Enterprise AI:
Data Democratization and Cloud Computing

Michael Krigsman: Today on Episode #804 of CXOTalk, we're talking about the democratization of technology and the impact on AI. Our guest is Bob Muglia. He is the former CEO of Snowflake and has had a legendary career. I'm here with my amazing guest co-host QuHarrison Terry. 

Bob, as we introduce you, I will hold up your book. It is called The Datapreneurs. You have this extraordinary background in the computer industry. 

Bob Muglia: I joined Microsoft and spent 23 years there. The first technical guy on SQL Server. I helped to build that business.

I spent all my time in the product groups. Ran, in some senses, almost everything at Microsoft for a little while, except for games. Never did that. 

I spent time in Widows Server. I ran Visual Studio. Helped to put that together. I spent time in the Office group and MSN. 

Then, for the last seven years, was running the server and tools group. Was president there and grew that business to about $17 billion. 

Since then, by the way, it's gotten a lot bigger. That business has grown a lot since then. I helped work with Scott Guthrie at Microsoft now, and it's many times that size now. It's pretty remarkable.

After that, I spent a couple of years at Juniper down in the Bay Area. Then I joined Snowflake in the middle of 2014 and ran the company for five years.

I took it from zero revenue. That's an easy number to remember. It's not hard to remember zero. And took it to just about $200 million in revenue before I moved on.

What is a datapreneur?

QuHarrison Terry: What would you classify as a datapreneur?

Bob Muglia: I'm working with entrepreneurs now. What I've been since I left Snowflake is really helping small companies to grow and investing in small companies. I'm on about five boards of private, small companies all involved in data in one way or another.

I realized that when I was at Microsoft, even though I was working for a really large company, I was working with entrepreneurs the whole time, data entrepreneurs or datapreneurs, really. My role has always been pretty consistently to take, help, and work with these brilliant people that have built something truly amazing and help them to turn it into a viable product that sells in the marketplace. 

That's kind of my sweet spot is helping to define the technology so that it's something people want to buy, and then helping to bring it to market, price it, package it, and sell it in the marketplace in a way that makes it successful, which is why I've been focusing on really early-stage companies because that's what they do. They're trying to build things from the ground up.

I realized, when I was at Juniper, I tried to fix a bunch of things. It's really great, and I really respect the people that go into broken areas and like to fix them. I decided that I prefer building things instead of fixing somebody else's broken something, and so that's what I've been working on.

Michael Krigsman: Do you want to give us just a very brief overview of the book and why you wrote it? 

Bob Muglia: I felt like I had something to say. I wanted to put it in a way that was reasonably easy for people to read, so I worked with a co-author, Steve Hamm, who was extremely helpful. Frankly, I could never have gotten the book done without Steve's help. We decided to focus on the people aspects of this, the datapreneurs that were actually building this incredible software. 

The arc of data innovation

One of the key elements of this book is something called "The Arc of Data Innovation." This is this idea that although we look around us and we see AI as this big, new thing, the reality is that while it's new to us, the technology has been built and created over a period of time. 

It's all built on incredible work that's been done over decades and decades, a period. And so, I told that story in The Datapreneurs, and I describe it in this context of this arc of innovation, which is key technologies that have been invented over the last 40 or 50 years that have led us to where we are today. We're now in a period where this work that's been done over these decades is really paying off in some very significant ways, as we have this new AI technology that we're all looking at and thinking about, "How is that going to affect our lives?"

Data management in 2023: Startups vs. legacy enterprises

QuHarrison Terry: How should I organize data in my organization if I'm starting at zero versus if I'm already an incumbent with a lot of legacy technical debt? What are the best practices? 

Bob Muglia: Well, it's easier when you're starting from scratch because, if you start in 2023, you have the incredible, modern data stack to build on top of. The modern data stack are really a set of services that work together to provide Internet scale, incredibly large-scale working with data, allowing people (companies of really any size) to work with all of their data.

