Podcast: Play in new window | Download
Subscribe: Apple Podcasts | Spotify | Amazon Music | Android | Youtube Music | RSS
Ontology engineering has its roots in the idea of ontology as defined by classical philosophers.
Casey Hart sees many other connections between professional ontology practice and the academic discipline of philosophy and shows how concepts like epistemology, metaphysics, and rhetoric are relevant to both knowledge graphs and AI technology in general.
We talked about:
- his work as a lead ontologist at Ford and as an ontology consultant
- his academic background in philosophy
- the variety of pathways into ontology practice
- the philosophical principles like metaphysics, epistemology, and logic that inform the practice of ontology
- his history with the the Cyc project and employment at Cycorp
- how he re-uses classes like “category” and similar concepts from upper ontologies like gist
- his definition of “AI” – including his assertion that we should use term to talk about a practice, not a particular technology
- his reminder that ontologies are models and like all models can oversimplify reality
Casey’s bio
Casey Hart is the lead ontologist for Ford, runs an ontology consultancy, and pilots a growing YouTube channel. He is enthusiastic about philosophy and ontology evangelism. After earning his PhD in philosophy from the University of Wisconsin-Madison (specializing in epistemology and the philosophy of science), he found himself in the private sector at Cycorp. Along his professional career, he has worked in several domains: healthcare, oil & gas, automotive, climate science, agriculture, and retail, among others. Casey believes strongly that ontology should be fun, accessible, resemble what is being modelled, and just as complex as it needs to be.
He lives in the Pacific Northwest with his wife and three daughters and a few farm animals.
Connect with Casey online
- ontologyexplained at gmail dot com
- Ontology Explained YouTube channel
Video
Here’s the video version of our conversation:
Podcast intro transcript
This is the Knowledge Graph Insights podcast, episode number 38. When the subject of philosophy comes up in relation to ontology practice, it’s typically cited as the origin of the term, and then the subject is dropped. Casey Hart sees many other connections between ontology practice and it its philosophical roots. In addition to logic as the foundation of OWL, he shows how philosophy concepts like epistemology, metaphysics, and rhetoric are relevant to both knowledge graphs and AI technology in general.
Interview transcript
Larry:
Hi, everyone. Welcome to episode number 38 of the Knowledge Graph Insights podcast. I am really delighted today to welcome to the show Casey Hart. Casey has a really cool YouTube channel on the philosophy behind ontology engineering and ontology practice. Casey is currently an ontologist at Ford, the motor car company. So welcome Casey, tell the folks a little bit more about what you’re up to these days.
Casey:
Hi. Thanks, Larry. I’m super excited to be here. I’ve listened to the podcast, and man, your intro sounds so smooth. I was like, “I wonder how many edits that takes.” No, you just fire them off, that’s beautiful.
Casey:
Yeah, so like you said, these days I’m the ontologist at Ford, so building out data models for sensor data and vehicle information, all those sorts of fun things. I am also working as a consultant. I’ve got a couple of different startup healthcare companies and some cybersecurity stuff, little things around the edge. I love evangelizing ontology, talking about it and thinking about it. And as you mentioned for the YouTube channel, that’s been my creative outlet. My background is in philosophy and I was interested in, I got my PhD in philosophy, I was going to teach it. You write lots of papers, those sorts of things, and I miss that to some extent getting out into industry, and that’s been my way back in to, all right, come up with an idea, try and distill it, think about objections, put it together, and so I’m really enjoying that lately.
Larry:
And I’m enjoying the video-
Casey:
Glad to be on the show.
Larry:
Yeah, no, I really appreciate what you’re doing there. One thing I wanted to, and I love that that’s how you’re getting back to both your philosophical roots, but also part of it is to evangelize ontology practice, which is that’s what this podcast is all about, democratizing and sharing practice. But I think, and I just love that you have this explicit and strong philosophical foundation and bent to how you talk about things. I think a lot of times that conversation is like, “Yeah, ontology comes out of philosophy,” and that’s the end of the conversation. But you’ve mentioned the role of metaphysics, epistemology, logic, all of which, can you talk a little bit about how those, beyond just I think a lot of people think about logic and OWL and all that stuff, but can you talk a little bit more about the role of metaphysics and epistemology and these other philosophical ideas?
