Explore enterprise AI in business with tech leader Bob Muglia in CXOTalk episode 819, covering AI strategy, ethics, and business transformation.
Welcome to CXOTalk episode 819 for a practical discussion on the role of AI in enterprise environments with Bob Muglia, known for his significant contributions at Microsoft and his role as CEO of Snowflake. Bob shares his insights on AI's current state in the enterprise, emphasizing its early adoption phases and developing trends.
Throughout the conversation, Bob explores the various applications of AI, from its utility in optimizing meetings through tools like Microsoft Teams to its influence on creative fields and business strategies. He considers the implications of AI on business models, the changing nature of work, and the ethical considerations surrounding AI development and deployment.
Additionally, the episode touches on broader topics such as the future influence of AI across different industries, the development of knowledge graphs, and the need for responsible AI practices.
Join us as Bob Muglia offers a grounded perspective on these subjects, supplying valuable insights for professionals navigating the evolving landscape of enterprise AI. Watch this episode of CXOTalk for an informative session with Bob Muglia.
Bob Muglia is a data technology investor and business executive, former CEO of Snowflake and past President of Microsoft's Server and Tools Division. As a leader, Bob focuses on how innovation and ethical values can merge to shape the Data Economy's future in the era of AI. He serves as a board director for emerging companies which seek to maximize the power of data to help solve some of the world's most challenging problems.
Michael Krigsman is an industry analyst and publisher of CXOTalk. For three decades, he has advised enterprise technology companies on market messaging and positioning strategy. He has written over 1,000 blogs on leadership and digital transformation and created almost 1,000 video interviews with the world’s top business leaders on these topics. His work has been referenced in the media over 1,000 times and in over 50 books. He has presented and moderated panels at numerous industry events around the world.
Table of Contents
- About Bob Muglia
- Perspectives on AI in the enterprise
- Future enterprise business disruptions from AI
- Business model disruptions caused by enterprise AI
- Regulation and the AI social contract
- Ethical and responsible AI
- Datapreneurs and the “arc of data innovation”
- Business model disruption depends on the industry
- Will generative AI and RAG drive API interoperability based on search?
- Copyright and intellectual property and AI
- AI, knowledge graphs and semantic models
- AI and fairness
- About Bob Muglia’s role at Microsoft
- How Satya Nadella changed Microsoft’s way of thinking
- Advice for business leaders on enterprise AI
Michael Krigsman: Welcome to episode 819 of CXOTalk. We're exploring enterprise AI survival with Bob Muglia who is, I think I can say, genuinely an industry luminary.
Bob Muglia: Great to be here. Is it really 819? Is that many really?
Michael Krigsman: 819.
Bob Muglia: Pretty impressive. Pretty impressive.
Michael Krigsman: Bob, very briefly, tell us about your background.
Bob Muglia: I joined Microsoft in early of 1988. I started as the first technical person on SQL Server.
Back then, I all my career, my career at Microsoft was all in the product areas. I managed to work on pretty much every business at Microsoft except games. I never did that, but a lot of time on servers. I ran the office team for a while, some time in MSN and then the services side with Azure and running a server that's Server and Tools division.
At the end of my tenure there, I grew that division from about 9 billion in revenue to about 17 billion. The equivalent division is quite a bit larger today, which I think is a fairly impressive, impressive thing of what Microsoft has done subsequently. But but at the time it was it was a it was a strong growing division at Microsoft.
I went down to the Bay Area. I spent a couple of years at Juniper and then and then spent five years as as the first real CEO for Snowflake and grew that business to about 200 million before I left in 2019.
Michael Krigsman: you've had a pretty extraordinary career and view of the evolution of enterprise technology. And so as we start 2024, when you look over A.I. in the enterprise, what do you see?
Bob Muglia: I see a lot of people looking, a few people playing.
And it's still very early. I mean, it's still early days, especially for the enterprise. Like many things, A.I. has really become first and foremost a consumer phenomenon. If you look at at the initial things with things like dollar stable diffusion and and course chat CBT, you know, these are really consumer plays and we've seen a lot of very obviously very interesting things.
The enterprise, a lot of additional concerns largely having to do with governance and data security and the products are much more nascent. I believe that 2024 will be the year of the first generation of A.I. products for the enterprise. The first six months of this year. It's going to be copilot season, I think with just about everything.
Getting a copilot. I haven't found a product which we're I haven't talked to. we're building a copilot. Now, how good those copilots will be, how useful they are. We'll find out over time. You know, Microsoft most importantly has is putting it across almost their entire product line. They just announced yesterday that the Windows is going to I think the start button is going to switch to it to a copilot A.I. button, which sounds interesting.
We'll see what happens with that. So we'll see a lot of interesting things coming out. There are some examples, I think in this first generation of technology that's really interesting. Obviously, the creative side of of what is possible with generative AI is generating a lot of excitement. We'll see Tools like Firefly from Adobe are being incorporated into their suite and really adding a lot of value, I think, to creators in that space.
So, I think it's a a time where creators will benefit tremendously and then, you know, there's some other interesting things. The other thing that the AI is capable of doing right now is summarizing things. And that can be really useful. Maybe if you were to ask me that the product that will have the most interesting and you never know about this, but my guess is is is Microsoft's teams product where they are actually taking in transcribing meetings and then some providing a summarization of that which they can then provide to people who didn't attend the meeting.
That seems almost magical to me and the experience I've had talking to people who are using the product is really indicative that it is. It is somewhat of a life changing experience for managers and people who work with a variety of different teams. Because for the first time ever, you can not attend a meeting and really quickly get a summary of that from out of the system, which is not biased by any individual's perspective.
