Knowledge Graphs 101

With Kendall Clark, the Founder & CEO of Stardog

Kendall Clark, founder and CEO of Stardog Kendall Clark
Founder & CEO of Stardog

Kendall Clark is the founder and CEO of Stardog, a data integration company that takes a philosophical approach to cloud computing. Customers like NASA use Stardog to understand data relationships with semantic data models and unleash new insights.

Satyen-Sangai, CEO of Alation Satyen Sangani
Co-founder & CEO of Alation

As the Co-founder and CEO of Alation, Satyen lives his passion of empowering a curious and rational world by fundamentally improving the way data consumers, creators, and stewards find, understand, and trust data. Industry insiders call him a visionary entrepreneur. Those who meet him call him warm and down-to-earth. His kids call him “Dad.”


Kendall Clark
Founder & CEO of Stardog

Kendall Clark is the founder and CEO of Stardog, a data integration company that takes a philosophical approach to cloud computing. Customers like NASA use Stardog to understand data relationships with semantic data models and unleash new insights.


Satyen Sangani
Co-founder & CEO of Alation

As the Co-founder and CEO of Alation, Satyen lives his passion of empowering a curious and rational world by fundamentally improving the way data consumers, creators, and stewards find, understand, and trust data. Industry insiders call him a visionary entrepreneur. Those who meet him call him warm and down-to-earth. His kids call him “Dad.”


Kendall Clark
Founder & CEO of Stardog

Kendall Clark is the founder and CEO of Stardog, a data integration company that takes a philosophical approach to cloud computing. Customers like NASA use Stardog to understand data relationships with semantic data models and unleash new insights.

Satyen Sangani
Co-founder & CEO of Alation

As the Co-founder and CEO of Alation, Satyen lives his passion of empowering a curious and rational world by fundamentally improving the way data consumers, creators, and stewards find, understand, and trust data. Industry insiders call him a visionary entrepreneur. Those who meet him call him warm and down-to-earth. His kids call him “Dad.”

Kendall Clark: (00:03)
It’s a little bit like asking the first person you ever date to marry you. It’s not that that’s definitely a mistake and will never work out, but it’s risky. This is really what happens when you commit to doing integrations in the storage layer, it’s committing really upfront once for all, how things are, and then when things aren’t that way, that’s why you have to rerun the jobs and make the copies. Because you’ve decided, you’ve foreclosed all the possibilities really aggressively.

Satyen Sangani: (00:38)
A data set might seem objective, a hard fixed immutable thing that represents the world exactly as it is. The reality is that most data is the result of a series of human choices: Which data to collect? Which data to transform? Which data to summarize or aggregate? And which data to count? Every one of these choices is a modeling decision. Every one of these transformations is a human choice, a choice to decide what to measure and what’s not important.

So what happens when an analyst has a new idea about how to organize and understand data? They have to start from scratch. They have to recreate a new version of the data, a new data set. And then, they have to go through the process of remodeling, retransforming and resummarizing the data all over again.

But what if they didn’t have to? What if we could make data modeling far more flexible?

Satyen Sangani: (01:31)
Kendall asked himself this question and it led him to founding Stardog, an enterprise knowledge graph platform. Kendall is a true data radical. In addition to founding Stardog, he has a PhD in the philosophy of religion. He was also CEO of Complexible and the managing editor of xml.com.

Satyen Sangani: (01:48)
Today, we’ll talk about data transformation, we’ll talk about knowledge graphs, and we’ll learn when and how they ought to be used, as well as which use cases and users will leverage them. So even though the terrain may be a bit technical, Kendall and I break all of this down. His background and knowledge make for a really fun conversation.

Producer: (02:14)
Welcome to Data Radicals, a show about the people who use data to see things that nobody else can. This episode features an interview with Kendall Clark, founder and CEO of Stardog. In this episode, he and Satyen discuss Knowledge Graphs, their use applications, the relationship between philosophy and management and so much more.

Producer: (02:33)
This podcast is brought to you by Alation, our platform makes data easy to find, understand, use, and govern so analysts are confident they’re using the best data to build reports the C-suite can trust. The best part, data governance is woven into the interface, so it becomes part of the way you work with data. Learn more about Alation at A-L-A-T-I-O-N.com.

