Podcast: Play in new window | Download
Subscribe: Apple Podcasts | Spotify | Amazon Music | Android | Youtube Music | RSS
Knowledge graph technology has been around for decades. The benefits so far accruing to only a few big enterprises and tech companies.
Juan Sequeda sees large language models as a critical enabler for the broader adoption of KGs. With their capacity to accelerate the acquisition and use of valuable business knowledge, LLMs offer a path to a better return on your enterprise’s investment in semantics.
We talked about:
- his work data.world as Principal scientist and the head of the AI lab at data.world
- the new discovery and knowledge-acquisition capabilities that LLMs give knowledge engineers
- a variety of business benefits that unfold from these new capabilities
- the payoff of investing in semantics and knowledge: “one plus one is greater than two”
- how semantic understanding and the move from a data-first world to a knowledge-first world helps businesses make better decisions and become more efficient
- the pendulum swings in the history of the development of AI and knowledge systems
- his research with Dean Allemang on how knowledge graphs can help LLMs improve the accuracy of answers of questions posed to enterprise relational databases
- the role of industry benchmarks in understanding the return on your invest in semantics
- the importance of treating semantics as a first-class citizen
- how business leaders can recognize and take advantage of the semantics and knowledge work that is already happening in their organizations
Juan’s bio
Juan Sequeda is the Principal Scientist and Head of the AI Lab at data.world. He holds a PhD in Computer Science from The University of Texas at Austin. Juan’s research and industry work has been on the intersection of data and AI, with the goal to reliably create knowledge from inscrutable data, specifically designing and building Knowledge Graph for enterprise data and metadata management. Juan is the co-author of the book “Designing and Building Enterprise Knowledge Graph” and the co-host of Catalog and Cocktails, an honest, no-bs, non-salesy data podcast.
Connect with Juan online
- Catalog & Cocktails podcast
Video
Here’s the video version of our conversation:
Podcast intro transcript
This is the Knowledge Graph Insights podcast, episode number 19. The AI pendulum has been swinging back and forth for many decades. Juan Sequeda argues that we’re now at a point in the advancement of AI technology where businesses can fully reap its long-promised benefits. The key is a semantic understanding of your business, captured in a knowledge graph. Juan sees large language models as a critical enabler of this capability, in particular the ability of LLMs to accelerate the acquisition and use of valuable business knowledge.
Interview transcript
Larry:
Hi, everyone. Welcome to episode number 19 of the Knowledge Graph Insights podcast. I am really delighted today to welcome to the show Juan Sequeda. Juan is the principal scientist and the head of the AI lab at data.world. He’s also the co-host of the really good popular podcast, Catalog & Cocktails. So welcome, Juan. Tell the folks a little bit more about what you’re up to these days.
Juan:
Hey, very great. Thank you so much for having me. Great to chat with you. So what am I up to now these days? Obviously, knowledge graphs is something that is my entire life of what I’ve been doing. This was before it was called knowledge graphs. I would say that the last year, year-and-a-half, almost two years now, I would say, is been understanding the relationship between knowledge graphs and LLMs. If people have been following our work, what we’ve been doing a lot has been on understanding how to use knowledge graphs to increase the accuracy for your chat with your data system, so be able to do question answering over your structured SQL databases and how knowledge graphs increase the accuracy of that. So we can chat about that.
Juan:
But what I’m interested next now is I’m calling this how do we extract knowledge from people’s heads? And this isn’t new, this is not new, this is back from the ’80s and early ’90s, all the work on knowledge acquisition. People wanted to create rule-based systems and expert systems, and you need to have knowledge engineers, who go talk to the domain experts and be able to extract their knowledge and codify that. Well, that was hard, and it was brittle, and it was not scalable. And now we have a very fantastic tool called LLMs that I’m very excited to use them to be able to find ways to scale that back. So I think what’s old is new, and I think the opportunities to use these new tools to do the work that we’ve wanted to go do, but we haven’t been able to go do, which is just to improve knowledge management.