Casey:
Yeah, definitely. You mentioned this in the pre-notes, “Here’s a topic we’d like to get to,” and I got into a lot of imposter syndrome on this, right? I’m trying to talk myself out of this, but I think most ontologists have this feeling there’s no solid easy pipeline into becoming an ontologist, right? It’s a very eclectic group of us. My background’s in philosophy, you run into a bunch of librarians, you’ve got computer scientists who do DB administration, you’ve got jazz musicians I’ve run into, it’s a weird group.
Casey:
I say that just to be, sometimes when I get asked about, “Okay, how does ontological practice work?” I think, well, I didn’t actually train to be an ontologist. I fell into it, so I’m ill-equipped to say things about what role ontology or philosophy plays in ontology.
Casey:
I just know I learned philosophy, and then I’m using some of those tools here, so there’s two different answers. One is historically, how does philosophy inform and shape the nature of ontology practice? And the other part is just, okay, if you’ve got a philosophical toolkit of metaphysics and epistemology and logic, how does that apply and make you a better, I mean, the obvious connection is that ontology is a philosophical term. It comes from metaphysics. We look back to Aristotle, and it’s the study of that which exists, so do we want to say there’s fundamentally fire, air, earth, water or something like that? Or fundamentally, there are these atoms and those are the sorts of things that are part of the inventory of reality. It’s not physics, it’s metaphysics. It’s the thing that in I think for Aristotle is just, it’s the book that sits next to his physics in all of his category, in his library of everything.
Casey:
But when we move that forward to computer science and data modeling, then we’re thinking, okay, maybe not for all of reality, although maybe it depends on how big you want your data model to be. But if I’m a retailer, what are the terms and ontology, what are the terms that I care about, the things that I need to model the constituents of reality that matter to me? That might be types, if you’re Amazon, it’s okay, medium-sized dry goods versus sporting equipment versus something else. If I’m doing a medical ontology, it’s patients and payers and providers, et cetera. In philosophy, in ontology, there’s a bunch of different tools and examples, but we think about, okay, what are some fundamental distinctions that we want to make? How can we carve nature at its joints in really sensible ways? That’s a phrase that you’ll hear a lot. We could say more about it if you want.
Casey:
But what I found is being a philosopher goes into an ontology space is that I have this inventory of examples from all of my grad seminars and various things that I’m looking through and going through whether I want to talk about gavagai and undetached rabbit parts, if that makes sense to anybody, or whether I want to talk about grue as a color, here are some examples, ways that we can chop up the world in unnatural ways versus chopping it up in natural ways and how do we make those distinctions? That applies straightforwardly when you get into building an ontology model for an oil and gas industry or something like that. There’s a bunch of ways that we can divvy up all the things you care about, what’s the right and sensible way to do it?
Casey:
I guess that’s the metaphysics, ontology way. Logic you mentioned, right? We need to think about reasoning. I don’t just want to assert a bunch of things about my data. A fundamental premise of an ontology is that we want to understand our data, we want to confer meaning on it, and that means that we have to be able to leverage the structure of the ontology to infer things smartly. Simple things like set containment are fine if all persons are animals, and then we say something about animals, they’re creatures. Then when I say that persons are a subclass of that, then I get for free that persons are spatio-temporal things as well. But we get a lot more complicated inferences as we go. We have to think about statistical reasoning. Just in general, if logic is the study of what makes for good arguments, what follows from what, that’s obviously got a lot of applications in ontology, AI.
Casey:
And then the third piece that we talked about is epistemology. Epistemology is the study of knowledge and belief, roughly about what it means to be justified. The classic example there is, if I know something, what exactly does that amount to? And then Plato says it’s justified true beliefs. And then the history of epistemology is littered with examples of trying to cash out exactly what does it mean to be justified. And if you get new information, how can that undercut your justifications? How do you update your beliefs?