 


 

Kendall’s path from philosophy to computer science

Satyen Sangani: (02:56)
Kendall is what former guest, David Epstein, would call a generalist. Like me, Kendall and I came to the software industry from a non-traditional path. I asked him to describe his journey.

Kendall Clark: (03:06)
I think have a unusual background. I’m not sure how distinguished it is. It’s a little bit of a strange path to being a startup guy. I’ve still never had a computer science class, so I’m self-taught in computer science, which is a little bit strange. But I did bachelor’s, master’s and PhD in philosophy of religion, and philosophy, generally, but philosophy of religion specifically for my PhD.

And I guess one of the things to say about that inquiry, that discipline is, you end up getting comfortable with rules and rule-like systems and with philosophical arguments [which] are sort of algorithms over a weird domain. So that when I realized about 80% of the way through my PhD, that I actually was not going to be an academic, that the academic world was not for me, I needed another thing to do effectively. So this is not distinguished. This is not interesting. This is just real world stuff like, “Hey, I got to pay my bills. I need a career.” And I was 26, 27.

Satyen Sangani: (04:05)
And the philosophy of religion, there was no job. I guess there was no nonacademic job, or job that you wanted.

Kendall Clark: (04:13)
That’s right. If you don’t do academia as a philosopher, you pretty much don’t do philosophy formally, there’s no other place to do it really. I guess Google hires a few ethicists from time to time to… And actually it does occur to me now that one of my buddies is VP of ethical AI at DataRobot. So that’s a thing you can do if you… I studied moral philosophy, ethics, philosophy of religion. That was a thing I could have done, but all these people who do this now are 15 years younger than me. No one was doing this in the late 90s.

Kendall Clark: (04:43)
So I discovered internet, the web, Linux, Unix and then Linux all basically in like March-April of 1995. And I was finishing my PhD in religious studies and philosophy of religion and I just fell down this rabbit hole of, I can have a workstation on my old PC if I get in a floppy disc and download Slackware and it takes six weeks. And I just fell into this world and I thought, “Okay, this is something I can do.”

Kendall Clark: (05:10)
I remembered my first IT job, we were building web apps. I was working in Dallas, and we built an internal portal for a large California home manufacturer, really big, big national brand home builder. And it was my first exposure to enterprise IT, so this is ’97, ’98. And I was astonished to learn that inside of a large company, there wasn’t just one big data source. I just assumed quite naively, right? I knew a lot about computer science, I didn’t know that much about enterprise IT. I just assumed all the data of a company was in one place. Because it just made sense to me to assume that. So you can imagine how profoundly shocked I was to learn, “No, in fact it’s the exact opposite of that. The data’s in many, many places typically.” And in subset, Stardog still works on that problem. So really the first big problem I encountered in IT is the one that stuck with me. I guess you can say I’m nothing if not stubborn.

Kendall Clark: (06:06)
So I really came to data by observing that, in the real world, data is it’s natural state, it’s disconnected, and that disconnectedness is a challenge for everything else an organization wants to do with that data.

 


 

Stardog’s approach to knowledge graphs

Satyen Sangani: (06:19)
And that stubbornness is what led Kendall to found Stardog years later. I ask him to describe the organization and its mission.

Kendall Clark: (06:27)
We call Stardog an enterprise knowledge graph platform. It’s a back-office technology. We started very simply, there’s a user interface for implementation, but it exists in some ways deep in the data and analytics infrastructure stack. Really sits between the presentation layer or the business logic layer, that’s the layer where things get done, reports get written, people consume data, people make decisions, people generate new insight.

Kendall Clark: (06:54)
And then below that, is the wild west world of the data, which is always a bit of a mess. A company like Stardog could only really exist in an era of effectively infinite computing power, which is relatively new for the human race. But we had a lot of storage as a civilization, much, much sooner than that, right? This is what motivates, we’ve always integrated data by moving and copying it because it was really the only choice we had, right? So it’s not that people were dumb, people were not dumb, they were constrained, right? So it’s the smart, best choice you can make, if it’s the only choice you can make. And for a long time, it was the only choice.