Juan:
And one of the reasons why I argue that, we still try to solve the same problems. We’re solving the same problems today that we had 30 years ago, but the core problems continue to be there, and I think it’s this lack of knowledge. Now, we have a way to scale this out so people can stop complaining, saying, “Oh, it’s just too hard, too expensive.” That’s my long-winded answer of what I’m up to.
Larry:
Yeah, well, I’d love to jump right into that, because there’s all these intersecting things, at least in my world. I run in a lot of different worlds, the content world and knowledge management folks. Anyhow, like you said, in the AI world, there’s this longstanding, the expert systems and all that kind of stuff, and then there’s the knowledge engineers and enterprise knowledge management stuff. It’s always been really hard work, but parts of that are now more amenable to, if not automation, at least acceleration. Can you talk a little bit about how LLMs have kind of advanced the stuff you’ve always been doing and always wanted to do more of?
Juan:
Well, so LLMs, first of all, we look at the LLMs, and I like to look at things like as inputs and outputs. So the input is a prompt, and the output is going to be some text around this prompt. That’s at a super high level. But what’s been fantastic is that if you give it some context as an input, and part of that context, you also give it a task of what you want it to go do, the output that comes out actually aligns very nicely to it. Now, yes, there’s hallucinations and stuff, but look, if it gets 80%, 90% done, that’s better than nothing. It’s better than actually doing things completely manual. And guess what, if a human does it, they are not always getting things 100% correct anyways. So that’s at a super high level.
Juan:
So the example I give all the time is organizations always complain, “Oh, how many customers do we have? Oh, we have so many different customers. It’s so complicated.” Okay, so what? You’re just going to cross your hands and just live like that? No. How about you figure out at least how many different ones are there, and what are the implications of those things? But how would you do that? Well, you need to go start talking to people, and then somebody needs to go talk to all these people and start cataloging and combine all this stuff. Well, what if I actually just posed that question to somebody, saying, “Can you describe what is a customer for you in a sentence or in 10 seconds, in 15 seconds, right? Give me a minute rant of what is a customer for you?” Just take that, and you got a transcript, you got text right there.
Juan:
So now, I could prompt the LLM to say, “Hey, here’s this definition of a customer. Can you please give some definitions?” Here’s where the prompting, this is where the interesting experiments are going to come in. And then you can now do that for so many different other people. And then you can do further analysis to figure out, saying, “Hey, where are the overlaps of this stuff? Actually, you know what? Can you codify this? Can you turn this into code? Can you turn this into already some sort of a knowledge graph on it?” And I can now start comparing these things and realize, “Oh, look. There’s really X different definitions of this, where the alignment is on 80%, and the 20% of this.” Now we know these things. Now I can actually go back. That’s the type of stuff that we can actually go do very quickly in the initial experiments that I’ve been doing. It’s like, this stuff is working so freaking well.
Juan:
I wish we had this back 20, 30 years ago, but now it’s time to go look at the problems that we’re trying to go solve there. But back then and now we can actually solve them with these new tools. So that’s what I’m really excited about and this is something that I really hope, one, I want to show that this can be done, but second, but most importantly is the so what? Once we’ve done this, what is the value for an organization? So I think those are the ties that we need to go do still. It’s like we all complain about we don’t know the definition of the customer. So what is the problem for that? What’s the money that we’re leaving on the table? How much money are we losing by not having a complete definition of a customer? Now that we do have that, we need to be able to say, well, I do have that now. How is that improving? I think that’s really important for the next step.
Larry:
So we’re a couple of years into this now, where is the value emerging? You’ve talked a lot about the workflow in making knowledge graphs and capturing knowledge, but how does that percolate up to business value? Where are folks finding that payoff?
Juan:
You mean from knowledge graphs in general or…?