Casey:
More recent stuff, and this is what I did in my dissertation, Bayesian epistemology, so doing probabilistic reasoning. Instead of just what do you believe, what degrees of belief do you have towards something? It might not be that I believe or disbelieve that a vaccine will prevent me from getting the flu this year, but maybe I have a credence of 0.2 that’ll prevent me from getting the flu this year. And then we get new information that says that there’s two flu strains out there that we didn’t expect, and so now my credence is maybe lowered from 0.2 to some other number. How do we address that? And when I’m building an ontology model, I should be thinking about this stuff to what are the things out there? What things am I asserting our true in my model? How do I think probabilistically about stuff? How do we represent probabilistic knowledge, which is super tricky in OWL. That maybe is too much into the weeds and would take us more than the 30 minutes just to talk about.
Larry:
No, but it also might be, and we can say this for later, it might be a pathway into the AI stuff, the probabilistic, I don’t know if there’s some intersection there of those ideas, but as you were just talking just now, you reminded me, I heard some, I think it might’ve been Walid Saba years ago say that one of the problems with the semantic web is that it was Boolean, and your research was on Bayesian, and then you’ve just talked just now about probabilistic and degrees of belief and stuff. Is epistemology the thing that gives us the tool to deal with fuzziness in ontology practice?
Casey:
I mean, part of what they study in epistemology does get to that. You can maybe smooth out the transition. So yeah, if triples fundamentally feel more Boolean, right? You just say subject, predicate, object, and you’re just asserting that thing to be true. But we can pack probabilities into that kind of structure too, where we can use things, if we want to say RDF star or something like that, we can say of this statement that it has a degree of probability 0.7. You can also use data type properties, we can use lists. There’s ways to inject numbers into this.
Casey:
The same problem comes up in epistemology and Bayesian epistemology is discussion. We call them credences, right? So if I say I have a degree of belief of 0.7, that’s a credence, or at least that’s the way that’s talked about in the literature. And there are some people that say, “Ah, that’s lame. That’s just more ways of talking about belief.”
Casey:
So to say that I believe that it’s going to rain tomorrow, okay, there’s a Boolean sentence. It is just either I believe or I don’t believe it, or maybe withhold belief, maybe it’s three values. If I say that I have a credence that it’s going to rain tomorrow at 0.7, maybe that’s just the same thing as saying, “I believe there’s a 70% chance of rain tomorrow,” and now we’re just back into a straightforward belief claim and we don’t need to make any hay about talking about credences. You can probably do a lot of the same things when we’re, you can put numbers into this more Boolean feeling, OWL RDF model. Yeah, it’s something I have to think about more.
Casey:
I would say in general, just getting some of the Boolean stuff to work pretty well is hard and valuable enough that that’s usually my first pass is like, “Let’s get that stuff in there.”
Casey:
And then if we want to get more nuanced about saying, “Okay, now I have X degree of confidence in this source that that statement came from,” so I’m going to interpret it as we can do competing arguments. There were some really fun things that we did back at Cycorp where it was just adding metadata to each assertion. So if you have the provenance of each claim, then you state all these claims in maybe a more or less Boolean way. But when you want to weigh them against each other, some of them might be more forceful than another. You just imagine it’s like you have two experts that are telling you contradictory things, and then you have to decide which expert do I think is more reliable on this topic.
Larry:
That’s where it gets really fun. Was the mechanism for that annotations about the assertion about the likelihood of its?
Casey:
Yeah, essentially. I mean, at Cycorp it wasn’t in OWL RDF, it was in Cyc, which is a Lispy-derived version and so it wasn’t quite the same. But yeah, essentially using annotation properties or something like that.
Larry:
If you could talk just a little bit about the Cyc project because I think a lot of people newer to this may not be aware of it. I think it’s pretty important and foundational. In fact, Doug Lenat was post posthumously awarded an award at KGC last year. Can you talk just a little bit about the Cyc project and what you did at Cycorp?