Kendall Clark: (07:30)
The alternative for us is to say, “Well, rather than integrating data at the storage layer, what if we integrated data to computation layer?” Entropy has happened always to data and what people want is a means to have that data connected so that they can effectively do their jobs, get the insight, win the prize, whatever they’re about.

Kendall Clark: (07:52)
The point of Stardog as a software system is really to let its users query, search, maybe train a machine-learning model, do some data quality, a variety of services, one might want to do with data to accomplish some end, to provide a mechanism for those services to be accomplished without the data having to be moved or copied anywhere. We just take very seriously the fact that all the data that, say an enterprise, possesses has a natural resting place, it lives somewhere. And rather than thinking about moving the data into a new place in order to process it, we could just build software systems that process that data where it lives and save a bunch of time, save a bunch of heartache, save a bunch of upset. And in the end get the answers and the insights we want faster.

 


 

What is a knowledge graph?

Satyen Sangani: (08:42)
A lot of people talk about knowledge graphs. I think it’s the term that’s been widely discussed, widely used, I hear it all the time more and more. Help educate the audience, what is a knowledge graph? What does that mean?

Kendall Clark: (08:52)
A knowledge graph can mean both, a collection of facts that constitute some knowledge or some data about a domain, about some part of the world, something that people care about. And it can also mean a software platform or software system that manages or queries or assembles or processes in some ways that collection.

Kendall Clark: (09:14)
The thing to say about a knowledge graph is that it’s a graph, which means it has a particular data model, it’s a very simple data model. You typically talk about really three things in a graph data model. You have things that are called nodes or entities, which are often depicted as circles. In a graph database about your podcast series, you would be a node, you would be the host node, I would now be a guest node, right? And then there will be an arrow or a link, a line, sometimes what’s called an edge between these two nodes and that would represent the relationship we have, you hosted a podcast on which I appeared. If somebody says, “Do Kendall and Satyen know each other?” Sent that query to a graph database about your podcast, it would say, “Yes, they know each other in this precise way. They know each other in the web host-guest, interviewer, interviewee…” However you want to think about it, relationship.

Kendall Clark: (10:13)
Now a knowledge graph is expected to be at a higher level of abstraction than raw data facts. And it’s supposed to include things like logical theories, axioms, rules, other constructs that let you make a semantic interpretation of the data, which is to say much more like the theories that we carry around in our head as people, right? Rather than just a raw collection of data bits.

Kendall Clark: (10:44)
A knowledge graph often it should have a richer, more expressive. And here, I mean, in the technical sense, a computationally more powerful data model and data language to express much more complex relationships than, say, a relational database, right? A relational data base is, again, a lovely bit of software and a very dense collection of human ingenuity for 50, 75 years. But its data model is relatively inexpressive, right? You can really only express a few basic relationships between the parts of a relational database, whereas in a knowledge graph, you have effectively some fragments or full-on, first-order logic.

Satyen Sangani: (11:27)
Fair to say that in a graph, you can model, I think you used the word more sophisticated, more rich relationships between the nodes, between the nouns, between the things. And you can also represent logic, right? You can say, “Satyen on the podcast by himself talking to himself as both a host and a guest.”

Kendall Clark: (11:50)
Correct. And that would be to say that relation can be reflexive, right? So logical reflexivity means you can have that relationship with yourself.

Satyen Sangani: (12:01)
Where you couldn’t do that, for example, in a relational database.

Kendall Clark: (12:06)
You can do that in a relational database if the people programming the relational database, not the vendor, but the user wants to write a bunch of code to interpret some arrangement of the rows and columns to mean reflexive. Right? Of course, you can do it, but in a knowledge graph, this just comes for free, it’s part of the mechanism, right?

There’s a much, much richer… And, again, richer here doesn’t mean richer like Shakespeare is more rich than, say, Dilbert. It doesn’t mean richness in that poetic, human, emotional meaningfulness. Rich here means strictly one computer language is more powerful than another because it can represent everything that the less powerful language can represent. And it can represent things that the less powerful language cannot represent, right? You can strictly speak and just compute more and say a Turing complete language than a language that isn’t Turing complete.

