Jonathan Lustgarten: [00:00:00] There's so many new things, so many products coming out there, so much innovation and technology space that I think 2025 and in a couple of years are going to be huge, but make sure that when you see something, you understand how you use it.
Megan Sprinkle: Welcome to Vet Life Reimagined. I have a returning guest today, Dr. Jonathan Lustgarten. His first podcast episode was two and a half years ago. And you know what is probably one of the episodes that I mentioned the most because his unique career path in veterinary informatics really intrigued me. And if you haven't heard that episode, please go check it out.
I'll link it in the description. Much has changed in two and a half years regarding technology and AI. So I wanted to see how it has changed his career and the opportunities for other veterinary professionals. Jonathan discusses the rapid evolution of AI in veterinary medicine from his perspective, as a director at Mars Veterinary Health, you will hear about the explosion of AI [00:01:00] tools from diagnostic systems to AI scribes that are reshaping the field.
You'll hear about the intersection of AI and veterinary informatics and understand the broader Implications for veterinary professionals, client communication, and career opportunities. So let's get to the conversation with Dr. Jonathan Lustgarten.
I would love to start with just kind of the past two and a half years. How has your career changed? Are you anywhere else? So what's, what's new with you?
Jonathan Lustgarten: Well, so I'm still at Mars Veterinary Health, I actually was promoted to director of AI and machine learning. So that was pretty exciting. But I think the real impact is that the explosion of AI in veterinary medicine, two and a half years ago, we were talking about like, you know, there are some results.
There are some tools out there, you know, people are doing diagnostic AI, stuff like that. But in the past two and a half years, it's been, I don't want to call insanity because it's not really crazy, but like the volume and number of [00:02:00] tools that either incorporate AI had built AI for specific or are looking to, you know, integrate with AI systems.
It's just astronomical. Like I was talking to someone earlier today about, two years ago, I went to VMX and there was like one or two, you know, maybe three or four at most. I think that like we're AI forward this year, just in scribes, AI scribes. There were 17, like, you know, what, five fold, six fold increase, right?
So it's, it's one of those things where just for one facet, let alone all the other places we're talking about, we're going to ease of booking. We have communication, we have prompting, we have, all these different areas, you know, different companies like Antech Rapid Read.
Now IDEXX has it, the integration of digital pathology with automatic. AI is now like. Almost a, "you don't have AI?" As a question of saying like, Oh, you do have AI. You know, it's one of those things where it's like, Oh, I, where is it going? So I think that has been a [00:03:00] huge change and obviously it's affected me personally because people are now, you know, a lot of people are saying, Oh, instead of saying, Hey, can we have AI, can we use tools up like, Oh, How does that work?
People are now saying, Hey, I'm thinking about doing this. Is this an AI project? Fun part is that, you know, a lot of times it's like, no, AI is not the solution in that point, right? Cause I think it's the, and this follows human medicine, you know, new tool, shiny, shiny toy. but it's knowing when not to, but I'd actually, it's been a massive shift.
And now when I talk with clinicians because it's such a mass consumer thing now, it's like, there's not as much, I don't want to say reluctance, but there's not as much confusion but there's now not as much saying like, Oh, well, Why do we need AI? It's more like where can we use it? So I think that's been a huge thing beyond the position. It's just overall, you're seeing new, like veterinary conferences pop up.
You know, I give talk at VIS that was about usage of AI and how to implement it. That was not a conversation I [00:04:00] was having five years ago, which is awesome. I mean, I tried to have that conversation in some places. You get it. But, you know, Association for Veterinary Informatics, you know, we talked about that last time.
, you know, more and more papers are about use of AI, responsible use of AI, so there's all these places and that's two and a half within the past two and a half years. Not that long ago. I mean, both of us have had, you know, changes as well in our life, but it's that, like, just the sheer number and volume and the rapid adoption.
two and a half years ago when we were talking, you say, where's it going? Like, I'm hoping it will increase now. It's like, how much did we increase? So it's been pretty awesome.
Megan Sprinkle: Yeah. so I think it might be helpful to, so when we were in our last episode and what you do, we talked about veterinary informatics and. I think it might be interesting to hear your thoughts about, is that the same thing as AI? Is there, is it different? how do you help people understand, you know, what [00:05:00] you are already working on?
And then this buzzword of AI, that's now like taking over our lives.