Casey:
Yeah, sure. Without Doug Lenat and Cycorp, I would not be an ontologist, that’s for sure. I mean, my history with that was back when I was getting my master’s in Wyoming, I remember talking to one of my mentor professors there and just in passing, we were like, “All right, I’m looking forward to this. I’m good at this. I’m going to be a professor someday. Heck, what else would you do with a philosophy degree?”
Casey:
And he was like, “I had this one guy friend of mine long ago who showed me his business card after he got a job out of grad school and it said ontological engineer,” and we laughed about that, what a silly job title that would be. Some foreshadowing for my life.
Casey:
And then when I got on the job market for philosophy, Cyc had a posting because Cycorp wanted philosophy PhDs. They saw this as a market inefficiency. These are smart people who can think symbolically and put stuff down. Maybe he also had an understanding that humanities grads would take way less money than people who are already in the tech field, which is a little money ball of the situation. And the Cyc project was all about, so it was formalizing and symbolizing things, but it started with, I mean I think of it as it’s big differentiators like common-sense reasoning where the thought was, okay, great, there’s a whole bunch of things we want to do with AI, and that’s wonderful. Machine learning, that’s going to be super cool, that’s going to do some great stuff. But fundamentally, to think about the world, you need to know basic facts about how reality is structured. Bigger things don’t fit inside of smaller things. Heavier objects are harder to carry than lighter objects. When I say the bee landed on the flower, I need to have a model of reality that helps me interpret what that sentence means.
Casey:
And what we need to do is go in and hand encode more or less. I mean we can automate some of it, but all of these facts about reality, and we need to generate a bunch of what we call common-sense tests. Philosophers are really good at thinking through this because they can think of basic sentences, think of exceptions to those sentences. Did a lot of really foundational novel, important things. I can’t do the history full justice, but if you have folks who are interested in this at all, looking into the Cyc project I think is really cool. Doug has some good material out there as well if you just search Doug Lenat. He had one with, I’m going to forget his name now, the MIT podcaster guide, Lex Fridman I think [here’s the recording].
Larry:
Oh.
Casey:
If you want to get an overview, that’s classic Doug and gives you a lot of the important things, but that introduced me to ontology as a whole and thinking about not just doing like the domain picture, but that our knowledge is reused all over the place.
Casey:
One of the pushbacks I find in OWL RDF that maybe sets me apart from some other ontologists that have spent most of their time in OWL RDF, I’m not that interested in only building a domain model that looks at say, biology or something like that. Because I know the knowledge that we have about biology borrows from a lot of more general knowledge that’s reused. The fact that bigger things don’t fit into smaller things is reused in all sorts of domains. You can think of it almost as the upper ontology versus domain ontology but, Cycorp was the upper ontology for common-sense.
Larry:
Yeah, and that’s a common concern of any ontologist is whether to adopt an existing ontology or to create one unique to your domain. But you just said, and you talk, this sounds like a foundational thing in your practice is that you do, well, you mentioned for example, you’ve adopted Dave McComb’s idea of categories whenever you build an ontology, and I guess I don’t know if you’ve wholesale adopt the just ontology or just that practice out of that. Can you talk a little bit about that? Is it just reusing upper ontologies or are there practices and principles and concepts that you pluck from these other existing artifacts?
Casey:
Yeah, I mean, my relationship with gist is, that was the first upper ontology in OWL RDF that I learned, and so it was a safe space or something that I went back to. The metaphor I use a lot is, and I encourage anybody else who’s doing this, don’t reinvent the wheel. If there’s an ontology out there that does what you want, kick the tires on a little bit, but you should absolutely use it. What will probably happen and what happened with me and gist is there are a lot of things I found wanting about gist, I’m not trying to start a fight with them or something like that, but I viewed it as, I took that house as a fixer upper, right? It was like, all right, I’m going to take that ontology, and then I’m going to start knocking down walls and moving things around and redoing the plumbing and stuff like that. But there’s certain fixed points. Maybe I like the house’s footprints and then expand it.