Satyen Sangani: (12:58)
The consequence of that is you can ask more sophisticated questions that are closer to how people actually think.

Kendall Clark: (13:07)
Yes. That’s exactly right. An analogy that everyone who listens to this will certainly understand is, think about how you tell your friend over the phone to manipulate a spreadsheet. It’s very procedural, you say, “The last column is where I put our social security numbers. So you should stick another column between that one and the penultimate one and we’ll put the update date in that column.” We talk about it very physically in terms of the layout, right? Where the data is.

But then to continue the spreadsheet analogy, there’s another part of how we interact with spreadsheets, one thing that makes spreadsheets quite wonderful tools, that’s purely declarative, right? We just say somewhere in a cell, “Here’s a formula for computing interest rate”, or, “The net present value of cash, it’s just in this cell.” And no matter what else happens in the spreadsheet, as long as the data is somewhere, that thing will produce the right value. And it just always does it, right? A spreadsheet just always has the values that the cells compute. A spreadsheet is both declarative and procedural in that way.

Kendall Clark: (14:15)
The other implication of using a richer data model to describe your data is precisely that people can trade in what people are good at, which are these high-level, what’s typically all declarative descriptions of how they want the world to be. And then again, just let the computer figure it out. As you said, we might write a rule that says, “On this podcast, the relationship between host and interviewer cannot be reflexive.” We don’t want to have any cases where Satyen’s interviewing himself-

Satyen Sangani: (14:47)
We definitely don’t. No, we don’t.

Kendall Clark: (14:48)
No offense, man. That just will be a rule. Let’s say it’s your rule. So we’re not imposing this on you, you agree with this. So we might say, in fact, the interview’s relationship is anti-reflexive. Anti-reflexive is a very crisp, well-understood logical description to say… Now what that turns into, is it maybe a data quality parameter. If we’re trying to do some data integration and someone’s made a mistake somewhere, the computer’s saying, “Nope, we can’t have that. I know that’s wrong.” “How do you know that’s wrong?” “Well, you told me that this relationship is anti-reflexive. If we have a podcast, we need at least one guest and we need at least one host.” Because we have to have two nodes, there have to have been two parties to an interview, two parties to a conversation. We don’t have solo conversations on Data Radicals, right? That’s not you or your producers saying of some future data you’re not even aware of, “Procedurally no, delete that. That’s wrong.” That’s you saying, “These are just what the rules are for this part of the world.” And then the computer applies that and good stuff happens.

Satyen Sangani: (15:55)
Should every software system be built on a knowledge graph?

Kendall Clark: (15:58)
Of course not, literally construed, no. I mean, software systems have a very wide range of requirements, and, in some cases the data problem isn’t really that important. To answer what I think the healthy, provocative part of your import of your question was Gartner thinks…

Now, I don’t believe everything Gartner says, but I think they do think about what the better enterprise IT landscape would look like fairly well and pretty consistently. They’ve been saying for a couple years that something similar, which is, applications increasingly in the enterprise should be really very user-interface and role-specific kind of views onto a common pool of enterprise data that just exists somewhere. I think this is a widely-held belief about where the industry should go, that applications are really about human affordances rather than directly manipulating some unique data.

Kendall Clark: (16:59)
I’m struck by the fact that, let’s say at the US Postal Service, there are people who work in the back office and they manipulate big systems, they sit at a desk and they’re standard, what you think of as knowledge workers, right? And then you got the delivery people who are out there dealing with the rain and the sleet and the snow and the dogs and the weather and all that stuff, and it’s a hard job. They have very specialized computers, little devices that help in that setting.

Kendall Clark: (17:26)
In that world, both of those US postal service workers share some common pool of postal service data, they want to be looking at the same underlying reality that the data represents. But the applications that they use may be very, very different because of the human differences between carrying a mailbag and being on a busy street and trying to help somebody get a package versus sitting back at the office with a big screen and a big computer and big mug of coffee. And you’re writing some report or some query, right?

Kendall Clark: (17:55)
In that sense, I do think that enterprises that treat their data as a first class object and really take that seriously, get the chance to rethink in a pretty thorough going way, what an application means, right? And what you really want from an application and what that looks like.