Larry:
Well, just from this whole scheme you’re describing, the ability to accelerate understanding of knowledge and what you just described is what I do. I’ve been doing content modeling for many years and it’s sort of very similar kind of work and a lot of that back and forth. I’ve already seen it accelerated in my content modeling practice as well. And so the way you’re talking about, I can see that and then so to me it just means I can model faster, but then the ultimate business goal is like you’re doing things with that model more quickly. Does that make sense? Is it?
Juan:
Yeah. So I would say that, look, the honest no BS is that we’ve been doing things with data and even though as much as we complain organizations are being successful, they’re making money, they’re growing. So you can argue it’s like, “This stuff is already working for me. Why do I need to go do more?” So that’s the first argument, which is a very valid one.
Juan:
Now I think, let’s look at the 80-20 rule. Let’s say, okay, yeah, you’re being able to be efficient and operational efficiency and you’re able to go solve your things, but at an 80% level, and then there’s that last 20%. Another question is it worth to go figure out that last 20%, that last 20% is, for example, I have different definitions of customers, so when I go ask for things about customers, I get an answer back. Is that the correct answer? Was that the… I don’t know, but hey, my business is working, so maybe that 80% was enough. So that’s the honest we need to acknowledge. So now the question that the last 20% is something that we’ve really not been able to go tackle because we just continue living in the status quo of today. So we live on making sure that current 80% today keeps working faster and efficiently and so forth.
Juan:
Yes, the problem is that we were complaining about 30 years ago, the same problem we complain about today, but we still have advanced on scalability. We have things in the cloud, we have things more secure. All that stuff has really made things more important. What I believe is that, that last 20% is I call the investment in semantics and knowledge. And my hypothesis is that that last 20% has always been expensive and hard. That’s why you have not untapped it. But the moment that we do untap that I believe that we are going to have a little bit more better competitive edge. We’re going to know better our customers because it’s not just one generic thing where, “Oh, so now if I know things better, I cannot make smarter, better decisions around that stuff.”
Juan:
So my hypothesis here is that if you untapped that last mile, that 20%, we’re going to be able to go do things that we were not able to do before. And I suspect that that’s going to help us grow more.
Juan:
Now, organizations who are kind of in their old ways, they’re going to be slow and clunky to start, and they’re going to have younger companies who are going to go get, who are AI native, who are starting from a clean slate, they’re going to be able to get to that a hundred percent faster and hey, they’re maybe going to come up and catch up.
Juan:
So I think that at the end of the day, my hypothesis, and my position here is that investing in semantics and knowledge enables us to reuse data and reusing of data means that we’re going to be able to go make sure that one plus one is greater than two. That’s the big thing right there. I think that is the economic argument that we need to go do. Anyways, let me throw it back to you.
Larry:
Yeah, well, no. One thing, I still haven’t sorted out my audience for this podcast yet, and I think some of this is going to be onboarding people who are new to the world of the semantic web and semantic practice in general and ontology design and engineering and all these practices. When you talk about how semantics might take us that last mile or the last 20%, talk about a little bit about what you’re getting with that semantic more, better understanding of the meaning. You’ve used the example of the classic one customers. What do you even mean by that? How can clarifying, for example, your notion of what a customer is take you a little further down the road in terms of business results?
Juan:
Yeah, I liked it. Let’s talk about what I call this a shift from the data first world to a knowledge first world. The depth I was give is I give you a table and the table has a column name, and has an integer. The column name is length and the values over there, you have a bunch of values underneath. Do you know that’s an integer? And then next to it has another column name called unit. Okay. So that’s all you have right there. Okay, what do you know? Well, that’s a number. It’s a natural number and blank must be measuring something, right? Okay, that’s data right there. Knowledge and adding all that context is to know like, “Oh, that length is actually the height of a warehouse.” And by the way, all warehouses have height. So if you have a null value there, that is an obviously error because all warehouses have height. So that value must exist. If we just don’t know where it is, we should go figure out what that is.
