Dell exec looks back at history when considering AI's future

As Dell Technologies looks to either provide or support all aspects of a customer's AI journey, Dell's senior vice president of AI strategy, Matt Baker, is also thinking about Dell's AI path and beyond.

As with most technology vendors, Dell is focused on AI -- more specifically, its Dell AI Factory that offers a portfolio of infrastructure, services and partnerships for AI development and deployment. The company's approach to AI was fully on display at the recent Dell Technologies World, where executives talked about AI's significant disruption and compared it to the Industrial Revolution where machines introduced new ways of working and transformed the economy.

Matt Baker

In this Q&A, Baker provides more details about how the Dell AI Factory fits into the company's strategy, talks about how AI affects storage and explains why it's important to look to history when considering AI's future.

Editor's note: The following was edited for length and clarity.

How does the Dell AI Factory set itself apart from other AI offerings?

Matt Baker: Let's go back 15 years to the early days of cloud and virtualization. There were so many different platforms to choose from; there was so much complexity. People didn't know where to start. [The AI era] is that times 10. There is no dominant stack that is complete; it is all an assembly of componentry. If we can help customers from making the mistakes that we're making internally -- because we've developed stuff internally -- then we saved them a bunch of time, effort and energy.

At a time when legacy vendors offer more cloudlike subscription models, Dell has continued to highlight its hardware. Will the company continue offering Capex and Opex models or will they be consumed by the AI Factory?

Baker: The AI Factory will be included in Dell's offerings; it won't take over them. Think of Apex as a way of transacting business. You can transact AI Factory, Capex or Opex as you choose. There will be a vehicle through which AI factories will be either a capital purchase, a lease or a consumption model.

Why have two separate portfolios? The Dell AI Factory with Nvidia and the Dell AI Factory separate from Nvidia.

Baker: The one that is in the market already is the Nvidia variant. But there will be other variants that include different hardware ingredients and different software. The other variants will be easy to consume because they, too, will be built for purpose versus you having to assemble them from [scratch].

Dell introduced a new unstructured data storage array, PowerScale, and a new structured and unified data array, PowerStore Prime. How does storage need to change for AI?

Baker: It depends on the workload. If you're training a foundational model, the storage architecture need not be fundamentally different, but it needs to be tuned to feed a significant amount of information into GPU clusters.

What I see changing more rapidly than core storage -- fundamentally, data is stored as zeros and ones either directly on disk in a file system or in an object platform -- is the broader architecture that storage touches. Think data management writ large. Not just data lakehouses and databases, but everything that falls under the umbrella of data management. We are seeing a blossoming of new database types and new disciplines such as information retrieval technology. I think that will influence what rides above storage and the solutions that we need to break into the market to facilitate that.

No need to reinvent the wheel for storage?

Baker: We've been reinventing the storage wheel for some time. What drove the adoption of shared storage systems was largely virtualization. In order to do vMotion, you needed shared storage systems. Storage has always reacted to the overlying architecture it needs to support, and it will evolve. Everything between the app and the storage will change -- and change in a way that needs to be more orchestrated.

AI is a transformative technology, so why are we mainly seeing more and more chatbots?

Baker: AI can do a lot more, but I think everyone has to get very familiar with it. I think the reason why you see chatbots, or information/knowledge assistants, is that it is the lowest-hanging fruit currently available.

A lot of people want to talk about generative AI as if it is different from AI, but it's not. You can't separate them. The dominant use case, knowledge assistants/chatbots, are typically built leveraging a [retrieval-augmented generation] pipeline. RAG pipelines combine semantic search, graph search, all machine learning technologies that have been around for a while. Combine that with generative AI with an easy-to-use interface -- that allows you to interrogate big data sets in a way that's conversational, getting to the right answer faster.

The way we see the world of AI and generative AI is largely through the lens of consumer use cases that are dominated by chatbots. The first project that people do internally is, 'Let me make a chatbot for my data,' and that is very useful and effective.

Before customers begin making chatbots, what safety policies need to be implemented? If something goes wrong, how much responsibility falls on the vendors?

Baker: I think the world is reacting to a change in the behavior of systems that AI represents. Before, we had systems that were deterministic -- you put inputs in and you get inputs out. That's not what's happening [now]. You have a system that is capable of reaching into wells of information and knowledge, and generating content that comes out the other side.

As for who's responsible for putting in safeguards, let's take General Motors as an example. It's not just General Motors, it's General Motors and a big ecosystem of automobile parts manufacturers; it's an integration exercise. But we don't punish an automaker unless they do something egregious. ... You don't punish the automaker if someone drives drunk and kills someone else.

There's a degree of societal responsibility that gets built up at all layers. Do we have a part to play? Absolutely. What we're really interested in doing is building a framework for customers to consider their own responsible use policies, and it includes a number of different factors. We want to help walk them through what is their AI policy, not [create] a dictated policy.

This is going to be a shared responsibility across the entire industry -- not just the industry of people who make it, but also the people who are using it. No one company is going to be able to solve this problem unless they're solving the entire problem end to end.

A common concern with AI is that it will displace or replace human workers. Is this a consideration as the push for AI increases?

Baker: When in the course of human history has technology durably taken jobs away?

Messenger boys, switchboard operators, lift operators ...

Unless [AI] is fundamentally different than every other technological wave that occurred, out the other side of it, there will be more economic activity, more jobs, more prosperity, and the human condition improves.

Baker: But the net effect was the technological advancements drove economic activity, which created even more jobs. If you look at the history of technology, there has not been a technological shift that has durably eliminated jobs. Instead, it led to more prosperity and economic activity that led to more jobs overall.

Are there temporal dislocations that occur because of disruptive technology? Absolutely. But the history of technology has been a massive net positive to society. When I hear that question, I ask myself, philosophically, have we had a period where jobs go away and no jobs get created? Do we need more reskilling, upskilling as a society? Absolutely. Dell just launched a big internal education/upskilling initiative around AI.

Unless this technology is fundamentally different than every other technological wave that occurred, out the other side of it, there will be more economic activity, more jobs, more prosperity, and the human condition improves.

Have you seen AI replace a human in the workforce?

Baker: You can find individual examples of people being replaced, but it didn't render that person unable to find other work. I'm not trying to be flippant or dismissive of the pain that causes. The movement will lead to more opportunities. Will there be examples of displacement along the way? Yes. But that's the history of human progress.

Adam Armstrong is a TechTarget Editorial news writer covering file and block storage hardware and private clouds. He previously worked at StorageReview.com.