Casey:
That was my starting point. And then it now somewhat, the ontology that I use somewhat resembles gist. There are a few terms that I have in there that I’m like, “All right, this is basically exactly what they meant by the similar term in gist.” I think gist is more, it’s designed to be flexible a little bit and fit onto a number of different projects, and it’s really important for Dave and the company to keep their ontology relatively light. He has this thought that if you have more than a certain number of classes or terms, then the human brain just can’t grok it anymore, and so then it won’t be used and effective. I disagree with that general philosophy. I think that it should be way denser and more robust, but I think if we have the right hierarchical layering of things, then the top level classes should be relatively sparse. We don’t want too many of them, but it can fan out and get really fine-grained underneath that.
Casey:
I have a much larger version of gist, I guess. But then when you’re talking about categories in specific, right? I just did a video on this so it’s fresh on my mind, categories, I don’t love them metaphysically. It’s not like, I think this is an important way to view the world. Categories are almost more important because of the restrictions that you have in OWL. That’s a bad term of phrase, not OWL restrictions, but the limitations of an OWL framework, right? It’s fundamentally, it’s a two-leveled system where we’ve got individuals in classes and that’s great if we’re talking about, I’ve got dishes or something like that, and my particular coffee cup is an instance of a dish or is an instance of mug, that’s great, but there are lots of places where we want to use more than two levels.
Casey:
I feel like periodicals are really maybe a little more tricky on this. We’ve got New York Times as a newspaper, might be the collection of all New York Times newspapers, and then we have, I might be tempted to say that, and then today’s edition of the paper might be an individual that’s an instance of that. But then I have my own copy of today’s paper, and that feels like an instance of today’s paper, which is an instance of a New York Times. Wait, now I can’t have that many levels of instances, and now we have a puzzle where it’s like, I’ve got two levels, but I want to put stuff in three levels or more, and you have two choices. Either we can go keep the individuals the individuals, have classes and then have metaclasses. If you’re thinking of sets, sets that contain sets, or we can mush stuff down and say that we have individuals and we have some other individuals that categorize those individuals.
Casey:
And that’s the category route is that instead of if I want to talk about, I think the W3C has an article on this or whatever, if people search it up for punning and you get that eagles are an endangered species and you want to say, so eagle is an instance of endangered species, but eagle feels like the class of all the eagles. And so now I’m in this difficulty, you could make eagle a category, which is I’m just going to have a instance level version of eagle, and then I can use object properties, because the other driving force here is that in OWL RDF, object properties are what we want to use to describe and deal with our data. And so there’s this impetus to make the things that I want to talk about individuals rather than making them classes.
Casey:
Myself, I’m cool leaving them as classes and just having annotation properties talk about things, but then I know that I’m, in some sense, I’m misusing it because annotation properties aren’t supposed to be substantive, they’re supposed to be just metadata about how I formed a class or something like that. It’s not supposed to give me substantive modeling claims.
Larry:
That’s so interesting because in any one domain, you’re just going to have a bunch of stuff, and then getting it down to, and I think Dave would say, “Well, let’s get it down to a manageable number of classes.”
Larry:
And then this sounds like the way you handle it, it’s like your individual practice decisions about how to do this. I think that decision you talked about in one of your videos to model an event as a subclass of a situation, I thought that was a really interesting model. Is that in that same category as what you were just talking about categories? Yeah.
Casey:
Yeah. Man, there’s so many, all these grouping words like category and class and object and thing and stuff like that, when you become an ontologist, they in some sense, they lose all their meaning or you have to reorient yourself every time because you’re like, “Oh, that’s not the thing we were talking about.”
Casey:
They’re like, “You mean OWL thing? You mean the most general?” You’re like, “No, it was just a random throwaway noun. That’s what I wanted.”
Casey:
Comes to thinking about events and situations, so categories, like I said, it’s not super philosophical or metaphysical, maybe I think of them as labels or something, but it’s a useful device in the OWL framework. When I think about events, I’m much more metaphysically committed to them. Events and situations I think are real parts of the world. There are things I see out there, there are things I want to describe. I want my model to match up with that.
