Podcast: Play in new window | Download
Subscribe: Apple Podcasts | Spotify | Amazon Music | Android | Youtube Music | RSS
Hybrid AI architectures get more complex every day. For Jans Aasman, large language models and generative AI are just the newest additions to his toolkit.
Jans has been building advanced hybrid AI systems for more than 15 years, using knowledge graphs, symbolic logic, and machine learning – and now LLMs and gen AI – to build advanced AI systems for Fortune 500 companies.
We talked about:
- his knowledge graph and neuro-symbolic work as the CEO of Franz
- the crucial role of a visionary knowledge graph champion in KG adoption in enterprises
- the two types of KG champions he has encountered: the magic-seeking, forward-looking technologist and the more pragmatic IT leader trying to better organize their operation
- the AI architectural patterns and themes he has seen emerge over the past 25 years: logic, reasoning, event-based KGs, machine learning, and of course gen AI and LLMs
- how gen AI lets him do things he couldn’t have imagined five years ago
- the enduring importance of enterprise taxonomies, especially in RAG architectures which business entities need to be understood to answer complex business questions
- his approach to neuro-symbolic AI, seeing it as a “fluid interplay between a knowledge graph, symbolic logic, machine learning, and generative AI”
- the power of “magic predicates”
- a common combination of AI technologies and human interactions that can improve medical diagnosis and care decisions
- his strong belief in keeping humans in the loop in AI systems
- his observation that technology and business leaders seeing the need for “a symbolic approach next to generative AI”
- his take on the development of reasoning capabilities of LLMs
- how the code-generation capabilities of LLMs are more beneficial to senior programmers and may even impede the work of less experiences coders
Jans’ bio
Jans Aasman is a Ph.D. psychologist and expert in Cognitive Science – as well as CEO of Franz Inc., an early innovator in Artificial Intelligence and provider of Knowledge Graph Solutions based on AllegroGraph. As both a scientist and CEO, Dr. Aasman continues to break ground in the areas of Artificial Intelligence and Knowledge Graphs as he works hand-in-hand with numerous Fortune 500 organizations as well as government entities worldwide.
Connect with Jans online
- email: ja at franz dot com
Video
Here’s the video version of our conversation:
Podcast intro transcript
This is the Knowledge Graph Insights podcast, episode number 20. The mix of technologies in hybrid artificial intelligence systems just keeps getting more interesting. This might seem like a new phenomenon, but long before our LinkedIn feeds were clogged with posts about retrieval augmented generation and neuro-symbolic architectures, Jans Aasman was building AI systems that combined knowledge graphs, symbolic logic, and machine learning. Large language models and generative AI are just the newest technologies in his AI toolkit.
Interview transcript
Larry:
Hi, everyone. Welcome to episode number 20 of the Knowledge Graph Insights podcast. I am really delighted today to welcome to the show Jans Aasmann. Jans is, he originally started out as a psychologist and he got into cognitive science. For the past 20 years, he’s run a company called Franz, where he’s the CEO doing neuro-symbolic AI, so welcome, Jans. Tell the folks a little bit more about what you’re doing these days.
Jans:
We help companies build knowledge graphs, but with the special angle that we now offer neuro-symbolic AI so that we, in a very fluid way, mix traditional symbolic logic and the traditional machine learning with the new generative AI. We do this in every possible combination that you could think of.
Larry:
Who?
Jans:
These applications might be in healthcare or in call centers or in publishing. It’s many, many, many different domains it supplies.
Larry:
Is it mostly large enterprises or is there a certain scale at which this stuff works better?
Jans:
Our customers are always Fortune 500, Fortune 100 companies. It’s all the companies that are trying to do innovation. Most big enterprises now believe that knowledge graphs is in their future. They’re experimenting with it. They do experiments and we help them build their first knowledge graphs in most cases. Once they get that going, they can do it on their own.
Larry:
Interesting. It’s often their first knowledge graph. Where in the organizations is this typically pioneered? As you come into an organization, is it IT or is it data science? Where are you typically entering the organization?
Jans:
That’s actually the wrong question.
Larry:
Okay.
Jans:
The thing is, all the places where we’re successful, there’s a champion, a person that’s really looking into the future and has this vision that it’s possible to not build a new silo for every new problem, but there should be a way to integrate all the knowledge in the organization into something incredibly useful with that. You can’t leave it to a single programmer or a single architect. It’s usually someone with some business experience and also some architectural role that believes in this approach. If you leave it up to an IT department, it’s just not another database, but it’s not the philosophy of a knowledge graph of integrating knowledge. It’s just, okay, this is a problem I can solve with the graph. Let’s do it that way.