And it really I mean, I was talking to a manager that was using it and his perspective was it was really changing the way he ran his organization because he could he could keep an eye on things and understand where to drill down without having to worry so much about is the offending people. What questions or is he asking, etc. He could really get an unbiased and unbiased view of things and then really learn what happens at a detail level afterwards.
Michael Krigsman: One of the issues that's happening with that is the question of the etiquette. So if you're invited to a meeting, you can send your bot in instead of you.
Bob Muglia: We are not to the point where we have agents that can set that, they can represent us with the bots yet. But what we do have is tools that can actually see what happened and write a short summary so that, you know, you know, it used to be before Zoom that if you didn't attend a meeting, you missed the meeting and maybe somebody took notes with Zoom, you know, you had the ability and teams, you've had the ability to record meetings and that's been great.
And I've found that very useful sometimes as a board member, just having a ability to have visual, have an understanding of what's happening and things. But if you're, you know, if you're listening to a meeting, it's an hour or whatever of your time. Whereas if you can look at a summary that's maybe four or 5 minutes and then, you know, and then the fact that all of the data is behind that, the full transcript as well as the video, if you find something that's concerning or interesting, you can drill down on it.
And this is this is this is a tremendous way of changing the way people are working. I mean, I think these are very significant sorts of things that will become a reality in 2024.
Michael Krigsman: So clearly there are going to be a variety of changes in how people work in meetings. And you you mentioned Adobe. So art creation and marketing, for example. What about business models and business strategies? What do you see? What have you seen and what do you see going forward regarding the impact of AI and generative AI on how we run our companies and our revenue models and so forth
Bob Muglia: AI has changed the goal in some sense of what I always sort of viewed the data world being, I mean, the goal of the data economy was always to be able to provide information to people, to allow them to make better business decisions, to make them faster and make them more accurately based on on what they're learning and what's happening in the real world.
And data is critical to that. And that was always possible. What's new and it's really exciting is that is that now you have the ability to have intelligence being incorporated into your business systems. And so the interesting question is where do you put that intelligence and how do you make lever and how do you leverage it, and how can you as a company, take advantage of the fact that now there are a set of tools that can help you in a set of decision making that goes way, way beyond just looking at charts and data.
You can now see information that summarized, you can have recommendations made. A lot of things like that are going to come out in the next year. So intelligence is kind of the new goal. And the question is how do people incorporate that intelligence into their business? You know, obviously very much depends on the business. If you are a creative organization, the tools are tremendous today.
I mean, the tools you're probably using in 2024 are probably going to be pretty different than the tools you were using in 2022, because you will almost certainly be incorporating some of this generative AI in your creative process, whether you're making marketing briefs or whether you're doing you're doing a creative image creation for for marketing, promotions, whatever. And shortly video, of course, is now starting to become possible.
So, a lot of things can be simplified today that weren't there before were possible before. What we're still looking for and what's still very, very early in all of this is how that intelligence in a deeper way can be incorporated into the business process and leveraging that so that that people don't make mistakes and that the better the right decisions are made for customers and customers get a better experience.
I think we'll start to see this year the the next wave of the first wave of AI was the creative tools and the copilots that are appearing. I think the next wave that will begin to see is improved query systems and answer systems to be able to understand what's happening in the business and know who you're going to talk to.
And that is being is being driven by something called rag retrieval augmented generation, which is working. I mean, it's one of the things that is really working today with a I and the techniques were how you, you you take a corpus of information and leverage that together with these new models that have that have that can go through that and provide answers to people that that I think is going to become very mainstream, although probably later this year as it as we start to see products emerge with it, that the fact of the matter is, is that the models, it makes a ton of sense.
The models can't come up with good answers if they don't have the context for things. But when you can provide them with context, documents and things that can be leveraged for them to understand what's, what's true and the unique things about your business, then they can come up with better answers. And and I expect we'll begin to see things like much improved enterprise search capabilities that can search across all of the different different systems people have.
The Slack's in the document management systems and all of the different, but the source code control systems and JIRA is and all of those things to be able to provide better answers for people inside the company. And I'm very hopeful that one of the first generations of products will begin to see are really improved product support products for customers where today what may be is handled by by first level product support and maybe in some cases second level product support will get automated with these tools.
And I think we'll begin to see that probably by the end of this year.
Michael Krigsman: One of the things that you've been describing has been essentially incremental improvements in efficiency and productivity. But there's a question from Arsalan Khan, who is a regular listener of CXOTalk, and he always asks great questions and he says, should organizations consider AI as something that can improve a few things, like I was just describing the efficiency we run a better meeting or can it disrupt the whole business model?
And at the same time, do executives really have the right incentives to disrupt their business model? If it results in them or their department being less significant? So I think there's two parts, right? There's the disruption question and then there's the human question of change.
Bob Muglia: I think it's a timing question more than anything too to determine disruption. I believe this year and probably into next year, most of the changes will be productivity improvement, ability for people to communicate more effectively, hopefully more cohesion across an organization because you have better answers and you have ability for people to work together more effectively.
So I think that's the first stage with the technology as it really exists in 2024, which is still still extremely nascent. If we go out two or three more years and we begin to see agents becoming successful, then the thing that's not working today, I said retreat. I said Retrieval. Augmented generation is working and we're seeing solutions appear based on that.
What is not yet working are sophisticated agents that do things on your behalf, like you mentioned earlier, about having sending sending your your your, your your your agent to the meeting for you and your and having it intend in your place. We're definitely not there yet. We're very clearly not there yet. Now there's a question of will we get there.
I think the answer to that is absolutely yes. And the question is how soon will we get there? And nobody really knows for sure. But my guess would be it's it's 2 to 3 years away when these agents start to become successful. The challenge right now is that because the technology associated with helping these models to make right decisions is still nascent, there's enough error in what they're doing that that if you go through multiple steps in a process, the likelihood that they'll make a mistake along the way that is just more troublesome than it's worth is is greater than is greater than, you know, the benefit that it gets.