Now it's really all of the different types of data. Classically, it was structured data coming out of business systems, as well as potentially semi-structured log data that people were analyzing to understand the behavior of applications. 

Now we're in a world where there are new data sources that are super-interesting like video and documents and speech. All of these are elements of data that are now being collected as a part of companies and can be analyzed.

What I would do, I mean if I'm starting from scratch, it's pretty simple. You subscribe to a bunch of SaaS services. 

Today, businesses are run on SaaS services, especially small businesses where they run very little in-house and in-house data centers. Maybe nothing, probably, in fact. Everything is up in the cloud.

You work with Salesforce, and you work with some product support company, and you work with Datadog or something for your operations. You have all these different SaaS companies you're working with that are running your business, and all of those are generating data.

Typically, you have some business system that is a core part of what you've created. That can generate a lot of data.

Now it's pretty straightforward to take products like Fivetran, which allows you to take that data off of those SaaS systems and collect it into a centralized data warehouse. Products like Snowflake, or many other products in the industry because there are really five vendors that provide the modern data stack.

Snowflake and Databricks plus the three cloud vendors (Amazon, Microsoft, and Google) all have their own offerings in this space. You choose a platform, you choose a data pipeline vendor, and you start working with data. 

You choose a BI tool (Tableau, Power BI, something like that), and you can start working with data. You can begin to do machine learning. 

Everything is available for you. Everything can be in one place. It's much easier.

In 2014, it was so much harder. It was so much harder before that existed. Now it's all coming together for people.

Michael Krigsman: Please subscribe to our newsletter. Hit the subscribe button on our website. Subscribe to our YouTube channel. Check out CXOTalk.com. We have incredible shows coming up.

Democratizing data and the growth of AI

Bob, this concept of democratizing data is so important. What does that have to do with creating the AI growth that we have today?

Bob Muglia: This idea of making technology available to everyone has been core to what I've certainly been focused on in my entire career. 

When I look back at what we did (and my teams did) at Microsoft in some of the early days (the 1990s and into the 2000s), what I'm very proud of – and maybe one of the things I'm the most proud of – is that when you looked at technology before that timeframe, it was really only available to the largest companies. 

People would buy mainframes, or they'd buy these big, expensive mini-computers. Hundreds of thousands or millions of dollars of outlay in order to buy the technology required to run a business. 

Well, Microsoft changed all of that in the 1990s with products like Windows Server and SQL Server, both of which are products I spent a lot of time on. That allowed companies of essentially any size to run a business.

Go back to 1990. Your dentist was totally paper and pencil. The small business, the dry cleaner down at the corner, was probably also paper and pencil. That's how business ran in that timeframe. It was in the 1990s when technology like Small Business Server, Windows Server, SQL Server, all of those technologies packaged together that allowed companies to build applications that ran, were used, and were available to small businesses.

Now, of course, you look everywhere and everything is computerized. Now it's services. But back then, it was on-premises systems running on a little server in the closet that was in those rooms. People could do that for tens of thousands of dollars; much, much less costly than the alternatives that were available. 

Fast-forward to today. We're in a services world where everything is a service and everything can be purchased through some sort of subscription. You just apply to it and you pay for what you use. Now it's very cost-effective, or at least reasonably cost-effective, for organizations of any size to work with information and to treat information and data as one of their core assets, which I think everybody should do. 

The data you have and that you collect is a critical element of everybody's business. Now it is reasonably straightforward to make use of that with these tools like the modern data stack and the products that are part of that. 

Now we have this whole set of artificial intelligence models that can be built on top of this data foundation that can be set up in organizations. What that does and what's changed in 2023, which really didn't exist 24 months ago – this is pretty new – is that we now have, with these artificial intelligence, this idea that there is literally intelligence inside a computer that you can take and teach to do things on your behalf. 