Juan:
More knowledge is that, hey, the height of a warehouse is always between 15 feet and 50 feet. So if you see a number there called 99, that is definitely a wrong value in there. And then you can add more. By the way, this is all semantics. These are all meanings. This is providing meanings, but what are the… Now, let’s add more context and business context around that. So what? What is the value of having the correct data about heights of a warehouse? So think about this as if you have a warehouse, you’re selling storage space. So what happens I have the incorrect storage space? Let’s say my warehouse, if I say it’s 40 feet, but it was actually 50 feet, oh, I have empty space. I could have sold more space, so I’m leaving money on the table, but what if I said it was 50 feet, but it’s actually 40 feet and I sold all that space. Oh, well, I don’t have a place to go put all these packages that I’m expecting right now, so I’m going to be losing money along this stuff.
Juan:
So every time you start increasing more of that semantics, adding that meaning, and in that context, that context is connected more to the business side, which makes then connected more to, at the end of the day, we live in a capitalist world, so how am I making more money and saving more money around that stuff?
Juan:
Now, everything I just said is already happening today. And my point is that we’re doing this in, I know cliche is cliche words, it’s not first class citizens. We’re doing all this work. All that knowledge work is already happening. It is embedded in people’s heads. It’s just embedded in a dashboard report. It is embedded in code, it’s embedded in transformation and ETL pipelines and all that stuff.
Juan:
And I think that’s what I call that 80% people are doing it. I’m like, look, I think if we start treating that knowledge, that context, that semantics, with the respect that deserves that is going to make us more efficient in so many places. One, we now know what the rules are because we can say, Hey, those rules changed because of regulations, whatever. So we need to update that. We now can onboard employees faster to tell them, look, you’re managing these warehouses. Now we know why it’s so important to keep the data quality of it’s so high, It’s so important, right? Employees are leaving organizations. We want to be able to… If they’re off board, I want to be able to extract that knowledge from them because that other knowledge people are going to be using, and where am I going to go stick back and connect that?
Juan:
So I think all of this at the end of the day is really to take an organization to the next level of what I think what it really should be doing. We should really… Your data and the knowledge is really that brain of your organization. And I think the vision here is that if you want it to be acquired, you want that company to survive the test of time to be around for the next decades, decades and centuries, you want that knowledge to be in a place where it can evolve and people can continue building upon it to make it better and expand on it. And I think you’ll be able to do that faster and grow your company better. I mean, that’s what organizations are, right? So I think we are doing it today, and I just like we can be more efficient and better at it. And that’s my point.
Larry:
As you’re talking about that, I’m reminded, I’m thinking back because I know you’re about as deeply enmeshed in the old semantic web and the original knowledge work stuff. I was thinking just now for example, of Doug Engelbart and his notion that all of this work we do should augment human intelligence and human effort and everything you just described is things that he was talking about in 1965 or whatever, and we’re just now getting to it. I guess as somebody who’s been around this a while, is that galling or is it encouraging that it’s finally happening?
Juan:
Well, if I zoom in, I obviously get frustrated. I’m like, “But why aren’t we doing this? Why don’t we do this?” But then if I zoom out, I realize, look, first of all, change is hard always, right? That that’s one thing to acknowledge. Second is that the pendulums come and go, right? So we were trying to go do this conversation we’re having right now, but talking about knowledge that people were doing this. This was the topic in the ’80s of doing expert systems. If you think about what the Japanese were doing in the ’80s, the Japanese fifth generation project was about creating these AI machines based on prologue and having parallel hardware to go do this stuff. They were trying to do that in the ’80s.
Juan:
Now we’re kind of doing this coming back to it also in a newer manner. And I think people are not realizing, well, it’s not just one or the other. We should find the hybrid around that stuff. But we needed to start from one side of the pendulum, move to the other side of the pendulum and figure out, okay, we probably need the best in both worlds, but we actually figure out what the best in both worlds are is because we’ve been to each side of the pendulums.