 


 

Architecting for flexibility

Satyen Sangani: (18:15)
It strikes me that ultimately there’s a trade-off between what’s easiest for the user to use, and then how much flexibility they want the system and how much they want to learn the technology and the like. What I think you are basically saying is, “Look, there’s a set of cases where you don’t necessarily know what you don’t know, and there’s a lot of implied flexibility that’s required because,” to your point, “to get to that level of usability and customization, you have to divine what you may not know a priori when you’re designing the system.” And those are great cases for when a knowledge graph could be useful. Do you feel that’s fair or right?

Kendall Clark: (18:52)
I think that’s right. I think the need for flexibility arises in many different ways. And uncertainty is one of the drivers of the need of agility or flexibility. Uncertainty is one. Just sometimes you can just be… I’m absolutely certain that my requirements are going to change a lot and very rapidly, right? But let’s say within a range that I can predict with high certainty. So it’s not that I have certainty, I just have pressure, right? The business moves fast, so I think that’s another case.

Kendall Clark: (19:23)
The other thing that we’ve seen where we do particularly well in what were called regulated industries, the paranoid industries, right? I didn’t quite understand that for a while, but I finally realized that the other advantage to a knowledge graph as an integration mechanism as opposed to, say, a relational model is knowledge graphs really do promote an extraordinarily high degree of reusability. It’s just always the case that a knowledge graph gets extended by getting more nodes and edges added to it. Customers never just throw it away and start over from scratch. And I still really never seen a relational database effort start with, by people saying, “Hey, we’ve got to make this new application. We’re going to use a relational backend. Let’s use the schema that we used for that other thing before and start there.” Nope. It’s always a fresh page from scratch new-data modeling effort every time.

Kendall Clark: (20:21)
I think relational systems have a very low degree of schema reusability, which in absent, any other specification who cares, frankly, doesn’t really matter, but in regulated industries, it’s super important to get it right because you’re going to have to stand tall before the man and account for yourself, that’s what a regulated industry is. One of the reasons I think knowledge graph technology seem more uptake there is exactly the reusability.

When you really have to get it right, once you’ve got it right, you don’t want to mess around with it anymore. This is the opposite of flexibility/uncertainty, right? This is certainty that comes from, “Hey, we finally figured out how to model those derivative credit default swaps. And it’s correct and the regulators are happy and we haven’t crashed the economy and we’re making money and just leave it”, right? Don’t start over. If you do have to do it again, that just introduces the risk that we may get it wrong. It informs how you manage data, how you think about data, how you build human processes, all that stuff.

Kendall Clark: (21:24)
This is the sense in which I published a piece last year that said something like, “Not only is data management not finished, it’s far more strategic than anyone ever imagined.” Because we tend to think of data management, ho-hum, it’s ETL jobs. It’s just plumbing its infrastructure. I think events have proven that to be quite false. It really is quite strategic because it’s a significant limiting factor to the questions you can ask and answer of the data. And ultimately that’s a significant limiting factor to what you can do as a business.

Satyen Sangani: (21:57)
In our founding story, the way I started thinking about and coming to the problem of solving Alation, got me down this… I wouldn’t call it rat hole is a pejorative and it sounds really bad, but it got me down an inquiry path of looking at semantic layers and thinking about knowledge, graphs and ontologies and sparkle as an example.

And it’s funny because I met my co-founder who at the time was Google engineer and has then resigned, and told him, “Well, aren’t we going to implement this on an ontology and in a graphic database?” And he’s like, “We’re not going to do that.” Because we know what the user wants and we know what the questions they’re going to ask and we know what they’re searching for and we know what the user’s going to need. We don’t need to reinvent the wheel every single time. And we don’t need massive levels of reusability. When you’re searching for a table, you’re going to do the same thing every single time. A lot of the system doesn’t need to be built on a knowledge graph.

Kendall Clark: (22:49)
What he was saying effectively, you were there, but I’m guessing, “We need those last few percentages of absolute control and performance and economy of scale rather than any more flexibility.”