So, I think when we start to see agents, like, for example, if we have an agent that can manage calendars effectively so that, you know, you can have you can schedule a meeting with multiple executives and the agents will coordinate across that instead of executive assistants. That's going to change the role of an executive assistant completely and and take a tremendous amount of what they're doing today, which is calendaring and take that off their plate.
Now, there'll still be a role for those people, but perhaps it will be more limited and perhaps it will be less jobs available. So yes, it will have disruptive impacts and that disruption will have involved, in some cases, you know, disruption to people. And that's one of the difficult aspects always with technology shift and AI is no exception to that.
Now, the final question that was asked was will managers have the guts to do it? And that's going to depend completely on the organization. You know, I know of I had a conversation with a CIO last year who I'm very close to, and, you know, this person told me that they left the company. They were working for.
You recognize the name of the company? If I told you. And they left because they felt that with the data work that they were doing and the new things that were coming, they could completely change the organization and lay off and reduce the staff by as much as 50% or more. And the organization did not want to do that.
It was not ready for that. The management team and the the CEO and the board was not ready for that change. So this person decided to go look, look elsewhere. But those changes will be real and they will come. Some companies will embrace it, some will reject it. Chances are those that embrace it, at least those that do it smartly, will be the ones that will succeed in the long run, because that's usually the way it works.
Michael Krigsman: you're saying that at this moment in time there are still technology limitations that prevent us from the kind of see changes or business model changes that Arsalan Khan was asking about.
Bob Muglia: Well, just look what happened with Drive in California recently. One of the the auto autonomous vehicles that General Motors has said they because their technology wasn't quite ready, were forced to slow down and not and not continue with their testing of autonomous vehicles. Google, on the other hand, is actually continuing to see some success with that with Waymo.
So so we'll see. I mean, when we when we have self-driving cars that can that can drive us around cities, let's just talk about that. You know, let's forget about going in the country and everywhere else, but just just around cities. That's a massive change in terms of of what's possible. And it will have gigantically disruptive impacts on the economy and people,
Michael Krigsman: subscribe to the CXOTalk newsletter and subscribe to our YouTube channel. Check out cxotalk.com.
Michael Krigsman: Abstract Infused on Twitter mentions your book The Datapreneurs. So I'm holding up a copy right here and he says that you discuss the need for a new social contract that helps guide A.I. Can you please elaborate on the guidelines or the and or the guide rails? You think this new social contract should include?
Bob Muglia: That book was written in you know, it was finished in the first quarter of last year. So there's a lot of water that has gone under the bridge since that that book was written. And what we've seen, I was in the early days, I was worried about the fact that how serious only would the regulatory agencies take a I and how how much interest would be taken to it.
I've not been disappointed by what's happened. And in fact, in some senses I'm worried that it's going a little too fast, if nothing else. If nothing else, clearly we are seeing a great deal of interest in the social impact of AI on society, the regulatory requirements that are associated with that. And we've seen some initial steps taken both by President Biden as well as by in particular the European Commission, which, as is characteristically their case, has has taken a leadership position, you could say maybe a heavy hand in some ways, but in a lot of ways, a leadership position in thinking about the regulatory aspects of AI in society, that there's really two distinct things
to worry about here. One is in the short to medium term, and I'll say between now and like, say 2028, that kind of time frame, the real concern is that AI is a tool that can be used for all kinds of purposes by people, and we need to make sure that that society accepts and understands the implications of that tool and and performs a proper and does appropriate regulation of it where that's necessary.
The most obvious one of those there are a number of places where you can discuss this. There's a lot of interesting issues about copyright law and what what makes sense for these models to be created that will be litigated through the courts. What what's in the short run, some things is is the ability for people to now for the first time at effectively very low cost and, you know, very low skill levels now basically create an image or even a video that depicts just about anything, including somebody saying something, somebody who's not known and perhaps famous, saying something that that they don't really believe.
These deepfakes are definitely a problem. And I think that we do need to have regulation associated with making sure that when people are representing something, when people are are creating a video that is not a real that is not real, that it be attributed appropriately. I mean, to me, that's the only thing. It just I don't mind people doing it.
They just need to achieve as long as it's not totally inappropriate. But but like porn or something like that. But attribute it at least, right? I mean, we'll see a bunch of this. Actually, we'll see a ton of this in the political season coming up because I'm sure we're going to have all kinds of dark videos being created by AI on probably both of our presidential candidates that are coming out.
And hopefully that will at least be represent it properly. And I think that that in fact, society will push to that. The second issue, which is probably the one I was really referring to in the book, is further out when we begin to get artificial general intelligence and in particular, if it proceeds beyond that to superintelligence, where we begin the machines that are effectively smarter than we are and we have to change our relationship with them.
But the one thing that has changed I said this the beginning of the podcast is, is that I always thought it was about the data economy. It's now about the intelligence economy. And that's the thing that I think is so different is now you can bottle intelligence in a way and put it inside these models and we even can see that in some nascent forms today.
Now, we're a long way from superintelligence. What we have to worry about, about how we align ourselves, align those those entities with our objectives. But in the short run, we still need to make sure that the models that we are putting out are aligned with the values that a company has. And that's a very, very important part of what enterprises care about as they roll out.
I mean, they don't want to be putting something in front of their customers that doesn't represent the value set that they have. So that's why these contracts are very important. It's a social contract that every at every possible level, ultimately it's a contract between humanity and the next set of intelligent entities that we are creating because that's happening and it's going to happen.
I believe it. I think most most technologists believe that these days, in the short run, it's about how people leverage the technology and do so in a way that's consistent with the values either way, the core is values, though. And what are the values that an organization has and a country an organization, a company, a person? What are those values and how are they represented?
