legal framework we have around the world does not allow for something explainable to be legally defensible. And that's where I wonder, like, sort of where we're going in the end is kind of, nobody's gonna sit there and go, well, you know, it was trained. We use the right data and everything.

It ain't our fault. Well, it is your fault in the end. If you bought it or someone else trained it, you're, yeah.

So hope you're, you're nodding. What, what do you think? Yes, In God, this, this actually, this, this ties back to an experience I had at, uh, a former large employer where, uh, we wanted to start, um, capitalize capitalization of our AI projects instead of running a project nine months later, oh, we can start capitalizing because it's finished now we're gonna start capitalization as it goes.

Um, we had to have a conversation with Coopers because at that point there were not clear rules written for capitalization of agile projects, everything, capitalization, waterfall. What it came down to, which I think will be similar for this, is due diligence. Did you document somewhere what you were planning to do?

Did you do what you plan to do? How can you prove it? Right?

So now there, it, it looks sort of like a chain of trust has been built. One of the ways organizations can help with that is if they come up with, or, or find a way to develop some sort of governance and accountability framework. So now they have a way to document the tools, the processes, how the AI systems were developed, deployed, monitored, uh, what guidelines they used, what ethical guidelines they used, um, and then if they are audited, especially at the stage we are now, it's just, did you do what you said you were gonna do in the guidelines that you said you were gonna do it in?

I think that's the best defense any organization could have at this point. I just think the law isn't there yet. Hope I could be wrong, right?

I don't think I, this winds up getting tested. Lisa, you talked to, to CEOs and CMOs about this sort of stuff. Yep.

Is it similar to, to, to what hope and I are talking about? It is, it definitely is. I think where I'm seeing more of the AI implemented within organizations that I talk to, um, CEOs are leaning in and the, the Delta airline, CEO talked about how they're leaning into generative ai.

This was back in like December with how they really modified their, their, uh, ask Delta chat bot. I talk with marketers all the time about how they're using it for things like content creation, message testing, AB testing, um, content creation, really kind as a primary use case right now. Um, so what, what I don't hear from them is concerns about the guardrails.

I think they're piloting projects now. We need those projects based on our, kind of our first segment today to start getting out of the pilot phases so we can start understanding and investors can, where is the proof in the pudding? Where is the ROI?

But I think most folks right now are looking at it project by project function by function phase. That's what I'm seeing. And the guardrail conversations haven't come up yet, but you bring up a great point guy that that's probably something that needs to be next in terms of assessing how they're using it.

Are they using it appropriately? Because every organization, we know that this is a, a strong data foundation is essential for, uh, AI governance, for, for responsible ai. Every organization, including marketing, everything that CEO owns needs to be able to demonstrate the data that they're using.

Where is it coming from? Where is the store would being storehouse being used? And that's part of this whole responsible ai, I think, framework that has to be there.

And I think it's a little, I I wouldn't say cart before the horse. I just think we're not focused on that as much as we should be. And maybe the timing is becoming apparent now.

So, you know, guy, you mentioned, uh, Mike who heart in his, uh, cynicism, this is where his part comes in. So I actually came across some research from Rand Corporation as a global policy think tank, and they concluded that over 80% of these ai, next big thing projects are gonna fail, which is twice the failure rate for non-tech related startups. And again, it's what Lisa had said, it, it usually comes down to a misalignment of goals between the key stakeholders.

So leadership has this view of what AI can or should achieve that's not really grounded in reality. And instead it's driven by this kind of preconceived notion of what AI is. And then it gets down to the rank and file where there's not enough upskilling or reskilling.

There are, uh, people scrambling to strategize, including the poor engineers. You're kind of stuck in the middle. So that's just a word of caution or a word of warning that we're gonna see this huge gold rush land rush.

And there are gonna be a lot of people who are gonna end up being bankrupt or, or, or not gonna find a kernel of gold anywhere. I don't know about bankrupt. Uh, you know, uh, the implementers of this, um, the ones who are, uh, uh, obeying the corporate masters and, and just going, they, they know when there's a wind at their back, they sail.

And I think that an 80% failure rate, um, especially, you know, post agile, post DevOps post, you know, uh, um, beta release, you know, MVP, all that, that world that we're in right now, 80% failure rate is not seen as failure. 80% failure rate is, this didn't work. We learned we're continuing, uh, I kind of shrug and just like 80%.

Okay, great. You know, it's early days. The 20 percent's gonna be maze balls.

Yeah, Exactly. I'm gonna say 20% is actually pretty high percentage, right? When you think about starting a business of any type, that's a high percentage.

I mean, compared to other industries like the restaurant industry or what have you. So yeah, it's, it's interesting. You, it's, it's wise though to just kind of flip that number around.

We'll see. I mean, we'll see. Yeah, it's, uh, is, is, is things are, are, are continuing to, to, to move along.

I mean, we're seeing no end in sight or slowdown in what's going on is, as you alluded to in the, uh, the futureum analysis. I mean, we're gonna have five years of super strong growth at the very least, so we shall see. But, uh, on that note, we're gonna wrap up our show for today.

Thanks so much for joining us, and please stay tuned to Techstrong TV for a daily deluge of news on AI security, DevOps cloud, what have you. Thanks to our panel. You were all great and, uh, want to wish everyone a happy, great holiday weekend.

Take care. Yeah, Great. Long weekend everyone.

Yeah, thanks. Thanks Everyone. Yeah, thank.