Jans:
You need a person that says, “Hey, I’ve got so many different sources of information in my organization. I know we’re not combining it in the right way, and it’s too complex to put in a relational database. I know we have to solve it with a thing that they now call knowledge graphs, but it’s even more than that. I know partly it’s data science. It’s machine learning. Partly it’s rule-based. It’s the logic, symbolic logic. I also know that I need the new generative AI in this, but how do I do this all? This is incredibly complex. I can see the future, but how do I do it?” That’s where we come in, help build them knowledge graphs, but with a symbolic angle.
Larry:
I love that, and that knowledge graphs are a key element in the architecture of the future and of the present, it sounds like. That champion who comes to you, it sounds like they’re somebody who’s been aware of the hazards and consequences of siloed information and data. Is that typically what they’re coming in for, of how can we better integrate and understand all of this?
Jans:
Again, I would say there’s two types of champions. One of them is just, they want magic. They see all the articles about generative AI. They see the Gartner articles, Forrester articles about knowledge graphs, and they think, “I have the feeling that something can be done with knowledge graphs and neuro-symbolic AI.” Those people are not very super technical. They usually have a technical background but then went into business, but they know something can be done.
Jans:
Then you have the second type of champion, of people that are literally always over 35, that have spent their active life building application after application. Every time when they created a beautiful application where their bosses were really happy about, the sad thing is, they build a new silo and they made their whole enterprise even more complicated. These are the people that say, “You know what? There has to be a better way of integrating my knowledge.” Those are the people that get interested in semantic technology and they say, “There has to be a way that we don’t build new silos every time.” It’s this thing that we call data-centric computing. Data comes first and applications need to go on top, but we don’t need to change the data all the time, rewrite and copy the data all the time. That’s the disappointed IT person that says, “There has to be a better way.” Does it make sense?
Larry:
Yes, that makes sense.
Jans:
One is really looking at the future like, “Wow, my company needs magic to make more revenues. The other one is, “Hey, we need to reorganize our IT house, because this is madness the way we do it now.”
Larry:
Yeah. They don’t want to do that, repeating the same mistake over and over again.
Jans:
No, let me try. How do I turn my sound off? Okay. Yeah.
Larry:
Cool. Yes. You said they’re always older, because these folks have been around the block a few times who come in with that. I’m going to guess. Are they generally? It sounds like both. It sounds like maybe in the first type, the magic seekers, you’re probably doing a lot of education, but with the second type, you’re maybe just more finessing the implementation. Is that?
Jans:
No, it’s a huge difference. The magic seeker will give you freedom to think together, “Okay, how are we going to do this? What is the first baby step we can do to show the rest of the company that this works?” You try to find the low-hanging fruit where you can show how neuro-symbolic AI and knowledge graphs can help and do things they couldn’t do before, whereas the second type of champion, the disappointed IT person that says, “We have to find a new way,” then it’s way more IT-oriented. Let’s find three databases where we want to do something extra with. Let’s build a semantic data catalog. Let’s build an ontology of the objects that we really care about in our business. Then let’s see how we can replace an existing system by something that is 10 times more simple, 10 times more easy to understand, and 10 times easier to do data science with. Does that make sense? That second champion is very more IT-oriented and wants to make the flow within the company better, and the first one just wants more revenue by magic.
Larry:
Right.
Jans:
Huge difference. Huge difference in the two champions.
Larry:
Is it pretty much an even mix between those, or do you seek out one or the other more?
Jans:
No, you can’t cold-call a company and try to find a champion. It doesn’t work that way. They have to find you. They have to go to conferences to find you and you have to tell the stories. What is neuro-symbolic AI? What is the use of knowledge graphs? Then you give them example stories of what you did for other companies and then they say, “Hey, I have something like that. Can you do that for me?” That is basically how I think all of us in the world of knowledge graphs work, right? You give examples. They look at it and said, “Hey, that applies to me probably.”
Larry:
Yeah. That’s interesting. In either case, you’re going in and you focus on neuro-symbolic AI, but a lot of the work, we talked last week in preparation for this, and it sounds like a lot of the work you do, there’s almost always a mix of logical neuro-symbolic stuff, machine learning, LLMs. Is that bespoke and driven by the needs of each organization, or are there patterns that emerge in that?