Juan:
This is why we just need to humanically learn these things. And it is frustrating, at least for me because I’m like, I’ve been seeing all these things. I’m connecting all these dots, but I’m realizing other people are not connecting it. It’s like, well, because we need the pendulum to swing on one side. For the pendulum to swing on one side, you need people focused on that side of their bubble in a way. And then when the pendulum goes back and forth, and then we’ll see more of these dots connected. And I think that’s now starting to happen. And if we look at it just progress in this area, I mean 50 years, 60 years is not that much time either. So we also have to have… I have faith in humanity. Things will come in and we’ll have smarter people always to kind of push things forward. So I think that’s what’s happening right now.
Larry:
And that faith, as you were talking about that, one of the things that occurred to me as you were just talking, I haven’t embarrassed to say I haven’t thought about this before, but so much of that stuff in the ’80s was about manufacturing and industrial applications of computing power. Whereas now, we’re in this almost 100% largely knowledge-based cognitive economy. You think that shift prepares us better to take advantage of, because from what you’ve said, a lot of this has kind of been known, but is it better or easier or more viable and feasible to apply it in the current era?
Juan:
I’d say, yeah. I mean, as time evolves with new technology, different problems become easier, then we have new problems around all the time. So I think, yeah, that’s what I’m saying. Now I can use these LLM to help me go do some knowledge acquisition work, which was something which was really incredibly hard to go do. It was completely unscalable, manual 20, 30 years ago, and now we can figure out how to go do this in a scalable way. And guess what? That’s going to lead to new problems, which we don’t even know what they are. So I mean, that’s the point.
Larry:
You remind me too, that you do so much research in this area. For example, the research that you and Dean did around improved question answering against the relational databases with knowledge. Aa number of people observed that that was the longest paper title ever, but I think it was necessary to convey that. Can you talk a little bit about that and maybe do some of the lessons from that research apply in what you were just talking about?
Juan:
Yeah, so when we look at how Large Language Models and ChatGPT came out, obviously everybody’s super excited about it. And from a data perspective, specifically from a structured data from databases, relational databases, people are like, “Oh, I can now do text to SQL.” And then big reminder is that doing question answering over structured data, text to SQL is something that people have been working on for over 30 years. So this isn’t a new thing either. Now, suddenly everybody’s like, “Well, we can just prompt this. It’ll generate SQL.” And then, yeah, you give it a very simple example, a very simple question. It works pretty well. So a cute question with cute data, cute schema, that works pretty well. And then people are starting to go write up all this stuff and getting excited. I’m like, “Okay, hold on. Let’s calm ourselves down and let’s go figure out what really is the state around it.”
Juan:
So the first question that we asked ourselves was, to what extent can these LLMs actually provide accurate answers over relational database, over enterprise relational databases? Because all the experiments and people writing blogs about this stuff are just about getting cute examples, cute data and stuff like that. So to what extent it was? Now our hypothesis was this is not going to be accurate at all because it doesn’t even know what this stuff means. So it means extra context, it needs that meaning, it needs that semantics. And then again, we’re like, “Well, our hypothesis here is that if you have a knowledge graph representation of your relational database, it is going to return more accurate results than if you don’t have that.”
Juan:
Now, an initial, not even experimentally initial test that we did was do these LLMs, and particularly we looking at ChatGPT at the beginning, it’s like, “Does it actually know these semantic web standards of ontologies of OWL, RDF and SPARQL?” I was very hesitant. And then I started playing around with this. So I played around with the… Dean playing around with Brian Jacob also played around with this, and we’re like, “Oh my God, this stuff knows it.”
Juan:
All these, it knows all the syntax. It’s generating everything correctly. I’m like, “What the heck?” And then I remember giving it like, “Hey, the first problem here is an OWL ontology copy and paste it, write a SPARQL query for the following question.” And I just did a bunch of just examples like this, like it was just coming out correct all the time. I’m like, “What the heck?” And I’m like, “Okay.” So this was kind of like, “Okay, now this thing has started to work. Let’s go test this out.” So that’s kind of the premise of how it all started. And then also we were talking to a lot of the folks at the cloud vendors like Snowflake and stuff, and they’re starting to put all these things out or like, “Oh, LLMs,” and they’re acknowledging that they need to go do some semantics and stuff, but they didn’t know what that looked like and actually how much would it actually improve?