The way I often describe this is that people have skills that they've learned in a given domain, and they understand the attributes of that domain: how things work, how this talks to that. That's knowledge that everyone has and it's up in their head. What's now possible for the first time is to kind of bottle that knowledge, to take that knowledge that you have and stick it inside one of these AI models and allow that model to do much of that work for you. That provides a whole new set of opportunities and makes things possible that just were not possible, like I say, 24 months ago. It's pretty exciting. 

Corporate culture and AI

Michael Krigsman: On the one hand, you've got this ease of access because of all of the SaaS tools and the lower costs, lower barriers to entry that you were describing earlier. However, AI is so different because of the open-ended nature of the result. And so, where does the cultural element come into play that, "Okay, we have all of these tools, but our organization is hierarchical, we're structured, and we have silos. We want predictable results, and AI is not doing that"? There is this cultural element. How does that fit in? 

Bob Muglia: One of the most important things I've said is that when you were building these services that you purchase or you create for your internal users, those services (in one way or another) imbue the values of your organization. You can see that in decisions that get made, in the way those products are offered, and the kinds of things that they present to end users.

The culture comes through. The values come through. 

This has been true for a long time. I could see it in products that I knew well. I could understand how decisions that were made by people and the values of those people were reflected in the products that they built. 

You can actually look at that and understand that if you can trace it back to the people that built it. Now multiply that by 100 because the values of the organization are going to get programmed into these models that get created. 

It's important to really understand your values. That's something that I think is incredibly important for every company to do. I encourage every small company I work with to build their values and really live by their values. 

That was something that was very important to me when I built Snowflake was that it was going to be a values-based company. Fortunately, the team, they wanted it, and we created a company that people liked to work with. It wasn't just a good product but it was a good company as well because it was a values-based company.

Well, those values are now going to get imbued in all of the AI that gets created by every company. I think this is just an opportunity for people to really understand what they're all about and then take the core of that and implement it in these models. Now it starts to become reproducible because what we're going to start to see is a lot of questions are going to get answered by these models.

One of the areas of initial progress that the technology is really ready for right now is to solve problems like helping people to answer product support questions without having to go to a product support specialist. Models are really good at answering questions like that. And if we can augment the models with knowledge, we can make sure the models answer the questions correctly for the customer.

But again, the values will begin to show through in this, so it's really important that we think about that and build it that way.

QuHarrison Terry: When you go back just ten years ago (in 2013) and look at it, you were ruminating on the concept of starting Snowflake and getting going in that engine. You looked at the industry, and you said, "This is where we're going. We're going to be more data-centered, and these are the things that are missing."

In your book, you talk about the next 20 years. There's a point where most futurists, they show the exponential chart, and they say, "This is the singularity." 

However, in your book, you added a few steps in between there. You said, "Yes, this is where we're at with ChatGPT and large language models. But before we get to singularity, there are all these fundamental steps that need to be filled in." 

Future advancements in AI and databases

Where is the future thinker in Bob at today and what do you think a technologist needs to solve for the next ten?

Bob Muglia: Well, in the short run, I think what everybody is really focusing on in the next 12 months or so is to take this technology that exists in foundation models (primarily large language models) and apply it in applications. We're mostly waiting for the apps right now.

We've had this incredible hype cycle in the first six months of this year. I've never seen hype bigger than the AI hype, which is good, I think. 

It deserved it. I actually think it deserved it. I'm pleased it happened, but it did. 

Now I think we're kind of at the peak of the hype. In Gartner terms, we're sort of starting to enter the trough of disillusionment right now as people wait for the applications.

But in this case, I think, again, following Gartner, I think that's going to be a short trough. Pretty soon, next year, we'll be onto the slope of enlightenment where people begin to understand how the technology will really be applied in their business. I think that will start to happen in the next year as we see more products become finalized and people can begin working with them; the Adobe products, the Microsoft products, plus the incorporation of AI in virtually every other product that people are using. 

I think the technology that sits behind the AI is improving dramatically so that, over the next 12 months or so, people will be able to leverage this AI as a part of their modern data stack solution using Databricks, Snowflake, Fabric, or BigQuery – whatever they want to use – and to work inside that environment to actually build AI applications for themselves. They'll understand better how to do it. We're at that stage right now where AI is being established into the industry. 