Jans:
It’s the latter. It’s the patterns that emerged over time, right? When we started out 17 years ago with the AllegroGraph, it was a knowledge graph, a graph database. The word knowledge graph wasn’t even invented yet and we had logic. It was declarative logic. That was important that instead of writing programs, you had declarative logic, but then we got these things like OWL and reasoning systems on top of OWL. That worked sometimes great, but these description logics that we worked with were really bad with time and place. All the customers I had had something that was temporal and was located in areas, so we added a full prolog compiler to our AllegroGraph, our pre-knowledge graph. Now we had a symbolic logic system, but then 10 years ago, I think, actually even more, my first articles about temporal and spatial knowledge were, I’d written about 11 years ago already, because every customer did something with things that change over time, and location was often a big part of it.
Jans:
I got into this event-based version of knowledge graph. Of course, what’s the most important thing in an event-based knowledge graph? You want to predict what is the next thing going to happen. What is my customer going to decide? What’s going to happen to this patient? When is this plane going to have a problem with a particular part of the equipment, and is it in my inventory? Prediction became important, so we started working with machine learning. It was a completely natural evolution. We started doing stuff with events in time and place. We wanted to do prediction, so now we needed to become experts in machine learning, so that was great. Then about two years ago, our world exploded literally for all the people in our community, because now we suddenly had these generative AI and LLMs.
Jans:
Now, by the way, the knowledge graph people, for the last 10 years have been working with unstructured text and using taxonomies to find the important word in your domain, and then use these taxonomies to extract the entities out of the unstructured text and then put those also in the graph, so now you can use logic and even machine learning to work with that. It still was very, very painfully slow to build these taxonomies and to work with them. Then two years ago, as I just said, our world exploded and now we have generative AI. By the way, we still need taxonomies all over the place, but now the generative AI can do things that I never even dreamt about. For example, in the old days, I could say, “I’m looking at a medical text and I can find the word aneurysm.”
Jans:
I can find that, but it’s fairly useful to just have the term, right? I want to know, “Okay, does the patient have it now or was it a little bit ago, or was it in the history? Did the doctor actually say that he denied an aneurysm or was it actually there? In what stage is the aneurysm? Is it treated? If it’s treated, what is it treated with?” We became experts in just annotating terms out of documents. Instead of just getting a term out, we can say 20 things about how that term fits into the pattern of that particular paper. Now then obviously, sorry, in this text. Then of course, it became very important that we can also put things in a vector store. We can start doing RAG, but not just dumb RAG where you dump everything in a big bag of vector-indexed text, but that you actually can do a mix and match of symbolic AI to pre-select where you want to do your RAG on.
Jans:
They call that graph RAG. That’s yet another thing you might’ve heard about, but that also became part of the mix. Now after, say, 17 years working in this domain, now we have the symbolic logic, the first autologic, because we have the full Prolog. We have machine learning, primarily recurrent neural networks for event prediction. We have generative AI for entity extraction in RAG and even graph RAG, so it’s blooming, and the whole industry is going in this direction. It’s just a pattern that’s emerging.
Jans:
By the way, we call this mix neuro-symbolic AI, because, I did my thesis in modeling car driver behavior. I’m a psychologist. I tried to model a car driver. I wrote about 2000 rules to model the behavior of car drivers, and I got amazing results when I compared my model with real human drivers, but I was lamenting in my PhD thesis that we really need neural networks to do object recognition and do pattern recognition, because just doing the rules wasn’t good enough. It was too complicated, and it just couldn’t do a lot of things. Now it’s 35 years ago that I finished my PhD thesis, and now I wish I could go back, because now I could do anything I ever wanted in my model of a car driver. Does that make sense? I’m going off-topic here.
Larry:
That speaks to not the rapidity, but the depth of the evolution of the tooling here, that you could, but even back then, you could see, you could imagine some kind of technology that could help you accelerate that, but it just didn’t exist and you couldn’t build it. It’s just the way it is. Was it frustrating, or was it just a relief when it finally showed up?
Jans:
It was more like, “Oh, this is what I always wanted.” Does that make sense? This is what I always, yeah.
Larry:
Right, because you didn’t even know.
Jans:
No, no.
Larry:
Yeah, no.
Jans:
Hey, I’ve been doing NLP for more than 30 years. In my previous jobs, NLP was always one part of it. Now I still, every day, can’t believe what we’ve gotten into with the generative AI parts.