Satyen Sangani: (23:01)
Right. We know that we have different persona-types, like an analyst or a data scientist, and we can build differences for them, but we don’t need the ultimate level of flexibility because we don’t have ultimate uncertainty. We know what the user wants. And I think-

Kendall Clark: (23:14)
Sorry to cut you off, I have the same irony, we don’t use a knowledge graph internally to manage the sales pipeline of Stardog. Oh, the irony, right? Oh, what do you mean? You don’t eat your dog food? No, we run a sales process like everyone else. Salesforce is sufficient.

Satyen Sangani: (23:29)
Yeah. And I think it ultimately comes down to, ultimately, with your software application with whatever it is that you’re trying to do? What are you trying to do and is a knowledge graph useful? It’s funny because when we started this catalog, a lot of people would ask me, “Well, are you capturing structured or semi-structured or unstructured data in your catalog and how are you doing search?”

Satyen Sangani: (23:51)
The interesting thing, and I think this comes back to that question, because the interesting thing is like any structuring of the data is in some sense an analytical exercise. You’re choosing to label what’s important, what’s not important. There’s a photograph you’re choosing to say, “The shirt’s important, but the background color in gray is not necessarily important and that’s part of my data model.” Every structure, whether it’s an OLTP application or OLAP model and a Snowflake, all these things are, I think on some level, somebody’s abstracted knowledge out of the world around them and chosen to ignore a whole bunch of things that don’t matter in the context of that model.

Satyen Sangani: (24:30)
And it does strike me that this idea of what you are saying is really powerful, which is like, “Look, all of the structuring that’s happening in the integration layer, that’s on some level lost knowledge or potentially lost knowledge, that if we could just allow the structuring to be itself a lot more flexible, could end up then unlocking a lot of knowledge because there’s more reusability. People can learn and build on top of things that have already been built, but they can also explore things that have not already been built. And so it just gives you a wider palette on which to paint, the ability to ask more complicated questions. And that could be a very powerful way of approaching the world.” That seems really cool as an idea.

Kendall Clark: (25:12)
I think that’s exactly right. I mean it’s a little bit like asking the first person you ever date to marry you, right? It’s not that that’s definitely a mistake and will never work out, but it’s risky, right? This is what we often call in large scale systems early-versus-late binding. How long can you defer or delay before you have to give the final answer? Maybe if you go on a lot of dates and you experience more of what it’s like to be in relationships, you can make a better choice. Why be in a hurry, right? This is really what happens when you commit to doing integrations in the storage layer. It’s committing really upfront once for all how things are, and then when things aren’t that way, that’s why you have to rerun the jobs and make the copies, because you’ve foreclosed all the possibilities really aggressively, right?

Kendall Clark: (26:12)
You’ve said, “I don’t want to take time. I don’t want to defer searching this complex space to be adapt, more flexible for what happened next. It’s just like this because that’s the way it seems to me.” Which is often true. Every schema is a choice that reflects human values and priorities and understanding. There’s no way around it. That’s just what it means. We impose this meaning and order on otherwise, what is just meaningless, like data. And our argument along has been like, “Look, it’s not that you should get rid of Snowflake. This is a net new capability that lets you relate data no matter where it is to the data that’s anywhere else based on what it means.” And then you can query it. And that has a lot of value for your organization, but it has value in the context of everything else you’re doing.

Kendall Clark: (26:58)
Right now maybe yes, it will help you save some money. It will help be more agile, more flexible. These are good things. We can capture these in terms of value, there’s ROI, but it’s not a knowledge graph or anything else you can only choose one, kind of situation.

 


 

How do you build a winning culture? Practice charity

Satyen Sangani: (27:13)
To close out our conversation. I asked Kendall to describe how his experience has influenced the culture he has built at Stardog.

Kendall Clark: (27:20)
With respect to the internal culture of Stardog, I think it’s, for me, been much more informed by my early career as a philosopher. In particular and some of the values we talk about as part of the Stardog family includes things like practice charity, right? And I always have to explain this one because typically when a corporate executive says, “Practice charity”, they mean, “Do philanthropy.” And I also mean that, but that’s not what I mean by practicing charity, by practicing charity, I mean practice moral and epistemic charity, which is to say, when you’re talking to someone, when you’re trying to work with someone to discover the truth, to try to discover the way forward, to try to discover even a dirty revenue hack that you need to put into place just for a quarter, you’re trying to accomplish some end, you need to always be construing what the other person’s position in the most charitable possible light.