Between now and then, I see advancements in databases. A lot of my focus has been on how databases will change in the next ten years.

I think relational technology is ready for a breakthrough in the sense that it's ready to begin to leave SQL behind. SQL and relational have been copesetic and tied together since IBM invented both in the 1970s. It's great, and SQL is fantastic. It's very appropriate for working with structured data. 

But now there are all these different kinds of data. We have semi-structured data. We have complex data in the form of videos and documents and things. 

Relational technology can apply to that but we're being held back in some senses by SQL. While I see SQL continuing to be incredibly important; very, very critical; still the standard for working with data and slicing and dicing data; I think we'll begin to see products that provide much more sophisticated database technologies that break free of the box of SQL.

Future of data, models, structured, and unstructured data

QuHarrison Terry: To the point where we'll have just unstructured databases that still can give us the same outputs as SQL? 

Bob Muglia: There's no such thing as an unstructured database. There's no such thing truly as an unstructured file. What would an unstructured file do? 

People call this unstructured. It's not unstructured. It's complex structure. 

A video is complex structure. A document is a very complex structure. 

These are not unstructured documents. These are not unstructured things. We've just talked about them that way because they've been opaque – not to us. 

You can take a picture, an unstructured picture of a horse eating hay in a barn, and a four-year-old can identify that. But until a few years ago, a computer just thought it was a bunch of bits. Now, with machine learning, a computer will identify that as a horse eating hay. 

It's not unstructured. It's now data. All of these formats are rich opportunities with data. 

Relational is a very broad set of mathematics. It can be applied to data of any shape. 

Where we've worked with data in the form of tables, we can now start to work with it relationally in the form of semi-structured documents as well as any shape. That's where this idea of a knowledge graph comes in that you can take and create a shape of objects that can describe almost anything. 

Most importantly, we will begin to see knowledge graphs model business process. The world is moving to modeling, in general, whether they are explicit models done through a relational technology or whether they are statistical models based on these new neural network-based machine learning artificial intelligence things.

There are still models and, more and more, we'll move to modeling our world both explicitly and statistically. That will really help people to understand what their business is all about and make better business decisions and drive things forward; figure out what to do.

Michael Krigsman: Let's shift gears slightly. This discussion of models is a very important one, but another important aspect of all of this is the cloud. You have been involved with the cloud, helping shape what we think of as the modern cloud computing. 

“The future is cloud”

And so, where does cloud fit into again the possibility of AI and what we're currently experiencing?

Bob Muglia: The future is cloud. I'll just say that. I think the future is that more and more companies will adopt the cloud in one sense or another to run their business. Even companies that think of themselves as running things on-premises will use a variety of cloud services in a lot of ways. 

The cloud continues to expand. One of the most interesting places where the cloud is moving to is the edge; devices that get closer to you inside cities and things. I think that's going to become progressively more important as we move into a world of robotics.

One of the key things that I think is going to happen over the next 10 to 15 years is we will enter what I think of as the era of robotics where robots of various shapes will be flying around as drones, walking around in our houses, eventually. They'll be running down the sidewalk doing all sorts of things. 

These robots will be working and interacting with us in our lives, and I think the cloud will control all of that. It will start in these centralized, massive data centers, but it will extend out to edge data centers that are in every city around the world to help control these things effectively in real time. 

If you have two drones flying around, they better know where each other is because you don't want these things to crash into each other. Traffic control is going to be a whole new set of things. You really need a cloud-like system to do that. It's the only way you can possibly solve these problems.

While on-premises remains important, customers have security concerns. Some companies feel that's very crucial to them. More and more, these things are going to be cloud-based services. 

In general, data collection in the cloud and analysis in the cloud is so much more appropriate because it's very bursty in its nature. And so, instead of having to have dedicated computer resources to do things, you just have compute resources when you need them.