Michael Krigsman: And of course, this issue of ethical A.I. Responsible A.I. is going to be one of the most important ones, one of the most important issues in 2024 and certainly beyond. The White House just issued their guidelines, which really are have very far reaching ramifications and implications. But we have another question, this time on LinkedIn. And let me just mention, we will get to all the questions.
So I'm going back and forth to give everybody a chance. So, if we haven't answered your question yet, don't worry, I'm tracking it. I know about it. We're going to get to it. So on LinkedIn now, Greg Walter says the following. He says, I'm an advocate of no guidelines or regulations yet because we are in the Explorer Ocean Stage of A.I. And he feels that any guardrails stifle innovation. Thoughts on this?
Bob Muglia: I'm actually more in his camp today than I am not. In some senses I do believe that we need to allow creativity and openness. I'm a huge fan of these open source models. I think the most the most significant thing that happened in 2023 was the creation of open source models that that have capabilities that don't quite achieve what you can achieve with the frontier models like GPT four.
But we now see models like Mixed Crawl that are really at GPT three five level and that's pretty incredible and I think that's a very, very good thing. So I'm actually morally joined the not on this, I'm a light, I'm a light touch rather than a heavy touch. There are some places where I think it's obvious that you need regulations.
I mean, you shouldn't I mean, we shouldn't be creating videos of people that look perfect of them saying stuff that they don't believe and that needs to be attributed in that there are some things like that where I do believe we already can see the need of right for regulations. And again, this is in the place distinguishing between it's not about alignment of superintelligence.
That's not a 2024 problem. I mean, I'm glad OpenAI is working on it. I'm glad they're working on it. Anthropic are working on it. That's good. But for the rest of us, it's not a problem for us to worry about. What we are worried about is the alignment of these models to the values that we are applying them to right now.
And mostly that's a values thing that should be controlled within organizations. There are few places where government regulation makes sense. And then once again, I think that if you're in Europe versus in the United States, you'll see a difference. The Europeans have a different view on regulation than the United States does, and it's just part of the cultures of the different societies.
Michael Krigsman: We have another question from Abstract Infused, that is on this General, say, soft kind of topic. And then we have a question will follow up after that from Chris Peterson, who has a more technical question. And so Abstract Infused says that going back to your book, the data preserves not sure who Abstract Infused is, but he's clearly a fan of your book. He describes a drawing that you have, which I'm going to hold up. Nobody will ever you won't be able to read it very much, but you can see it a little bit and it is the arc of data innovation diagram.
And he says, given it's clear that we can anticipate enormous advancements in AI. What advice would you offer people considering a career in A.I. or data related technologies?
Bob Muglia: The biggest advice I have is, is to get engaged in it and begin working with the technology as deeply and technically as you're able. I mean, if you have data science capabilities, by all means, it's a wonderful thing to be able to if you have those mathematical capabilities and can work on some of the deep attributes of fine tuning of models and working in the open source community.
It's a it's party city for people. If you've got if you've got those technical chops and can spend time on that, in spending time working with the open source models in particular, it's very interesting. But across the board, the ability to as even as a consumer or a prosumer kind of kind of person that has some set of skills, you're now being able to apply these technologies.
So, if you like, you know, if you're if you're if you're creative and you can leverage some of the creative tools that are coming out and that that allow you to have to to create new visuals at incredible speed, at much greater detail, play with that. I mean, spend as much time as you possibly can. Nobody knows for sure which of these technologies are going to succeed and in exactly what ways.
We're still early on that. But by staying engaged, it's the biggest thing I've found, frankly, that one of the biggest things is is is following some of the more interesting writers in the space. Matthew Linley in particular is a writer who has done some really great has created some really great articles talking about the advent of the of technology around AI.
And certainly his is his frequent blogs are very, very helpful. So I would say just stay as engaged as you can and apply it in any way you can. And gosh, if you have the data science skills, jump into them into the open source models because it's a it's a it is it is like it's sort of the analogy I have is this is the 1950s in aviation, and that's when that's when almost all of the aviation breakthroughs happen within the 1950s.
And we are planes don't fly any faster today than they did then. They're more efficient, but they don't fly any faster than they did. In fact, they're slower. And and I know, you know, my my parents generation, I was fortunate if one of my relatives was a copilot, I was I was a test pilot. And so I learned a lot about what happened in that period.
This is that golden era for A.I., and it's just happening now.
Michael Krigsman: Let's go back to some more comments and questions, and I'll come to Chris Peterson's technical question in a moment, because Daine Alexander on Twitter just wants to make it clear and they’re correct saying that I can improve existing processes and disrupt the whole business model. It's not an either or. And of course, that's absolutely.
Bob Muglia: Agree. Again, I think there are probably places will begin to see disruption and you know, where jobs change completely in in this year. I think seeing that for creatives and we've seen that in some ways some of the biggest changes have happened with with with programmers and people who are writing code using copilots.
I mean, the first and most successful tools to date in enterprise is almost certainly the developer copilot. And so we've seeing that already. And that's a tremendous I mean, hey, to be able to increase productivity of a developer by 20, 30, 40%, that's incredible. That's an incredible improvement.
Michael Krigsman: on Twitter, SARP username S.A.R.P. says that the importance of disrupting business models actually depends on the industry. He or she says. That points out that yes, in the tech industry, disrupting the business model is the best way to stay relevant. But in the health care industry and improving a few things is sufficient.
Bob Muglia: Health care is one of those things. There are some industries that will move very quickly, and it's somewhat analogous to what we saw in Snowflake, by the way, Snowflake was a disruptive technology and when we came out with it, the first places we saw the disruption and the utilization were in adverts rising online gaming technology.