Juan:
So that basically was like, well, we got all these unknowns. All these open questions, and we formulated the research question is to what extent can knowledge graphs increase the accuracy of LLM question answering? So we wanted to know the extent. It’s like if I start from zero and my goal is 100, if I don’t use knowledge graphs, where do I go? If I use knowledge graphs, how much does that improve? So that’s the question. And then the framework to go figure this out was we created a benchmark. We looked at an enterprise domain of insurance, enterprise schema in the insurance domain. We broke down questions into different quadrants, which was important, not just a laundry list of questions. And then, yeah, so we did a lot of experiments and that’s where we started to see the results from that. And then the results from that work was three times more accurate with very basic prompting, and that was the whole thing.
Juan:
We wanted to eliminate as many variables as possible. So the simplest prompt ever. Basically, here is your SQL schema, copy and paste, write a SQL query for the following question. Here is your OWL ontology, copy and paste, write a SPARQL query for the following question. That was it. And just with that basic prompt. And what was really cool is that this resonated with so many people and other vendors were starting to replicate the work. And I think obviously the vendors and the people in the space were very pro-semantics. They’re like, “Hey, look at this. We now have evidence around that.” So I know a fact that at all the large tech companies, our paper had been circulated to everybody around these things because it’s like people who’ve been trying to make the case for semantics said, “Okay, here is a paper that gives us quantifiable evidence of why you should go invest in it.”
Larry:
Yeah. I mean, that’s more than a year ago now, right?
Juan:
Yeah, that was actually year and three months. So we published, we put the paper online in November of 2023, but we actually started to talk about the initial results I think in September. So actually we decided to start doing the work after the Snowflake Summit in June 2023. And then we started doing the work, we started getting our first results after setting everything up around September. I think I gave a first talk around this at the Turing Institute in London back then, and then just that’s the work that came out. And then we did some subsequent work of how to further increase the accuracy by now leveraging the ontologies and that got said more to four times more accurate. Point is invest in semantics, invest in knowledge graphs for all this.
Juan:
And now you see that right now with graph RAG is a reason people are using graph RAG. And the whole knowledge, the graph RAG manifesto from Neo4j sites our work of like, “Hey, graph RAG is better than just only vector databases because we have the evidence that knowledge graphs improve the accuracy.” That’s just one of the reasons. So it’s great to now see that how it’s been a catalyst to people to say, “Yeah, we should start investing in semantics and knowledge.”
Larry:
Yeah, I hadn’t thought about the relationship between the arrival of graph RAG and your work. Do you think that was related? Do you think it was…?
Juan:
Yeah, for sure. No, it was related. I mean, you can see the graph RAG comes from Microsoft and Neo4j started pushing this, and I mean, you could see all the talks that even Emil Eifrem from the CEO of Neo4j talks about it and giving us somewhat of the credit of where it comes from. Yeah.
Larry:
Nice. Well, and that comes right back to the original question that I wanted to get you on the show about is how LLMs have accelerated all this, and you’ve given just tons of examples. Well, one other thing about your research, you intended that to be like benchmarked and benchmarkable. Has it become sort of the standard by which people are measuring the success?
Juan:
No, it hasn’t. It hasn’t. And I think the reason why it hasn’t specifically for the question answer is because there continues to be the very large benchmark systems like BIRD or Spider where they automatically test all these systems they go do, and that has become a de facto one. So when anybody’s doing text to SQL, they just continue to use those things. All those systems over there, what they do, from what I can perceive, and if I’m getting this wrong, I would appreciate people to kind of call me out on it, is they end up being training a lot of these models to go understand what those benchmarks are. And the point in our experiments was you had to invest in the semantics. There was work that needed to be done, it was not fully automated. So once I invested in that semantics, we should go do this. Now the accuracy is going to now increase.