QuHarrison Terry: We have some members in the audience that have a ton of questions. I'm going to take one of the questions right now. It's from a frequent CXOTalk listener and viewer of the show, Arsalan Khan. He's got a really good question. 

Regulatory policy and AI

What he's asking is, if we all become data collectors and we start to just unearth and just say, "This data is valuable. This is that. This is this," who is going to be the enforcer and really make sure we need to collect this data, how do we store this data, what are the privacy laws pertaining to the data? What do you feel is the right outlook for that?

Bob Muglia: The issues of regulation is always a really interesting conversation. There are clearly steps of things that are very important. 

Different parts of the world have different views on privacy. Europe's view on privacy is different than the view in the United States, and certainly different than in some Asian countries. 

Those sorts of compliance requirements have to be taken into account in everything. I think that the tools are going to continue to improve to make this easier and easier. 

GDPR was actually, in my opinion, a big step forward. The alternatives were much worse before then. The alternative was confusion before then. GDPR at least makes things clear (or generally, reasonably clear) as to what you need to do.

QuHarrison Terry: Remember there's no such thing as unstructured data, so there's no such thing as confusion in your world. 

Michael Krigsman: [Laughter]

Bob Muglia: Well, there can be confusion. There are a lot of things that are confusing in this world. 

QuHarrison Terry: I was just going to say that. [Laughter]

Bob Muglia: There are a lot of confusing things in this world. 

QuHarrison Terry: Yeah.

Bob Muglia: Frankly, a lot of regulations are pretty confusing. You look at some of these regulations, and they're almost impossible to understand. 

But compliance will always be an important part of this and, in particular, access control. All of these things are very important. 

Semantic models in business

Frankly, there are some problems that still need to be addressed in the modern data stack. 

It's not hard to put access control on your data in the modern data stack. What's a little hard is to manage that and to understand exactly what access people have. That's still pretty hard. Those problems will get solved over the next few years, though. 

You actually need knowledge graphs. The thing that's interesting is that all these problems are appearing where it's becoming clear you need a semantic model for something. You need to describe what you're trying to accomplish. 

Semantic models for businesses, this is really important. Where are the business rules? Where do the business rules exist inside your organization? 

An interesting question for any CXO to ask: Where are those rules? Do they exist? 

Well, I'm going to tell you where they mostly exist. They exist inside applications, often very opaque. You don't know what the hell they're doing, but they're in there. These are applications you're running.

They exist in Slack messages. They exist on whiteboards. They exist in people's heads. They are almost never declared explicitly and never are they declared explicitly in a way where they can be operationalized. 

That's all going to change with knowledge graphs. That's what knowledge graphs are all about is to take what's today implicit about the business and make it explicit, and to define the semantics associated with the business.

Now here's the interesting learning, and I've heard this now from at least eight different sources that are people trying to use these large language models to do things. 

One of the areas of advancement that people are trying to solve is to use English as a BI language, as a business intelligence language. Instead of having to write a SQL query directly, you could just ask a model to find your data for you. Well, what people are finding is they try and build these systems is that, to solve that problem, they need to have some kind of semantic model sitting behind the large language model to explain how the data works. 

There are these very simple demos. You have orders, and you've got customers. You do a very simple demo, and you run a query. Boy, isn't it cool? You can get an answer. Just ask a question in English.

Try that with your real data and the complexity of the data environment that you have. It won't work the way you would expect. 

To make it work, we have to help the language models understand our business. How can they possibly...? If we don't understand our business, how can they understand our business? The only way you can do it is being explicit.

QuHarrison Terry: One thing I want to ask you here, and then I'm going to pass it back to Mike, is that you have very clear, explicit expectations of the technology and where we're going. However, one of the things that I see when I look at just the current landscape is the way we structure our teams if we get these technologies to work the way they should. 

This is past the hallucinations. This is the past the semantic models actually working. Once that works, why do I need the current team structures that I have? How do you think of reskilling and upskilling your organization? 