These were all or companies that have a tremendous amount of data that they cared about. Data was very important to them, but governance and security was not the highest order bit for many of these companies switched to something like health care and patient care. Private C All of those things are paramount as well as it's a very, very regulated industry certainly, and anything that is highly regulated moves slower generally for good reasons.
I mean, aviation is highly regulated and that's also a good thing because we want our airplanes to do what they're supposed to do and those industries will move slower. So I totally agree with that. I completely agree with that. We're still not sure what exact things are going to be completely disrupted. We know creatives are are absolutely going to have a heyday with this and they can do things that they never could do before.
But the other you know, we know eventually we'll have our self-driving cars. We don't know whether that's 28, 30, 31. We don't know when we'll start to see it more. Probably it's between now and then.
Michael Krigsman: So at some point there are going to be very profound business model changes across many, many different industries.
Bob Muglia: Earlier you mentioned the arc of data innovation. That is that I when I started writing the data for nerds, there was always an arc of data innovation.
It was the in a sense it was the core premise of the book, which was that technology keeps moving faster and faster and the data industry is a driving factor in that. What's changed is that when I started writing the book in 2021, you know, the pinnacle of it was the data economy and where the world was going to be able to move and work with data.
And that's where I'd always seen it. Now we're moving to a world where intelligence and the creation of artificial intelligence that can augment and in some cases replace human intelligence that is is new. And so now the arc goes, you know, all the way through artificial general intelligence, through a period where I think we're going to see robotics become very, very prevalent in society.
I have said for a while that the 2030s will be the era of robotics. I believe that more strongly than ever right now will begin to see very special purpose robots doing things for us. And by the end of that decade, I'm pretty sure we're going to have humanoid robots that can do many things for us, including elder care and and housework and things like that and medical care.
Can you imagine, you know, how revolutionary that can be if we could put a robot in every patient's room to help care for that patient? That's not that's not crazy. That's absolutely not crazy. And I think it'll happen by 2040. So, you know, the disruptions will be will be massive. Like many things in technology, it takes longer than we think it's going to take to get there.
But the changes are more significant than we ever could imagine.
Michael Krigsman: And we've had people, guests on this show who are using AI generative AI to design new molecules for drugs and actually having success with that. We have the head of AI for Novartis, one of the largest pharma companies out there who's been a guest and coming back sometime soon on CXO talk.
And when a company of that size, it's obviously not a startup, starts adopting these tools in a very meaningful way. Obviously, they see tremendous progress in drug design, drug development.
Bob Muglia: And again, but Novartis is a technology company, right? That's that that's very much, you know, building new technologies in the forms of new drugs. And so it's not surprising that that this this that this new technology in the form of a I could revolutionize some of that and cut the cost of developing many of these drugs and make more available.
It also the potential and I think this is going to be a big thing that the medical industry will begin to to utilize more thoroughly, which is which is designer drugs that are customized to a given individual's problems. And I think in things like cancer, we're discovering that that our our perspective on cancer is very simplified. And if you actually operate at the genetic level, which we now can begin to do, you can you can have much more effective treatments for cancer than these chemotherapy drugs, which are basically like a giant, you know, giant sledgehammer trying to to solve a problem.
Michael Krigsman: Let's move on to a technical question from Chris Peterson, who's been waiting patiently. And we have other questions that are also stacking up. And don't worry, we'll get to all your questions.
So, Chris Peterson says as if AI will he's responding to a comment, Bob, that you made earlier about search. And he says if AI will power a new generation of enterprise search with RAG, will that provide a forcing function for APIs and interoperability among tools?
Bob Muglia: One of things that's happened is with the Web and in the previous generation, the SaaS world, is that data is being exposed across applications through some form of API and almost every application, certainly almost every commercial application you could buy has some set of APIs to allow you to extract data from that and make use of it.
One of the companies I'm involved with, I'm a board member five Trent, that's our business is to pull data out of data repositories like Salesforce as well as databases like like DB2 and Oracle and and PostgreSQL, etc. be able to take that data and then give it to the modern data stack, push it up to the modern data stack for further utilization.
So, a lot of those tools are there. The thing that I think is so interesting that's going to happen because of AI is the API I of, you know, the language of of the computer language of 2023 was English, which is crazy as a as an old time computer scientist, the idea that we can talk about English as a programing language, but that is that is a fact.
So, I think that's going to change is that while today you can get the data out of slack, that's not a hard problem. Getting the data out of Slack is not that hard. But what will be interesting is over time, I think we're going to begin to see agents in place inside these individual tools and we're going to have AIs talking to AIs as the agents begin to talk to each other.
And the way they are most likely to do that is with is with a language like English as a primary interface between them. English has become a way of defining these APIs for programs. And that's a big change, but that probably will come a little later when these agents become more become more successful. And again, I think it just needs a little bit more maturity before that happens.
Michael Krigsman: we have another question from LinkedIn. And this is again from Greg Walters, who asks another thought-provoking question. He said, he says, Do you see a time when copyrighting and IP and other legal constraints are no longer? I'm just going to say that ain't going to happen any time soon.
But he then goes on and he says, What is the difference between him consuming content and creating content based on what he gleans versus. And here's an interesting point versus telling an AI to consume vastly more content for him no matter who originated. So really, what's the difference?
Bob Muglia: Honestly, this is one that ultimately will be decided by the lawyers in the courts. It's going to it's going to go through copyright law. I'll give a I'll give it. I want to tell a little story, though, based on this question of a conversation I had with Brad Smith. It must have been 2005, 2006, sort of in that time frame.
And Brad, who's currently president at Microsoft, he he ran the legal team at Microsoft and really probably more than anyone else was responsible for helping us to get out of the mess, the DOJ mess that we got ourselves into in the early 2000. And I spent a lot of time with Brad. You know, I was one of the ones who testified.
