Nvidia VP Manuvir Das says enterprise AI is at ‘an inflection point’ in the wake of GTC
Join top executives in San Francisco July 11-12 to hear how leaders are integrating and optimizing AI investments for success, learn more
Amidst the dizzying array of product announcements at Nvidia GTC today, I was particularly interested to hear how the news could impact the adoption and application of AI in the enterprise in 2023.
I was happy to get the scoop straight from Manuvir Das, vice president of enterprise computing at Nvidia, whom I spoke with last month for an in-depth look at Nvidia’s rise to AI dominance over the past decade. This one-on-one interview has been edited and condensed for clarity.
>>Follow VentureBeat’s ongoing Nvidia GTC Spring 2023 coverage
VentureBeat: How would you characterize the GTC announcements in terms of how they are going to impact AI in the enterprise?
Join us in San Francisco July 11-12, where top executives will share how they’ve integrated and optimized AI investments for success and avoided common pitfalls.
Manuvir Das: I would say that enterprise AI is at an inflection point. Generative AI and ChatGPT have now made it so clear that AI can change every business. We’ve seen this coming for a while, and [have pushed] companies to adopt AI, but I think ChatGPT really created that iPhone moment, if you will, for enterprise companies to say ‘I need to adopt this into my business and into my products, Otherwise I will be left behind.
If you look at many of our announcements, they are in tune with the moment. DGX Cloud [which democratizes Nvidia DGX AI supercomputers and is optimized to run Nvidia AI Enterprise] There is a place where you can get the ability to do your training.
The Nvidia AI Foundation is a place where you can customize Generative AI for your company and then deploy it. Nvidia offers AI enterprise software, the Nemo framework, which you can use to train your own models in the cloud.
The new portfolio of chips we’ve announced are all about the inference of generative AI. Our Hopper H100 was designed in large part specifically to do a really good job on these Transformer models that are behind generative AI, and Hopper is fully in production and through different channels like different clouds and OEMs. has become available since.
So on the enterprise AI front, the whole theme of GTC is to say this is the moment for enterprises to really embrace AI. For all those enterprises out there who are on the fence and wondering, is there a compelling business use case to justify investing in AI? I feel it staring them in the face now.
VB: What was the biggest pain point for enterprises addressing these various announcements?
Slave: The first pain point is that AI is a complex thing. For the very small company or single data scientist, there are great turnkey solutions. You can use SageMaker on AWS or Vertex AI or Azure ML and get up to speed very quickly. But the moment you move up to like a single GPU, or a single node, the solutions don’t really scale easily.
On the other hand, you can be a really big business of AI consuming thousands of GPUs like Meta or OpenAI and you have your own team of hundreds of engineers who can deal with all the complexities and do all the engineering work really I need to use that large scale environment.
The difference is for a typical enterprise company trying to adopt AI. You need a solution in the middle like 50 or 100 GPUs so it scales but it’s relatively turnkey. So you can just do your data science but you don’t need to do engineering. That’s what we addressed with DGX Cloud.
The second pain point is that generative AI feels great. ChatGPT is awesome. It’s just a giant model for the whole world. But an enterprise company needs a model that is useful in their context, with their data, their customer information, and their employees. So they needed a way to build their own custom models, but they couldn’t start from scratch.
The kind of model they need requires information from the company in order to accurately answer questions, but it also requires common sense, which is how ChatGPT learned from the Internet.
So we believe what was needed in the market is a system for enterprise customers. We have some pre-trained models that we’ve trained on the internet enough that they have common sense and ability to interact, going back and forth. And then we give you a way to optimize this content with your data, so that the model can actually be used in your processes to provide the right answer, and express it properly, intelligently. That’s why we created the Nvidia Foundation.
The third pain point was that for enterprise companies in general, if they are going to adopt AI in production, they need some reliable platforms. For me as an enterprise customer, I use SAP HANA or VMware or Microsoft Office 365 or Windows Server or what have you. These are platforms I can trust — bugs will be fixed, there’s a release cadence, things are proven. Whereas today, a lot of AI I just lift it from open source, that is, from open source, tie it all together and hope for the best.
We address that by taking our Nvidia Enterprise Software Platform and putting it in the cloud so you can consume it when you have an instance – Google, Azure or AWS or wherever you choose to do your work. So you can actually run a reliable software that has done the work of bringing all the parts together.
VB: Which of today’s Nvidia GTC announcements are you personally most excited about?
Slave: It’s tough to play favorites, but I definitely think DGX Cloud is going to be a huge advantage for enterprise customers because it’s going to be a lot easier for them to really scale that middle-ground. Nobody’s really paying attention to them because they’re not like a unicorn renting out 10,000 GPUs and everyone’s attention is on them. But I think enterprise customers are the ones that will really see the benefits of AI being carried over to their customers. They need some love, they need a better solution.
VB: Obviously GTC has historically been talking to the millions of developers who are going to be using these tools. But I’m curious — if you had a group of C-suite leaders in a room, people who know more or less about Nvidia, what would you tell them about what they need to know about enterprise AI and Nvidia right now? need to know?
Slave: I think what they need to know is, number one, this is a full-stack problem. It’s not just about the hardware. You need the hardware and the right software optimized all together to make it work. And I’ll tell them that, unknown to them, Nvidia has spent the past decade discovering all the different use cases that AI can enable, and the software stack to enable all of these use cases. are building.
So if they’re looking for a one-stop-shop company to talk to or work with on how to steer the AI journey, there’s no better bet for them than Nvidia, because we’re the one who did it. all this. You talk about a company like OpenAI that has ChatGPT, and kudos to them. How did they get it? Because they worked closely with Nvidia to design the right architecture, so they could do all their training and build their models.
VB: What about the people out there who are melting down their GPUs right now? I mean, are there enough GPUs for all this to happen?
Slave: It’s our job to provide lots of GPUs. But it’s an interesting thing you say about melding, because I think another thing that’s been different in traditional enterprise computing over the years is the general conventional wisdom that you over-provision. You don’t want your CPU running at more than 50 to 60% of its capacity. You leave a lot of headroom there. And just think about the waste that comes from that.
But because GPUs were designed from the start to solve the problem of performance, we always put GPUs in a mode where they’re designed for people to actually use them. When we do our full burn-in testing for every new GPU that comes out before we put it into production, we subject it to intense loads where it’s running at max load for long periods of time. , because we know that’s how people are going to use them. So I think there will be other benefits when it comes to GPUs. I’m sure people are trying to melt them down but I think we’re good.
VentureBeat’s mission To be a digital town square for technology decision makers to gain knowledge about transformative enterprise technology and transactions. Find our briefing.