The Art of the Wait on Artificial Intelligence

[00:00:00] This episode is brought to you by Vanta. Stressing over cybersecurity, whether you're a startup, growing fast, or already established, Vanta can help you get ISO 27001 certified and more. Plus, it allows your company to centralize security workflows, complete questionnaires up to five times faster, and proactively manage vendor risk to help your team stay compliant. Head to Vanta.com slash Spotify to learn more.

[00:00:31] Welcome to Tech News Briefing. It's Tuesday, March 25th. I'm Victoria Craig for The Wall Street Journal. We spend a lot of time talking about the ways artificial intelligence platforms can make life easier and the pace at which people and companies are adopting it. But not everyone is getting in on that trend. Today, stories of the not-so-fast adopters, including some corporate executives who aren't quite buying NVIDIA's push to upgrade to the latest chip rollout.

[00:00:58] Then, if in the frenzied race to leverage AI, your company has been slow to integrate its varied uses into day-to-day operations, one researcher says that might actually be a good thing. But first, new product launches can be hard to resist. And that's what NVIDIA is hoping will be the case when it comes to the latest update to its Blackwell chips, which help power AI systems.

[00:01:26] CEO Jensen Wong took the wraps off his company's latest technology, which is the successor to its Hopper chips, at last week's Developers Conference. And he did his best to convince company decision-makers in attendance that now is the time to level up. There are circumstances where Hopper is fine. That's the best thing I could say about Hopper. There are circumstances where you're fine. Not many. If I had to take a swing.

[00:01:56] But WSJ Enterprise Technology Bureau Chief Stephen Rosenbush talked to some executives at the Developer Conference, and not everyone is rushing to order the latest and greatest technology. So Stephen, who are some of these companies, and more importantly, why are they waiting it out? The companies are making decisions that are very specific to their particular business.

[00:02:18] It's not that easy to generalize about which companies are keeping up with the leading edge of NVIDIA's chips and systems, and which ones are more content to sort of hold on to an earlier generation of chips. But I did talk to several companies, including HPE and Ford.

[00:02:40] And in both instances, they're content to work with pre-Blackwell chips, Blackwell being the current generation of NVIDIA's system, with an eye toward potentially upgrading in the future. Antonio Neri, the CEO of HPE, not a small company by any means whatsoever, said that he had 250 pre-Blackwell chips and that he had more than enough computing power to run his entire company internally.

[00:03:07] I think it's helpful to think of this sort of situation between NVIDIA and its corporate clients from a consumer standpoint. So if I see Apple or Samsung is launching a new phone, I have this sort of FOMO. I don't want to miss out on the latest, greatest piece of technology. But also, Apple or Samsung wants me to buy that phone because they've got a bottom line to think about. So when we think about this from NVIDIA's perspective, what is their pitch to their corporate customers to buy this latest generation of chips?

[00:03:36] The pitch to customers from NVIDIA's point of view goes something like this. In the aggregate, companies are spending more and more money on AI infrastructure every year. But the unit cost of computing within AI is going down dramatically. And the company ought to stay on the leading edge of that price performance curve to keep their budget from exploding.

[00:04:02] So if you're buying more and more and more computing power to process AI, you're going to spend more and more money unless the unit cost of that computing comes way down. And the really compelling thing from the buyer's point of view about NVIDIA's latest technology isn't simply that the performance is so much greater, which it is, but that the price performance is so much greater.

[00:04:28] And NVIDIA's argument is that you need to stay on the leading edge of that price performance curve or you will be crushed by the economics of the technology. NVIDIA's boss Jensen Wong acknowledged some aspects of this at the developers conference last week. He said that building out company AI infrastructure takes time, planning and billions of dollars. Is this him sort of trying to manage his own expectations?

[00:04:55] Because as you just said, some of these companies, they don't necessarily need the latest and greatest right away. While demand for Blackwell is strong, it is not ubiquitous and it is not universal. And the need to really pay attention in the coming weeks and certainly going into the next quarter to see how that demand holds up. It takes time and it's something of a negotiation.

[00:05:18] While NVIDIA will say that all companies should have the most powerful price performance infrastructure right now, they understand that not every company has the same needs, but that every company really does need to have a clear understanding of where the product roadmap is headed so that they can plan and they can upgrade as it suits them. That was WSJ's Enterprise Tech Bureau Chief, Stephen Rosenbush.