Bob Muglia: Well, I think that team structures will change. They always change as technology advances. Technology advances change the way teams work.

If you just look at data teams, in general, if you look at a modern data team today, it looks a lot different than a modern data team did ten years ago. A data team ten years ago, if they were really sophisticated, they were working with Tableau, and they had one data warehouse. 

QuHarrison Terry: Yeah.

Bob Muglia: But the other team over here was working with Excel, and this team over here was doing this other different thing. 

Today, they have these things all centralized. That required different ways of structuring and thinking about teams. 

One of the areas of conversation, there's been an organizational structure described called data mesh. People have heard of that. 

Data mesh is essentially a way of organizing teams and thinking of data as a product within that team and then providing that to other groups within the organization. Totally appropriate. It's exactly the way to think about things for large organizations that have multiple data teams.

My point is that as technology changes, the organizational structure that you put in place to support that has to evolve. It will be different in different companies. 

Having talked to many companies, some are very centralized with very centralized IT and everything goes through that. Some are super decentralized and they have to think differently.

I would never encourage anyone to make a giant change in their culture unless they feel there's a strong business reason to do it. But you need to have adjustments in the way you work with your teams within your existing culture (whether it's centralized, decentralized, whatever its element is) to allow you to run your business as the technology changes.

Large language models will have an impact. For sure, they will have an impact on that. 

Exactly how, I don't know. Honestly, I'm not sure.

It'll change. It'll be very dependent on how the technology evolves. 

I'm sure we have not seen the coolest applications of this. There will be new things coming that people will be wowed by that will change the way business works that we've not even seen yet, and so we have to see how the technology gets applied. But it will change. 

Importance of adopting a modern data stack

Michael Krigsman: Bob, what you're describing, of course, makes perfect sense. But in a way, it's a kind of science fiction for many organizations today. And so, going back to a question that Qu asked earlier, how can organizations adopt this? Then maybe that's a great lead into Isaac Azimov and literally science fiction.

Bob Muglia: Everybody should adopt the modern data stack, some incarnation of the modern data stack. You should be looking at your data sources, whether they're internal applications or whether they're third-party SaaS applications, or they're the apps you purchased and you run. 

Whatever you're doing (on-premises or cloud), take that data. Put in place one of the modern data stack providers like Snowflake. Adopt that as a centralized repository for your data. Buy tools that will allow you to move the data out of your different applications, your operational apps, into the centralized repository. And begin to work with it. Get the data together. Begin to allow people to analyze it. 

Begin to become data-driven. This is the first major element for a company is to adopt a data-driven mentality where they begin to use their data sources as the way to answer questions. This is the biggest cultural transformation that I think many companies have to go through is to begin to trust the data. 

What I have found is that valid data that people believe is the fastest way to close a discussion and to make a decision. In an organization, decision-making, you're trying to make the best decision you can in a timely manner is critical for every organization. By far the best way to do that is to do it data-driven. 

What I have found is that I think my intuition is pretty good and it is often very wrong. What I intuitively believe is often disproven by the data. And I have learned to follow the data and not follow my intuition.

I think, if you look at really successful companies, if you look at a company like Google or Amazon, these are data-driven companies. 

QuHarrison Terry: Build off the data, literally, right? I look at your resume, your track record. You've been a part of some of the most influential and impactful technologies to touch base in the technological landscape the last three, four decades.

The question that I have is, in the last decade, you did the unthinkable. You could have just rode off into the sunset and chilled, been a VC, and just literally been the iconic force that you are. But instead, you hopped into the saddle and became an entrepreneur. There are a lot of learnings there that I'm sure you've had. 

You weren't just any entrepreneur. You actually took your company public. And you did it in less than ten years, which is already pretty astronomical in itself. 

The question I have for you is, how do you think about the team and the structures within the org? Obviously, the thoughts that you have here (and if it's the data you're following), I want to know more about just how you led meetings, how you organized teams, things of that nature. What are the quick takeaways in that department?