Casey:
This is another piece that ties into my history at Cycorp, so the approach there for modeling events and situations was a Davidsonian approach. I feel like I maybe even just talk about, this is our Davidsonian model, and I don’t say anything else about that. And some of my YouTube videos, which may be confusing, but named after Donald Davidson, philosopher who has his own account for how we should think about events and things happening about particulars and hanging off of each other. We were talking a little bit before this call. It’s like, okay, I remember that and I need to go back and read the paper to really say any more meaningful things, so apologies to Donald Davidson lovers out there that I’m not giving a good account of that.
Casey:
But essentially the way that you can think about in events, the event itself is the Christmas tree, and then everything else gets hung off of it. We’re doing a podcast interview here, and then there are a bunch of role players within that podcast interview event. You can hang me off of it as the interviewee, and we can hang Larry off of it as the interviewer. And then we’ve got a particular time that this is happening and place, or multiple places since it’s occurring online in both Europe and the United States. And so you’ve got this fundamental central object that then all of the other stuff is attached to. In an OWL RDF, those role-playing predicates are, I have a super predicate that’s like situation involves, and then I have sub-properties of that that attach the various sorts of roles that you might have something like doneBy requires relating an agent to the events that is the doer of the events. And maybe we have more specific versions of doneBy, completedBy or startedBy or et cetera.
Casey:
And not only do I pretty much like the metaphysical view of events as this subclass of situation that’s a Christmas tree that everything hangs off of it, it’s very amenable to graph structure. OWL RDF, one of the puzzles, if I want to say that Casey does podcast. They’re like, “Oh, great, there’s my triple: Casey does podcast.”
Casey:
And you’re like, “Wait, but Larry does, wait. All right. Podcast is the events. That’s the thing that we’re doing. And then all the things get related to it.” That’s what it’s pushed towards.
Casey:
But one of the first approaches that some people do to events and activities is they say, “Casey kicks ball,” or whatever as my triple, and I want to say when that happens, now how do I do that? What do I attach the time to? I can’t attach it to the property because that’s just not the way OWL RDF works, it’s not an LPG. Then I have to do this RDF star or something or a reification. And essentially you can think of what’s going on here as a reification have to think more about what-
Larry:
You know what I’ll do-
Casey:
They might be isomorphic, but they’re pretty close.
Larry:
I’ll link to your video where you talk about “the block,” LeBron James’s famous block and with a 2016, I think NBA finals because that was a great example of that, but you said a minute ago the word agent and immediately I was triggered. It’s like, “Oh yeah, AI, we got to talk about AI.” But also, that’s also how I first found you. I can’t remember how I came, I think somebody shared it on LinkedIn, but you did a post about what is AI? And then you were like, “I don’t know,” but then you did elaborate and you have a lot of good ideas about what it might be. But I’m wondering, that was maybe a year ago or something. I’m just wondering how your thinking on AI has evolved and especially in relation to the relationship between the neural side, all the LLM, GPTs, conversational, gen AI kind of stuff, and then our knowledge representation stuff. Where’s your thinking at these days?
Casey:
Yeah, some spoilers for upcoming videos. Yeah, I’ve got the next couple there. I feel like I know what I want to say my answer to what AI is now, but I’m always someone, maybe this is the analytic philosopher in me. When we’re starting something, I ought to be able to define it or say what I’m talking about first. My approach to defining AI, well, we can start with some examples of it. And this is something else that I borrow from methodology that we did in philosophy a lot. In epistemology, we know what it is to know something or at least we talk about knowing stuff. And so the first thing I do is, okay, say some stuff that we know. I know I’m holding this cup, I know I’m in this room, et cetera. And then we try and come up with a theory of what knowledge is such that it explains, makes sense of those things.