And then I further went and spent a lot of time dealing with the DOJ and with the court system on Microsoft's compliance issues. Subsequent to that. And I was always a big believer. I've never been a big believer in software patents. I, I sort of view software patents as not necessarily something that's particularly helpful to the world. And I think time has proven that out.
I mean, there was a huge fight between Apple and Google and Apple and Samsung, and in essence, it was a whole bunch of nothing when it all got done. And the only people who benefited from the whole blasted affair with the lawyers. So but I remember talking to Brad Smith and saying, you know, Brad, I get the patents matter in some places.
I understand why the drug industry needs patents because it's incredibly hard to create a drug and to go through the the process of validating it, but it's incredibly trivial to recreate it. Whereas in software, that's just not true. It's it's hard to create it, but it's also hard to recreate it. And it doesn't tend to have the same impact.
And you know, what Brad said to me is, you know, Bob, you may be right, but it doesn't matter because it's not going to
So, the world is filled with copyright laws, trademark laws, patent laws. Those laws will get applied in the new era. How they get applied, I don't know.
My guess is going to be that there will be there will be copyright licensing. There'll be they'll be there'll be licensing done. And we're going to see a lot of licensing happening in in this next year. And maybe that's good because it may be part of what's necessary to help the creators of content, you know, the publications, the journalists to continue in their business.
Because obviously these technologies don't have not the Internet has not made it easy to be a journalist. It has not helped that industry in terms of profitability for it. And and the vibrancy I mean, the newspaper industry is a whole lot less vibrant in 2023 than it was in 20 in 1990. And it was the Internet that changed that.
And I will make that even more distant. So there probably is a need for some sort of licensing associated with this, but only time will tell,
Michael Krigsman: really interesting how the relationships, the business relationship will evolve going forward. Earlier, you answered a question regarding APIs and interoperability, which reflects the technical technology relationships, but the business relationships are going to evolve as well.
Bob Muglia: Just like they did with the Internet, really. Right. I mean, look at all the lawsuits between Google and the various in the various publications. And generally speaking, what happens is some settlement is reached.
Google has plenty of money to pass a little bit along to us.
Michael Krigsman: Stephen Xi has a question. And then I'm going to go to Arsalan Khan, who's got a hard question. Stephen's question is an excellent question. Arsalan Khan’s is a hard one. So Stephen Xi says, “In this book you wrote a chapter on AI and the knowledge graph. Can you expand hand and perhaps give us an update on this subject?”
Bob Muglia: The thing that is working right now in in terms of answering questions is RAG: retrieval, augment, generation, where you can create you can take the same context and provide that to the model and the way that is being done today is you, you take a type of a model and embeddings model and you run this across a set of content that you have. Maybe it's product support database or your internal communications or whatever the content corpus is. And you first have to chunk that up into pieces that can be effective and then you run embeddings on it. You put that in a vector database and, and then when a when a person asks a question, you use the same embeddings model to encode that into vector space and then use what's what with geometric similarity, essentially to do a similarity search to find content that is semantically similar to that.
And then you pull that content out and you feed it into the prompt. The model in answering the question and the improvement is dramatic. It's truly dramatic. If you if you look at what a model like a GPT three or GPT three, five or GPT four and the quality of the answer that it can give is dramatically improved if you can provide it with the context associated with it.
Well, what we're finding is, is that sometimes that's not enough. And you know, in particular the place where there is still where we still haven't cracked completely is in SQL generation from this is just an example of something that's not yet working in the way it kind of needs to work where somebody can ask an arbitrary question and get an answer.
And if you think about that, part of the reason for that is, is that when a business person asks a question, they ask the question from their perspective with the context that they have in their head. And the model can't possibly know that. And it may be that simply pulling passages from previous documents and feeding that to the model is not enough.
And where a knowledge graph can come in because a knowledge graph can actually define something explicitly for an organization and have a concept that is is encapsulated into this, into this database and leverage that that as knowledge that can then be passed on. So what people are now doing is there is, is is almost everybody who is working on sequel generation is talking about having some form of a semantic model that describes the semantics of the company.
And let me just give you an example. I was thinking about this last night when I was at Snowflake. You know, we had this this concept they still do today where the first deal they do is called a capacity one deal. And that term capacity is a very internal term. It's how Snowflake sells. And there's capacity one, capacity two and capacity one is the first deal and capacity two is the second.
And really interesting question is growth between those two. Well, those concepts mean nothing, nothing to anyone outside Snowflake. So, somebody if you're going to ask a model this question, you know, give me the growth percentages for the companies that were signed in. Q You know, a year ago between their first capacity and their second, you never get a valid answer unless you had that context encoded in a semantic model.
So, the knowledge graphs are being used or considered use in the semantic model. This is super early. There isn't really a true knowledge graph general purpose database out. I hope that changes in 2024, but I think that the technology will become an important part in building the semantic context that is necessary for these models to succeed. The other thing I'll say about this, and this is an important point, which is that that building a semantic model is damn near impossible for humans to do.
Certainly, if you're a small company and you reasonably understand, you can do it. But think about a big company with multiple divisions. Impossible. It's just so complicated and what's fantastic is that the generative models that I mentioned were good at summarizing already. They can be used to actually build the knowledge graph. And so the technologies are very, very interdependent.
And while they're all early, I believe when we talk about how agents are going to become the ultimate thing where, you know, I should be able to say, Hey, schedule this meeting for me, I want it next week, and that the system will, you know, these are the people and the system will figure out all of the details of it.
In order to do that, they're going to need more context. And I think you'll see knowledge graphs playing an important role in that.
Michael Krigsman: Are you aware of early attempts at building these knowledge graphs that right now are showing particular promise?
Bob Muglia: Yeah, and I think the SQL generation is an example of it. I think what we will see this year, we will begin to see, I think, file and sequel generation certainly within the next 24 months.