Bob Muglia: I actually have a very, very short toolbox of management techniques. It's not enormous. I don't have 100 different things that I pull out of my toolbox. There are only a few.

My absolute favorite technique is the regular meeting that you have with an organization to drive an outcome. The cadence of that, if you were the leader of the organization, the cadence of that is often a week, that you do a weekly meeting with your team to drive an outcome.

Let me give you an example of one of the first times I think we did this incredibly successfully, and that was at Microsoft. It was not the first time, but it was a really exemplary time. 

When I was running the Windows Server team and VMware was exploding in the marketplace, this was back in the early 2000s. This was the time when virtualization was just taking over in the IT industry. It was a huge threat to us in some senses, VMware was, and we had our own product, HyperV, which we were competing with them on. 

I ran a process because this was a massive change to the Windows Server business. We were literally moving from physical licensing to virtual licensing. It's a big change. 

It was a huge business. It was a $5 billion business even back then, so you don't take these things lightly. Mistakes can be very, very costly.

We ran a process over many weeks where I had about a dozen people across the company meeting. Strategists, financial people, product people, marketing people, sales people, people from every part of the organization thinking about the problem.

We talked about how we would restructure things. Over time, we came to a new licensing model that ultimately turned out to be a real win-win.

I look back on that. While we were very competitive with VMware back then, you look back, and what happened? Both companies won. 

What a great outcome. VMware won and Microsoft won. Actually, all three won because the customer won, too. It was a win-win-win, and that's always what you look for, especially when you think about partners.

I always think about partnerships as tactical. Every partnership is tactical. There's no such thing as a strategic partnership. There are just long-term tactical partnerships.

To me, you're always seeking that win-win with your partner. If you can't find that, the that partnership is going to dissolve. I look back, and the process of doing that is always a process I find where it requires the thoughts of multiple people working together. 

I am so much smarter when my brain is combined with the thoughts and ideas of so many other people. And those all come together to create a stew of the best possible ideas.

Michael Krigsman: Mike Prest, who is CIO of a private equity investment group, asks on LinkedIn. He says, "Customer-centric transparent data collection policies can help industries self-regulate. Your thoughts on the issue of companies having opaque data collection policies and the fact that consumers are less likely to trust and use these companies?" The issue of data and trust.

Bob Muglia: Well, it's a huge issue. Right? Companies will have their own policies associated with it, and different businesses require different things. 

Companies that are advertising-based, for example, as their primary revenue model. Their primary customers are advertisers, not their consumers. 

What I think of is I believe in transparency on these things as much as possible. When we created Snowflake, we did this in a very transparent way, the way we worked with data, because that's what customers were doing; they were trusting us with their data. 

I am a big fan of transparency. I know how important it is. Yet I recognize there are businesses where there probably won't be transparency. Mostly those are businesses where the interest of the business is not aligned with myself as a consumer. 

Michael Krigsman: Arsalan Khan comes back, and he says, "An image is essentially data. If we create an image through text-to-image, basically the bottom line is who owns the work product: the AI or the artist? How do we manage that mess?"

Bob Muglia: These are derivative works, and so you're creating new things. 

I think what's going to happen is creators (authors, publishers, people who create something, artists, whatever), they're going to decide what they want AI to do with their work, whether they want AI to be able to work on it. If they don't want to, I think they'll be, "No. Don't tread on me," and the AI will avoid those people.

My guess is going to be that that is a temporary thing and that everybody is going to want the AI to understand your work product. 

If OpenAI, Google, or anyone wants to read The Datapreneurs, please read The Datapreneurs with your next large language model and make sure it knows. I want it to be in there, and that's a decision that I, as an author, makes. I think it will be a decision made by the creator, basically. 

Why Isaac Asimov was a “prophet”

Michael Krigsman: You refer to Isaac Asimov as a prophet. Why?

Bob Muglia: Over 60 years ago, he foresaw all of this in amazing ways, really; in truly amazing ways. He wrote his first robot novels and defined what's known as the Asimov's three Laws of Robotics in 1942. Okay? That was before the digital computer was invented. Just think about that. 