Casey:
Maybe we have to throw out some examples, but that’s a good basis for us and for AI, I’m going to start with, okay, well, what are some things that we’re commonly calling AI. ChatGPT is AI, maybe this camera like autofocusing on me to some extent, is AI, a number of things like that. Okay, what’s the definition of it that works? Let’s look historically, and the more that I’ve pushed on it, the more that I think it’s wrong to try and define something as AI. It’s okay. It’s a fine pass for us to say, I’m going to call this thing AI. That’s clearly the way we talk and I’m not going to force us to change that. But I think it’s probably better to think of AI as the discipline. It’s the thing that we’re studying, so the science of artificial intelligence is trying to create non-biological minds or something like that.
Casey:
And then the scientists who do that, the people who research that end up building a bunch of stuff, and the byproducts of that research, we can call AI colloquially, but I don’t think of those things as being AI themselves. If you think of the parallel to this for science, science is very hard to define. It’s a classic, another philosophy of science, the classic demarcation problem. How do you describe what’s science versus pseudoscience? It feels intuitive to everyone until you push to some difficult cases and then you realize, well, for a long time they thought chronology was science and shaking an eight-ball or reading tea leaves was science, now we don’t want to call that. How do you do it? How do you not? What’s the line of demarcation? If you find yourself thinking, “Oh, it’s obvious,” please come, solve the problem. Philosophers would love for you to solve that problem, it’s harder than you think.
Casey:
But if we talk about science, so people who study a specific type of science, whether we want to say it like a material science or physics or whatever, we don’t say the products that could produce by that are science. Undoubtedly, someone studying material science came in and helped produce whatever polymer goes together to make this toy that I’m holding. For those of you on the podcast, I’m holding a little squishy forklift. I wouldn’t say, but see this I’m holding, this is science. No, we’d say the study of the sciences enable us to build this sort of thing. That’s how I feel about what people are calling AI. Is ChatGPT AI? No, I mean you can say it if you want, but those who were interested in creating non-human minds produced a thing and it was this, and it was a tool that did these various sorts of things.
Casey:
I think that’s important, one, because we should be clear about our concepts, and I think if you’ll push anybody on what they mean by AI, you’ll find out they’re not very clear about their concepts. It almost always comes down, I think to AI is something that either could be confused for a human, or that could do something that a human might otherwise do. Well, look, I’ve got a door over there, and a human could stand in front of it and keep people from getting in. And before the invention of deadbolts, people would do that. And now we’ve got a deadbolt that does that. I don’t want to call that deadbolt AI, but it’s doing a thing that an intelligent human otherwise might do.
Casey:
We can come up with these cases that feel grotesque. They’re not really AI, but maybe at some point in time somebody will, like, why is this door closed and not opening? Even though I’m pushing on it, there’s some artificial person holding it closed for me? They could have said that until we figured out how deadbolts worked and they were no longer so magical. And I think the same thing’s going to happen with LLMs and other things along the way. A thermostat that senses the temperature and regulates the temperature to your house feels super-smart, but now people are not probably calling that AI so much, we’re not so mystified by it. And then eventually we’ll be like, “Yeah, what LLMs are doing, they are taking a bunch of information, munging it together, and then spitting it back out at me.” Hopefully in smart kinds of ways, the algorithms have primed to be really helpful to me, but that’s just another deadbolt or paperweight or something like that in my view.
Larry:
Nice. I wish we could have you moderating all of the hype coming out of Silicon Valley, have a philosopher approved version of it.
Casey:
It’s so frustrating, man, because you’re just like, okay, the people who are giving the hype are the people who are profiting a ton off of it, which is, this is again, logic and epistemology and rhetoric 101, don’t trust those people. Maybe some of them are right, and some of them are reliable, but they have problematic biases and conflicts of interest.
Casey:
And then the other part that’s really tough is, and I know you want to talk somewhat about talking to business folks, that’s part of the target audience here, but you set ridiculous expectations that AI will do these magical things, and it puts those of us, hopefully I’m in this camp that’s not over-hyping the industry, but thinks it does really cool things, puts you in a bind because you don’t want to go to a potential client and be like, “Hey, this product, it’s not as good as you think.”