Larry:
Yeah. Are there still, I don’t know, old-fashioned, but conventional NLP, the understanding and generation stuff in these LLMs, or have the GPTs and LLMs just replaced all that?
Jans:
Oh, no, absolutely not. There’s many facets to a generative AI. You can ask questions of a text. It will do really well if you have a concrete text, but you still want to do entity extraction, extract the concepts out of a text and link them to your company-determined taxonomy. It’s still important that you have a list of all the things you care about in your company and then take texts and relate it to the terms you find important, because doing RAG on a billion documents without pre-selection of what you actually want to look at gives you very, very bad results. You still need to pre-process texts. Just imagine you have a billion documents, sorry, a billion emails between people in your organization. At some point, you say, “Okay, find me all meetings that were canceled due to COVID between this group, this project group, and that project group.”
Jans:
You have a billion emails and you have that question. You can use RAG and talk to the email and you will get a really bad answer. The chance that you get a right answer is almost nil, because what you really first have to do, “Okay, what does it mean that you’re part of this group or that you’re part of that group?” Let’s only select all the emails from these two groups. Then let’s look at the time period where this happens. Now you have to do a time selection on the data, and then you can start doing, maybe first, another selection of the fact that there was a meeting request or something about a meeting. Then you can start doing RAG about which meetings were canceled due to COVID. Now you get a really good result, because you did a lot of work with symbolic logic just to pre-select the domain in where you want to look, because otherwise it just, how do Americans say, that’s a shotgun approach? You hit something, but you don’t know if you hit the right thing.
Larry:
Yeah, that’s really… the way you just said that, that’s super interesting. It gets into use cases for just AI in general, but then how you would tackle that particular use case. How has that evolved? It sounds like that particular use case might not have changed that much with LLMs. Maybe there’s some?
Jans:
No, but it makes a big difference, because now when I do my RAG, I only look at the texts that contain any word that resembles meeting that were between someone from Group A to Group B. Then I do my RAG within that subdomain and that works, but also, I have a database with about half a million clinical trials. I say, “Give me the negative effects of ibuprofen if you are already using aspirin.” There’s 40, 50 years of clinical trials. It’s nonsense to ask this question, because there’s too many articles to look at, but if you really, really, really can do a pre-selection on ibuprofen-related drugs, NSAIDs and aspirin-related products, and then find a much smaller subset, then the RAG suddenly becomes very viable and useful.
Larry:
It’s more… Yeah, that’s a really good example of that kind of architecture where you do that machine language pre-process. Is that ontologically driven or is that? You mentioned that everything-
Jans:
It’s taxonomy. There’s an ontology of how we represent the clinical trials, with the predicates that I use for clinical trials. Then there’s a taxonomy component, because we have UMLS with more than several million concepts that it can link to. I still use the ontology to find the right part of the clinical trials. I still use the taxonomy to get to terms that I want to relate to. Then ultimately I use the LLM to fine-tune and find actual answers or give answers to questions based on a very small subdomain. Yeah, we’re getting in the weeds now of one little part called graph RAG, right?
Larry:
Yeah. That’s funny. It comes up all the time. It’s not, yeah.
Jans:
For me, the neuro-symbolic AI is just about the very fluid interplay between a knowledge graph, symbolic logic, machine learning, and generative AI, but the use cases are almost infinite and it’s really fluid. For example, we now have in our query language these things we call magic predicates, where we basically, instead of looking at a database, you apply functions. One of our functions is to talk to our vector store. For example, I want to match the term aspirin with other drugs in UMLS. I now can say aspirin matches, and then I get a list of other medications and medication dosages. That’s part of a query, but AllegroGraph can also use natural language to create queries for you, so you can say, “Okay, give me all the patients that used both ibuprofen and something familiar to aspirin,” so it will write a query for you, but part of the query is that calls this magic predicate again.
Jans:
Can you imagine? You have a system that writes queries that uses itself to do subqueries, so it calls itself within the query, but that magic predicate also uses SPARQL to do preselection, so basically it’s like turtles all the way down. You go from a, anyway, it’s very different ties. That’s just completely normal for us now to do this. We don’t even think about it anymore, but this is an example. Another example is, we always, say I have medical databases and I can do a prediction based on logic. I can say, “Someone will get atrial fibrillation if,” and then I have a whole series of conditions that you need to have, and then it’s very likely you’ll get AFib. We have enough data that we can also build machine learning models to, given a sequence of events of a patient, I can say, “Hey, this person is going to get AFib in the next three months with a likelihood of say 70%.”