Asimov was thinking about a world where people would live with intelligent machines. And the big thing he did is he didn't treat these robots as Frankensteinian, creations that shouldn't have been made. He thought about them as machines that were created by humans to serve humans. 

What are we doing now? We're doing exactly that. 

Remarkably, if you look at his books, a lot of what he was writing about was happening when? The 2030s and 2040s. Guess when it's actually going to happen. His timing is almost exactly right on.

I believe, by 2040, we will have humanoid robots that will live amongst us and perform tasks working for us. Helping to take care of elderly people. Helping us in our household. Helping us in a whole bunch of ways.

This is coming, and Asimov foresaw it. And in order for it to work, we need, effectively, the laws of robotics.

To give you an idea about how far ahead he was, later in his career, he augmented the laws of robotics. The original laws were must not harm a human or allow a human to come to harm. It must accept orders, except when those disobey the first law. Then it can preserve its existence. 

Later on, he created the zeroth law, which is that a robot may not harm humanity or, through inaction, allow humanity to come to harm. Now think about how prescient that is. Thinking about that so far ahead of time and look at where we are and what we're doing. 

He was thinking about that then. That's a prophet for you.

QuHarrison Terry: What is your least favorite word in the realm of technology?

Bob Muglia: I have learned, over time, that technology is going to make almost everything possible – over time. It's just a question of getting the timing right and realizing what's possible.

Impossible. I would say the word I like the least is "impossible" because things that we thought were impossible now are possible. I think that that's going to continue to be true.

Michael Krigsman: Apparently, you would not be a great default to know CIO. You don't believe in that. 

Bob Muglia: I believe in yes. I believe in making things happen. I believe in solving problems. I'm a big fan of that.

Michael Krigsman: This is from Lisbeth Shaw. "How can we avoid an AI-mediated dystopian future?"

Bob Muglia: Our values. By making sure that we think about the values we create.

The machines will reflect our values. Because people will create these machines, we will have machines that reflect every value, everyone, that people have. 

That means the good, the bad, and the really awful. We're going to see some really awful things too, but we have to manage that just like we manage everything else. 

I've lived my entire life under the nuclear umbrella. I grew up in the era where I ducked and covered in the 1960s when it was very real. 

Here we are. We've been able to survive as humanity. We can survive AI too.

Skynet. There's no... Skynet only happens if we want it, if we make it happen.

Michael Krigsman: Again from Arsalan Khan. He's really on a roll. He says, "In an organization, which is more important: the data or the people?"

Bob Muglia: Always the people. The people are everything always. The data is what comes from people. People create everything.

QuHarrison Terry: Are you going to start another company?

Bob Muglia: I help people start companies. I help people build companies. That's what I'm doing now. 

QuHarrison Terry: But would you do it again? I feel like you've still got something in you (from this conversation).

Bob Muglia: For a bunch of personal reasons, it's probably not the right thing for me to do. One of my challenges is that I don't know how to do this anything less than 100 hours a week. I wish I could be good at that. I'm not. I'm not.

I have to recognize my limitations on that. And so, now, by working through others, I can keep my time reasonable and help a lot of other people be successful, which is great. That's what I love to do: help people be successful.

Michael Krigsman: Okay. With that, I'm afraid this very fast conversation is out of time. A huge thank you to Bob Muglia. He is an industry legend and the author of The Datapreneurs. And to my excellent co-host, QuHarrison Terry. Qu, it was a pretty fast and furious conversation. 

QuHarrison Terry: It was. Bob put it in fifth gear pretty fast, and we stayed there the whole time.

Michael Krigsman: Everybody, thank you for watching. Now, before you go, please subscribe to our newsletter. Hit the subscribe button on our website. Subscribe to our YouTube channel. Check out CXOTalk.com. We have incredible shows coming up and special surprises. Check it out. 

We'll see you soon. Have a good one, everybody. Take care. Bye-bye.