[00:05:48] Coming up, if your company lacks a strategy around artificial intelligence, that might not be a bad thing. That conversation after the break. Here's the truth about AI. AI is only as powerful as the platform it's built into. ServiceNow puts AI to work for people across your business,

[00:06:15] removing friction and frustration for your employees, supercharging productivity for your developers, providing intelligent tools for your service agents to make customers happier, all built into a single platform you can use right now. That's why the world works with ServiceNow. Visit ServiceNow.com slash UK slash AI for people.

[00:06:42] Being an early adopter isn't necessarily all it's cracked up to be. At least, that might be the case when it comes to developing a company-wide strategy around artificial intelligence. WSJ contributor Joe Pepperd is a professor at University College Dublin's Graduate Business School. He says rushing to centralise AI resources and expertise is actually, in many cases, a mistake.

[00:07:06] Joe, you have four reasons for making this assertion that companies are really wasting their time investing in and implementing AI strategies. And one of them is that most companies just aren't ready for it. Just walk us through why. There's a lot of hype out there, a lot of fraud in relation to AI. AI is sort of presented as this mystical technology.

[00:07:30] And it's referred to in a sense that nobody really talks about, well, what sort of branch, let's say, of AI are you referring to? Are you talking specifically about large language models? Indeed, are you talking about chatbots? Are you talking about, let's say, computer vision, machine learning? It's just used now as a kind of pejorative term.

[00:07:49] And in my work with executives, I kind of find that a lot of them have, to some extent, have bought into that sort of Kool-Aid and are seeking to, you know, build kind of an AI strategy because they kind of have this fear of missing out. That sort of led me to think, hang on, that's probably not the right path for you and your organisation. So if I look at the first reason and that company's just aren't ready, they really just haven't done the foundational work.

[00:08:16] So as I say in the piece that if an AI strategy was somehow to land in the inbox of a CEO, they just wouldn't be able to implement it because of that foundational work. So how do companies develop an AI strategy from the ground up? Is it this idea of digital maturity that you talk about where you sort of dip your toe in first? First, you use AI as a tool to do some little tasks and then you develop an overarching strategy from that?

[00:08:43] When you talk about using AI to do some kind of simple tasks, you're probably thinking about kind of large language models. We have seen employees experiment, let's say, a lot of times unbeknownst to maybe their managers that we know can cause a risk for companies, particularly if they are putting kind of confidential or copyrighted customer or indeed employee information, uploading it,

[00:09:10] and then perhaps prompting maybe ChatGPT to, let's say, summarise a company document, for example. You would hope that at this stage they would be abreast of the risks of doing that, not just in terms of risk around privacy and confidentiality, but also in terms of the fact that we know that these large language models tend to hallucinate, which is sort of a polite way of saying make stuff up.

[00:09:35] I would have any issue with, you know, employees, let's say, experimenting with the likes of large language models, just to maybe understand the technology, to get a sense of its capabilities, and also maybe to begin to explore how the technology might be used in their day-to-day work. So how do companies know when to really go for it? Or are there some industries maybe where they just shouldn't?

[00:10:01] AI is just one of a number of technologies, but it's not so much the technology, it's the use to which the technology is put, that that's the transformational piece. So we've seen for many, many years a lot of manufacturing companies, industrial products companies, for example, have shifted their business model away from actually selling products to customers to selling services. Now, to deliver that value proposition profitably, they actually do need data.

[00:10:31] And it's generally data about the product in use and the ability then to collect that data, analyse that data, and then quickly make decisions in respect of that particular product or that particular asset. Because with this kind of shift in a business model from product to service, that means there's also a shift in risk. We see that the value proposition is based around outcomes, let's say.

[00:10:56] And of course, if the product is unavailable, for example, if you're an airline and we've got engines on the wing and you have availed of a value proposition that is based around what Rolls-Royce would call kind of power by the hour, the availability of that engine. If that engine is unavailable, that actually is now something that the engine manufacturer needs to be concerned about. So in essence, it's really about being specific to your industry and to your business

[00:11:25] and figuring out which tools fit better rather than approaching AI as a whole, a fix-all for something or solving for some problem that you're identified. I tend not to refer to AI as a tool. I sort of see it as a technology. You might say it may be a large language model, maybe a chat GPT is a particular tool. I see it as a technology with particular capabilities. And obviously, AI is an umbrella term. It's bringing different capability. And that's the challenge.

[00:11:54] That leadership challenge is to marry the capabilities of technology with business opportunity. That's the strategic challenge and the strategic opportunity. That was WSJ contributor Professor Joe Pepperd. And that's it for Tech News Briefing. Today's show was produced by Jess Jupiter with supervising producer Emily Martosi. I'm Victoria Craig for The Wall Street Journal. We'll be back this afternoon with TNB Tech Minute. Thanks for listening.