Jans:
A generative AI has read at least 36 million PubMed articles plus 100 million other articles about healthcare and patient care, so it also has an opinion. If I give a sequence of events of a patient to an LLM, it will tell me whether or not a patient is going to get Afib. It’ll even give me an estimate. Now I’m there, and I have patients in my database with sequences of events. I can mix and match. Each three of these AIs can tell me something about the likelihood, and then the question, who’s right? Which one? We even ask doctors to look at the sequence of events and what do you think? Now neuro-symbolic AI is also that you have all these different predictions. We’ve found that if it’s a rare disease, LLM is way better at predicting what’s going to happen, whereas for standard diseases, machine learning is way, way better, but symbolic logic is not doing bad for very well-known diseases. It’s a very interesting research domain. I have these three Ais. They all can predict, but when is what AI better? Does it make sense?
Larry:
It makes perfect sense. I’m wondering about the mix of just empirical study against that, yep, you were right; nope, that was wrong. This mix works better for this kind of condition or kind of pathology, but then I’m wondering also-
Jans:
The interesting thing is, you can just test it in the data. You have all the data available, so at any point in time for a patient, you can make three predictions and then you look what happened after that, so you can benchmark your own models.
Larry:
Oh, right, because you have all the data, the full course of the patient history and their treatment and outcomes and all that. Okay, got it. You don’t need to do that. To what extent are physicians and doctors and researchers in the loop on that, or is it just the fact that these AI tools are looking at their reports? Is that how they’re involved, or is there human oversight at any point?
Jans:
I’m a strong believer of a human in the loop. We have a very nice example of restaurant reviews, where we analyze restaurant reviews over time. We look whether or not the rating for the restaurant goes up and down, and can I find the reasons why the ratings go up and down? If I make a mistake with my LLM to get, say, some insight out of the text, who cares in the universe, because I’m looking at thousands of reviews. If I get a few wrong, who cares, but if it’s about your life, the prediction, you’d better put a human in the loop. A big hospital that we worked with had a model to predict whether or not a patient needs to be intubated. They had this beautiful model that could do the prediction with machine learning, but then the FDA would never approve of letting the machine decide whether or not you get a tube.
Jans:
What you do is you give the doctors a choice to say, “Okay.” The doctor could say, “Hey, it’s not my patient. Okay, let someone else decide,” or “I disagree with the AI,” or “Yes, I’ll put an order in.” They had a whole list of questions so that the human can make the decision. There’s many, many examples where we need a human in the loop, for example, with complex knowledge graphs. We need a human in the loop to teach it how to write queries for us. Say you build a complex knowledge graph. A developer would go around in the company and say, “Hey, what are all the things you want to ask of our knowledge graph?” then he starts writing. It probably takes two weeks doing example queries. Once you get enough example queries, you can train an LLM and you can do some intelligent prompting. Then it gets 99% of the queries right.
Jans:
When you just ask natural language what you want from the knowledge graph, it will find it for you. It will write a query for you. Every time when I look at it, I just can’t believe how good it’s gotten, but it’s never 100% trustable. You still need an intelligent human looking at it.
Larry:
Is it its ability to write the queries? Is that its main utility in that architecture, or is it because it’s looking at so many examples? Like you said that before, though, that it depends on…
Jans:
Oh, no, no. Okay, let’s see how old-fashioned symbolic logic works together with generative AI. We have a complex knowledge graph. The first thing we do is we ask AllegroGraph, build a SHACL image of your knowledge graph. Instead of a human being writing SHACL code to do the, SHACL is an alternative to building ontology. It’s a schema-based language that you can use to check the fluidity of your data, but you can also reverse engineer it. Basically, we don’t write SHACL anymore. We let the machine write the SHACL, and then the human being looks at the SHACL and says, “Do I agree with the SHACL?” Then you say, yes. Now you have fixed that part. You let the machine write the SHACL.
Jans:
Then when you ask the machine to write queries for you in the prompt, you say, “Okay, here’s the SHACL of my database.” Now it knows the structure. “Here are five examples that look very much like the thing you’re asking. By the way, here are some of the magic predicates that you might need for this query.” We do this in multiple steps, but the point being, it’s not that you just ask an LLM that doesn’t know your knowledge graph to create a query. No, you have to help it in every possible way. Part of it is, you give it the SHACL of your database. You give it successful examples, and then it writes queries. If you like the query it wrote and the results you got, you store it as a successful example, so you teach the LLM. The symbolic logic, the human and the generative AI work together to become expert in writing queries automatically.