Kendall Clark: (28:12)
Like here, we’re having a conversation, and one of the things that makes this conversation go better rather than worse is that I’ve misspoken a few times because I’m a person, and I didn’t say precisely what I meant, but you always came back with this interpretation of what I meant that put me in a good light. You assumed the best, right? Now, you could have done otherwise. It’s always very frustrating when you’re talking to somebody and everything you say they’re putting the worst possible spin on it. Now they’re not practicing charity. They’re not being generous in how they’re resolved the uncertainty of the act of interpreting another person’s words. “What did he mean by that? Well, I can see he could have meant this good thing. He could have meant a thing that really makes him quite a nasty person.” That’s a free choice I can make.

Kendall Clark: (28:57)
One of the things I learned from being a philosopher, I know this is not necessarily philosopher’s reputation, but it’s what the actual work of philosophy is, that when you’re working, let’s say, with your professor on your thesis, or with maybe your department mate on some paper, some argument of a paper, everybody gets to the truth faster, right? It’s a more efficient, more effective, more goal-directed process for everybody to be construing what everyone else says in a very positive light. And I find that that is not just a philosophical virtue, it’s a virtue of just good, decent humane people.

Kendall Clark: (29:31)
And I think of human systems like a startup that are effective, that are joyous, that are supporting of people’s both individual aspirations, talents and goals, as well as the shared aggregate talents or goals of the organization, it’s just a good way to be, right? And it applies to a lot of life. We’re often faced in the mystery of the other person’s behavior with some choices about how we construe what they’ve done, what were they trying to get at? And the people who are very pessimistic and uncharitable and they’re just not fun to be with because no matter how talented those people are, people who have a high cost of interpersonal relations are never worth their skills, their ability as salespeople or engineers or CEOs for that matter. Because that tax on interpersonal on human-to-human interactions, it’s a pernicious tax, right? And it systematic and it spoils everything. It’s not limited to individual interactions.

Satyen Sangani: (30:37)
Well, Kendall, just a phenomenal conversation. And just the range has been super cool, so it’s been fun to have it and glad to meet you. I think glad for our listeners to be able to hear about all these topics that sound very complicated, but I think that you’ve done an amazing job to help simplify. So thank you for taking the time to come on the show and we look forward to having you on at some point in the future.

Kendall Clark: (31:02)
I’ll just say, who says you can’t teach old dogs new tricks? You’re a former economist, I’m a former philosopher. Everybody always says it’s the second act of American life that distinguishes Americans from… I think this was in the Great Gatsby or something, I’m butchering the quote. But I very much think of myself in that way. So it’s always good to meet people who didn’t do computer science as undergraduates and find themselves running software companies.

Satyen Sangani: (31:28)
We’ll try to expand that cool club. Not sure how big it is, but we’ll find them.

Kendall Clark: (31:33)
It doesn’t have to be big to be cool. I agree with you.

Satyen Sangani: (31:39)
In business, as in life, the only constant is change. Data can help us adapt to change and help us seize new opportunities. If you take nothing else away from this episode, remember that if our systems are rigid, moving quickly becomes impossible.

Satyen Sangani: (31:56)
Kendall’s insights reflect why in many situations, a flexible design for data can be so valuable. Thank you to Kendall for joining us for this episode of Data Radicals. This is Satyen Sangani, co-founder and CEO of Alation. Thank you for listening.

Producer: (32:13)
This podcast is brought to you by Alation. Is your organization ready for its next compliance audit? Data governance can help you pass that audit while also supporting innovation, accelerating analytics and mitigating risk. Read this evaluation of 12 data governance solutions at alation.com/DGQ3. That’s alation with an A .com/DGQ3.

Other Episodes You Might Like :

Start with Story, End with Data

Ashish Thusoo

Ashish Thusoo

Founder of Qubole and Creator of Apache Hive

Subscribe to the Data Radicals

Get the latest episodes delivered right to your inbox.

Marketing by