Juan:
So it really is kind of a driver to go for the enterprise to go do all that core metadata catalog work that happens. And I think what I’m seeing when I talk to people, they hit that barrier. Yeah, you’re doing all this stuff without the semantics and you’re getting increasing, decreasing, but you’re hitting a wall. I’m like, yeah, you need to start treating your semantics as a first-class citizen and you aren’t. Then people realize that. Right? Now in the paper, the work, we were very specific on knowledge graphs and RDF and all at the end, put it, do whatever you want, but just treat it as a first-class citizen. Don’t just embed that semantics in some code or in the prompt or whatever. Treat it as a first-class citizen because at the end of the day, if you don’t, you’re just going to be living in eternal POC land and yeah, I guess cool if you create a great POC, but you want this to have an impact in the organization, you need to treat it as a first-class citizen.
Larry:
So that part of your hypothesis, I guess was born out that just add semantics and you’ll get better results. Is that taking off all the vendors at Semantics and Connected Data London? They all seem pretty busy. Is this world taken off a little bit?
Juan:
I would say so, and I hope so, and I think there’s a lot more to be done around this stuff. But again, remember is, so invest in the semantics. Now the question is, how costly is that investment? Which goes back to my initial point. It’s like I need to reduce that cost of that investment, therefore now, the next problem I’m focusing technically on is like, how do we make sure that we can now scale this out, make that cheaper, faster? That’s the next work that I’m doing.
Larry:
Nice. Yeah, I can’t wait to see what the future holds. Hey, Juan, I can’t believe it. We’re coming up close to time, but before we wrap up, is there anything last, anything you want to revisit from the conversation or just make sure we share before we wrap up?
Juan:
Well, I think the main thing I want folks to think about today listening is, it is time to invest in semantics and knowledge. And I want you to realize that you are already doing it, and every single conversation that you’re having with people, you’re talking about things, right? In every single code that your data engineers are writing pipelines, you are codifying that. In every single dashboard that you’re creating, somewhere you’re embedding the meaning of things. So you are already making that investment. Now, I think what is important is that to realize we should start treating all of that work with more organization, with more respect. It deserves more respect because this is basically the meaning of our business. So let’s acknowledge first that we’re already doing that work and it’s hard work. And the thing is that it’s being spread around and nobody has the mandate to manage that.
Juan:
And I think we just need to start realizing what we’re already doing that work, it’s hard work, it’s important work. Let’s start organizing that and figure out where it happens. I think where does it happen? I don’t know. I mean, every organization is different. I mean, is it a new role that we need to go do? Maybe if you look at the data engineers, there’s all this technical work that happens by the end of the end, there’s all this kind of social people work that’s happening. We probably need to go separate that type of work. Does that fall into a new role? Is that what I call it, the knowledge engineer, the knowledge scientists? Or is this work that should be done on data governance, data stewardship, this is what we need to go figure out? I think all that work is kind being spread around.
Juan:
So my takeaways, this is the time to start in figuring out how to go invest in semantic knowledge if we want to really go scale this out, and otherwise, we’ll just live in eternal POC land.
Larry:
It seems like there’s reason for more hope that we can get out of POC land and actually ensconce some of this knowledge and systems. But like you said, there’s still some details to be worked out there exactly where and how that happens. Hey, one very last thing, Juan, if people want to connect or follow you online, what’s the best place to find you?
Juan:
Oh, find me on LinkedIn. That’s easy. Just go follow me on LinkedIn and just shoot me an email too if you want. Juan, J-U-A-N at data.world. That’s it.
Larry:
Cool.
Juan:
Also, follow our podcast.
Larry:
Yeah, and I’ll include links to that. I got to say, the Catalog & Cocktails is awesome. I listened to every episode and I always enjoy it. The recent episode with Ethan Mollick really grabbed me. Anyhow, it’s fantastic. So I’ll include notes to all that stuff in the show notes as well.
Juan:
Thank you so much for listening.
Larry:
Yeah.
Juan:
All right. Thanks for having me.