Larry:
Yeah.
Jans:
That’s just one of the million ways you can combine these technologies.
Larry:
Yeah. We’re all figuring that out as we go.
Jans:
Yes, yes, yes. For us, it’s now second nature.
Larry:
Yeah. That’s it. Actually, let me jump back to your old study of human behavior and cognitive science days and how these architectures and the decisions about when to have humans in the loop, I don’t want to talk about AGI as a thing, but how truly helpful to humans and how close to human reasoning are we getting with some of these? If you really nail the architecture of some of these things you’ve talked about, it sounds like you can get pretty close to being a good diagnostician or call center predictor. How much are you hearkening back to your old behavioral science and cognitive science days?
Jans:
Have you already interviewed Gary Marcus?
Larry:
Not yet. He’s on my list, for sure.
Jans:
Yeah, you’d better call. He has a lot to say about reasoning. Of course, I’m talking to scientists. All I thought about when I was doing that work was thinking about reasoning. If even Bill Gates says that, “Hey, you need to combine generative AI with symbolic logic,” and if the CEO of NVIDIA is hiring knowledge graph engineers to combine LLMs with knowledge graphs, everyone in the industry now are getting to the point where you say, “Hey, just using generative AI isn’t enough. You need a structured companion next to the generative AI.”
Jans:
In the beginning, I got so many customers that, “We don’t need knowledge graphs anymore. We can do it all with the gen AI,” but that phase lasted only a few months, I think, and then that was over again. Then for a while people said, “At least we don’t need taxonomies anymore,” but we still need taxonomies. I see now the cycle is turning, and people realize that I need a symbolic approach next to my generative AI if I want to store the results, if I want to preselect. There’s so many ways where the two need to work together.
Jans:
Now, I didn’t answer your question about reasoning. I think that we’re getting very close to having really good reasoners in many subdomains, but they need to be treated, but it’s going to be this agentic approach. It’s going to be multi-step. If you looked at o1-preview and its reasoning, you see 20, 30 sub-questions it’s asking just to make sure it covers the whole thing and then still can get it wrong. Yesterday, I tried to get o1-preview to write a prolog program to do something complicated. After about 20 minutes, I gave up because it just couldn’t do it and I, as a human being could do it. The point was, I was even thinking, “Should I write an article about this?? I gave it every help I could, but when it tries to write, say, Prolog programs, to me, it feels like it doesn’t really understand prolog. From the inside-
Larry:
Is it just because there are many fewer examples of that than Python or JavaScript?
Jans:
Yeah, but I’m reading also a lot of literature about how people use Python in production. Instead of getting 10 times better, you get maybe 10% better. Then yesterday, I actually read a great article about that if you are an experienced Python programmer, it can be an incredible productivity tool to use it, but if you are a beginner Python programmer, then you might accept Python programs that are just plain wrong and you don’t get it. You still need to be a really good Python programmer, and then it’s a fantastic help. If you are a beginner, you just don’t learn the language like you should. At least that was the opinion of the author, and I really, really liked his. I can send it to you later, that article.
Larry:
Yeah, I would love to include that in the show notes. That sounds fascinating. Yeah, there’s some kind of matching going on there with level of how much you can get out of these tools depends on how much you’ve got in your head already.
Jans:
Yes.
Larry:
Is that accurate? Yeah.
Jans:
Absolutely. Yeah.
Larry:
That’s good. Hey, Jans, I can’t believe it. We’re coming up close to time, but before we wrap up, is there anything you want to revisit from the conversation or just make sure we share before we close?
Jans:
Apart from saying, if you have anything, I make a plug for myself. If there’s anything where you need a combination of symbolic logic, machine learning, and gen AI, come play with us.
Larry:
Yeah.
Jans:
That’s all I can say.
Larry:
Yeah, my intent in doing these is to advance practice, but some people are already advanced, as to what we were just talking about. You’ve got more of this in your head, so yeah, that’s perfectly legit. Hey, and if folks want to connect with you or follow you online, what’s the best place to find you?
Jans:
Well, sent me email at J-A, my initials at Franz, F-R-A-N-Z, .com.
Larry:
Great. I’ll put that in the show notes as well.
Jans:
That’s great. Thanks.
Larry:
Yeah. Thank you so much, Jans. I really enjoyed the conversation.
Jans:
Okay, it was fun to do, and I hope to talk to you again. All right.