Casey:
That’s not a compelling sales pitch, but at the same time, you can’t be like, “Hey, look at this product. It’s better than anything you’ve heard.” You’re like, “Nah, I can’t cash that check.” I mean, I’ll cash their checks until they realize I can’t cash the metaphorical check of getting the product to do the things. And so it’s not magic, but it can do some awesome stuff. You’ve got to find that balance and it’s really true.
Larry:
Yeah, and I think we’re all looking for that.
Casey:
If there is an AI winter, another one, it’s going to come from this ridiculous over-hype of LLMs and what we can expect them to do and then the layoffs associated with it, and then we’re going to have to backfill for those people if those products don’t do what we need them to do. It’s going to be interesting.
Larry:
It is, and you’re reminding me there’s a number of reasons we need to check back in in a year or so. Yeah, we’re running close on time, so I want to wrap things up, but you just mentioned rhetoric, another aspect of philosophy that’s so relevant to the conversational nature of AI. There’s probably all kinds of stuff there. The business stuff, we can talk about that more later, but right now I want to make sure before we wrap up, is there anything last you want to make sure we mention or anything you want to revisit from the conversation before we wrap up today?
Casey:
Man, we’ve hit a bunch of different things, I’ve rambled a lot. Apologies, I guess that’s just my style. I hope folks get from this, that there are a bunch of interesting philosophical tie-ins to ontology, and that if we think clearly about things and try and build our models in ways that reflect the way the world works with the constraints that are given to us by OWL RDF, you can do some cool stuff. There’s a lot of further conversation I want to have with you, but not a specific thing that jumps in here. I guess one thing that came up a few times in my head as we were going over that I should mention, don’t forget that the thing that we’re building is a model. And when I describe a model, a model is a simplification of reality.
Casey:
One objection I could hear someone having when I talked about, “Well, you’re restricted when you build stuff in OWL, because you’ve got two levels.” You’re like, “Okay, well that’s an impoverished model and that’s a problem. We need the more robust one.”
Casey:
And that was definitely part of the Cyc project is that it wanted put everything in it as many intricacies as you could, and that does some really cool stuff. But just like whenever you’re building a model, it has to be a simplification. We have the real thing, it’s the world, it’s out there. Our model has to be okay, and here are some relevant details we want to abstract away from that, that we can reuse. The job of an ontologist is to pick what’s the right level of abstraction for that so that it’s manageable and good and workable, but that we didn’t lose too much information because we lose too much information that we can’t prove the things that we need to be able to prove.
Casey:
And if you’re going back to that sort of, the debate we had with Dave McComb who’s not here about how many terms to have in the ontology, I think that’s the right lens to view that from if you’re trying to figure out, okay, maybe a more robust model that has got more terms in it is just worse because it’s less usable by people. You’re painting the steering wheel in a little toy car where the goal of the car was to generate wind resistance, and then who cares what color the steering wheel is, that doesn’t factor into it. So make sure the details you’re providing are the details that you want to use.
Larry:
That reminds me, I do a lot of talks about modeling, and I always include the George Box quote: “All models are wrong, but some models are useful.”
Casey:
Yes.
Larry:
Which is way too true. Hey, Casey, one very last thing. If folks want to connect with you or follow you, what’s the best place to find you online?
Casey:
Yeah, so the best place to find me I think is just probably on LinkedIn. You can look me up on my YouTube channel. This is a thing that I’m trying to figure out. Maybe I’ll send you a link later if there’s something to put in the description. But I’ve had a few people reach out to me from the YouTube channel via LinkedIn, and it’s annoying because you’ve got to go through, get a pro account to message somebody, so I need to get an anonymous email account or something, I’m not going to put my cell phone number up. I got to figure out the right answer. But yeah, go look up Ontology Explained on YouTube, find me there. Yeah.
Larry:
Okay. I’ll put your LinkedIn profile and your YouTube channel in there so folks can find your ultimate email solution there, I’m sure. Well, thanks so much, Casey. This was a really fun conversation.
Casey:
Pleasure is mine. Yeah, hopefully we can do another one sometime. Great to meet you.