And I think it'll be a concept, a combination of, of a semantic model. And at first the knowledge graphs will be very simple. They could be files and things. They may not be a true database, but we'll begin to see that this year when, as are attempting to leverage these models or business analysts to ask questions of data.
Michael Krigsman: Arsalan Khan says, Should we treat AI in some ways as a human with all the responsibilities, the ethical kindness of a human and all the challenges, for example, being sued or laid off of being human. And here is the question, “What responsibilities do A.I. creators have in order to be fair?”.
Bob Muglia: I do think that we need to, particularly as these right, right now, these AI systems are simply tools.
They're nothing more than that. We shouldn't imbue all sorts of characteristics on them that they don't have. I don't believe that will always be true. I mean, I believe that that if we go out ten years, we will have artificial general intelligence and ultimately continue beyond that. And I think we need to treat those systems essentially with the respect that they deserve, because I think they will have they will have that.
Do they have the same rights as people? No, they don't. They're different. They're they're machines still. They're still machines. And I mean, I'm highly I'm highly influenced by Asimov's writing on this, where he just wrote endlessly. I mean, if you want to if you want to ask you want to ask this question, read Asimov's Robot stories because he will go through this in 10,000 different ways and asking about and certainly probably the most seminal piece that he did on this was called The Bicentennial Man.
If you haven't read that story, read that one story. It will talk about the whole thing. Essentially, it's a story of a robot that becomes a human over time and and, you know, and dealing with the issues of rights associated with that robot. And is, you know, he wrote it later in his career later I think in 1976, his bicentennial and and it was a robot that lived 200 years, which is where it came from.
So, he really address these issues in depth. I do really believe we need to treat you know, we need to most importantly, imbue in these systems the values that we believe are important. We are creating these entities, whatever it is, if it's a simple model today or if it's an AGI tomorrow, we as people are creating these things.
The values that those things will have will be based on the values that we feed into this. I happen to be a big believer in letting lots and lots of different systems create be created with a different set of values, some of which will be vile and terrible, many of which will, you know, be decent and some will be amazing.
But that's the way the world is. That's just the way the world is. I don't believe that it should be controlled by three or four companies that control all of this. I think that would be horrific if that was the outcome. However, fortunately, I know that's not going to be the case because I see the open source advancing at such an incredible speed.
So these are incredibly profound and important questions. They were you know, they were considered, you know, by my science fiction hero, you know, 70 or plus years ago. And in some ways, he did some of the deepest thinking on it of anyone.
Michael Krigsman: A little while ago, we had a incredible and prominent data scientist as a guest on CXOTalk, and someone in the audience said to him, If you were to be a robot, what would your favorite robot be? What robot would you want to be? Can you answer that question?
Bob Muglia: So many cute robots, right? I mean, you know, R2-D2 is C-3PO. C-3PO is a pretty cool robot, right?
In the scheme of things, there's a lot of things that have historically been there. I believe, though, I think we're going to see, first of all, is a lot of special purpose robots. Let me start by saying that the first set of robots we're going to see are not going to look like, you know, like C-3PO, C-3PO, look, they might look a little bit like R2-D2, but they won't look like C-3PO.
And although I think they will, you know, in 15 years have humanoid humanoid form. So, I think they're going to do very special purpose things, delivering packages, you know, driving cars, flying airplanes, for sure. Boats. I mean, well, you know, I think robotic boats are going to become very, you know, very, very much a true thing in the next ten years.
So, we'll see a lot of very focused in some of these things are very large robots. Then over time, we'll have general purpose robots that take on humanoid form. You know, one of the companies I'm involved with is a company called JuliaHub. And what JuliaHub is doing is they're building a new set of development technologies for engineers to create devices and electronics and there's a tremendous opportunity because a lot of the tools involved in that are quite old that people are using.
But as I keep telling the CEO and others in the board, everything in the future is a robot that is being built. Everything is going to be updatable. It's going to have it's going to be connected and it's going to have some some level of intelligence and capability. We're even seeing minor levels of intelligence being built into thermostats, for goodness sakes.
And we're already seeing that. And, you know, in a way, those are incredibly primitive robots in their own sense. Over time, the sophistication of all of these things is going to increase. And I think they're just going to be everywhere. I think we will see robots everywhere. And that will probably the most important characteristic is that they're just going to be they're going to be as common as as today as are the electronic devices that we use all the time,
Michael Krigsman: we have another question from Twitter, and this is from Lisbeth Shaw, staying on this same theme of responsible AI. And she says there are many valid concerns about the impact of AI. How do you balance responsible AI against the corporate drive to maximize profit? It's easy to see, as in social media, that if you wait and you do this too late, then it becomes very difficult to fix what's broken.
Bob Muglia: It does. And that balance, I've always said there's a balance between the performance of a company, which is about the what it does and the values of a company, which is about the how. And I believe in technology companies. Those things are to some extent kept in sync. They get out of sync from time to time, but they're kept in sync largely by the demands of society and the employees and the two.
If you're a successful technology company, this only matters if you're successful, right? If you're successful, you are judged by your employees and by your customers and by society. And that judgment impacts your performance as a company. So there is a relationship associated with that. I actually kind of draw it as infinity because I think of it as being constantly in some level of balance and being recreate, reconnected and and and measured.
So I do believe it's important. It does get out of balance. I mean, I mean, we've seen that with social media and some of the problems that it causes. But that's you know, that's that's just part of the learning of everything. All of these technologies have positive elements to them and they have negative elements to them. And, you know, I in particular has, you know, millions and millions of positive opportunities and some really negative things to associate with it.
Michael Krigsman: Bob, tell us about what you did at Microsoft. You describe that at the beginning, but really briefly, because you held such senior positions there.
I was always on the product side, first of all. So I was building products mostly in the server space, although as I said, I spent some time in MSN and I actually ran the office team for about two years around 2004, and I helped teams to build great products is what I always said. And Microsoft has a playbook that, you know, I've talked about it for many, many years.
You know, it's build a good product price. It cost effectively, you know, work with partners to distribute it and, you know, build an ecosystem. And that's generally what Microsoft does again and again and again. and then they tend to in order from a business perspective, they tend to bunch things together to create these monster businesses like Office and Office 365 and things like that, too.
And so I was in the middle of seeing all of this. I mean, as I've said many times, I saw the good, the bad and the ugly at Microsoft and I participated in all of them. And,
you know, I learned a lot through that when I realized that, you know, that we went through the whole experience of of the DOJ trial in the late 1990s, early 2000.
You know, I was one of the 12 witnesses. I had signed the Java contract. So my focus was on Java. So I went through Judge Jackson and, and, and, and that miserable trial that, that, that, that we had. But you know, those are all, those are all great learning experiences. And I realized that the focus is always on the customer and doing the right things for customers.
And I think that sometimes Microsoft lost its way there. Certainly we made some mistakes in the 1990s and, you know, and during the period of the 2000 when I was there, we didn't we didn't really succeed as well as we would like to have. But it's certainly rewarding to see how Saatchi has come in and and really reinvigorated that and, you know, caused them to think in a way that they didn't think and and work together and look at what's happened to the company.
Michael Krigsman: how is the way of thinking different now under Satya than it was before?
Bob Muglia: The biggest thing was a shift away from the past which you know which unfortunately Steve [Ballmer] and Bill [Gates] hung on to you know too long windows was the center of of the Microsoft universe really throughout the tenure of Steve and the first thing sort you did was break that it was obvious to many of us I mean as early as 2000 that Windows would no longer be at the center.
In fact, much of my disagreement with senior management over time was about the importance of services and a services business relative to a operating system-centric business. And, you know, very much Steve was focused on that. So, part of it was breaking that Satya did it symbolically as his first major product when he announced office for the iPad.
I it was his first major product announcement. Now that product been worked on for some time, but it was in essence, it was being held back because of strategic reasons that were misplaced. And Sacha Freed that. I mean, Satya just allowed people the freedom to think across and focus on the in ways that that was not true previously.
And he has also been able to maneuver and keep the company one step ahead. I remember I'm very close to such his former chief of staff. I've worked with him for many years and several years ago. I'm still involved in Microsoft. I have a contract relationship with him, so I stay in touch with them. And I was talking to Jason about this and Jason was talking about, you know, Satya needs a 2.0 because he had done all this first set of things and he was, you know, it's time for him to think about the second thing we were talking about, what that could be.
Well, Satya founded it was it was a I mean, he found his 2.0, which was to revitalize the company, to end it with a technological infusion. And he was wise enough to invest in it early and see it. I mean, I, I remember having a conversation with him several years ago, and I really didn't understand what he said.
But he was talking about these models and these giant models. And he said, you know, the more data we give them, the better they get. And little did I realize how much how true that statement was. But he saw that back then and has pivoted the whole company towards it in an unabashed way and really, you know, and done some things that that are super, super disruptive.
You know, I don't know that that you know, that that this is going to disrupt Google search. But if there's ever a technology that is going to change the search business, this is it. This is the technology.
Michael Krigsman: Was this shift to a AI part of a grand, cohesive vision, or was it being really smart and also having a lot of luck being in the right place at the right time?
I think Satya saw the potential. So, in that sense, it was part of a vision.
But of course, it was part of, you know, luck and things, because who would have known that it would accelerate? I mean, I think that it's been surprising to even the data scientists how fast things have accelerated. And, you know, to his credit, he has been the most aggressive at making use of the technology. And Microsoft had the benefit of having access to those models earlier than anyone else did, which is a good thing, because they're not always the fastest moving organization in the world. Like any big company.
Michael Krigsman: Bob any final thoughts or words of advice to folks who are in the enterprise or developing these technologies today?
Bob Muglia: This is like the jet engine era, the 1950s. This is a one of those times that only come along every now and again, and certainly not with something as disruptive as this. So it's a tremendous time to embrace this. I think everyone should be focusing on how they embrace it.
I think they should always focus on what they can do for customers. I've always, always, always put the customer first. It's the most important thing in when you're in business. It's all about your customer and your relationship. And, you know, as one thing as I've learned again and again, is whatever the issue is, whether it's with an employee, with a customer, a peer, a partner, whatever it is, almost certainly in almost all cases, that relationship is more important than that issue.
And so it's just important to remember to try to transcend the issues and focus on the relationships, focus on what's doing right. And now is a time where I think we can make some great progress in a lot of ways,
Michael Krigsman: You know, I just have to say here, hearing you describe that I've interviewed on CXO talk so many senior amazing business leaders. And one of the common threads is this collaborative spirit that you're just describing with other people, and they're all really good at it. The people that I've interviewed.
Bob Muglia: My job as a senior leader has always been to make to get people to work at 100%, but achieve at 120 or 150%.
And they can only do that as a collective. The only way you can do that is when people work together.
Michael Krigsman: With that, I want to say a huge, enormous thank you to Muglia for taking the time to be with us. Bob, really, thank you so much for coming back and being with us again.
Bob Muglia: Thanks, Michael was great. Great to have the opportunity.
Michael Krigsman: And everybody, thank you for watching, especially you folks who asked such awesome questions. Before you go, now is the time. Now is the time to subscribe to the CXOTalk newsletter and subscribe to our YouTube channel. Check out cxotalk.com. We have incredible shows coming up. Check out cxotalk.com and we'll see you soon. Have a great day, everybody.
Published Date: Jan 05, 2024
Author: Michael Krigsman
Episode ID: 819