Jonathan Lustgarten: Yeah, and before you remember, it was machine learning. You don't say AI, you know, the whole general
general.
Megan Sprinkle: scary.
Jonathan Lustgarten: It's scary, right? you know, 2001 Space Odyssey, you know, how right? So then informatics uses AI uses machine learning to either analyze information or feedback information to people for use. Right. So I consider informatics kind of like the super set, you know, and there's people who designed purely AI, right? So it's like one of these giant Venn diagrams is actually in 2009, 2008, there was a article I published on it in like one of that, one health, magazines that show the overlap, right.
You know, where it's medicine and it's computer science, things like that. So informatics is kind of like. The bridge between, computer scientists, data scientists, even those things, these all are kind of blending together. But informatics is really understanding the data and then trying to find the right tool and sometimes building [00:06:00] your own tool, right?
Building your own AI algorithm or machine learning algorithm interchangeable in my head, at least, to do it. The only exception is generative AI. So to me, there's Machine learning AI interchangeable, but generative AI tends not to be because that tends to be more creation, right? You can use it to do things, right?
You know, you can use it to diagnosis and standard traditional algorithms, but one of its strongest thing is generating text. That's a whole subset that is still informatics, but also requires some focus, because let's face it, it's such a new technology. What we use today, like with the 03 it's all in one area. But to me, veterinary informatics is trying to translate veterinary practice and information. Into extracting value information or value items from it and then sending it back to the user to make informed decisions.
So I use AI tools and I build some of them, but to me, it's kind of like, some people used to get very nervous about it. Now people are like, oh, it's an accepted [00:07:00] part. So, AI is kind of, again, like a tool, your stethoscope is a tool, your CBC machine in your clinic is a tool, now there's, you know, your urinalysis clinic is using machine learning as a tool in the clinic for, you know, it's like, it's all, it's all intersecting, but to me, AI is very much a tool to use, and the veterinary informatician, the person does it, is knowing how do I use it in my scenario in veterinary medicine, it's to give information to find valuable information to give back to whoever your customer is.
Megan Sprinkle: Yes. And I'm thinking back to something that Adam Little said, and it, when it comes to AI scribes, he said that was. The most rapidly adopted tool that he has seen in technology and the veterinary industry, and he gave several reasons. And one of them was that it was something that the veterinarian could just decide to do on their own. They didn't have to wait for it to be adopted by the hospital , and it clearly solved. A pain [00:08:00] point for them. So it was an easy. Okay. I want to go home early when you're thinking about informatics and how the information is delivered to the veterinarian or the veterinary team. Can you describe how they interact with that? Are you seeing a faster adoption with that? Can we improve in any way? How is that interaction with the team?
Jonathan Lustgarten: Yeah, so the biggest thing with veterinary informatics, it used to be PowerPoint presentation, niche papers, things like that. I coming from Pitt Biomedical Informatics, like one of the main core courses is, user interface design. And I think user adoption is a critical aspect because it's like if I handed you, you know, machine code.
Yeah. That gives you how to treat something. I'm giving you the information, but you can't absorb it, right? It's a that's a that thing. I think what has happened is with things like chat GPT is that they abstracted some of the more technical to make it easier to use, but [00:09:00] that comes with its own challenges.
So I think people are looking for it. But 1 of the things even with scribe, there's still time to shift. So these, these scribe machines, all these algorithms, they're not veterinary specific. You know, some people are working on it that that will change. That's, that's a short term thing.
It's not like just for right now, at least they're not really built for veterinary medicine. The scribe apps, cause they tend to use these large language models out there. So the problem is that when you're shifting it, right? why did they just go use the scribe tool? Cause maybe the system they were using was very inefficient in typing.
Okay. Maybe their system didn't allow them to enter as much data. It's like a stream of consciousness can work really well, but it has its downside. You know, it's like people's memory. Like for me, I have a good auditory memory, but I know I will memorize it once I write it out. Right. And same people, some people when they're typing out in a medical record, they'll think of it.
So there's shifts there. And then, so you've [00:10:00] shifted the. You know, maybe the hard part is getting it in, but what about the review? These systems make errors, like 30%, 30 percent errors, you know, one in 10 major medical errors. These are things that we're finding today, even with more advanced thing. Now you can do a lot of it, reduce it, but there's still errors, right?
It misses ideas, it hallucinates, Is the veterinarian stopping and reviewing what was written? And I said, for simple things, you know, particularly for the simple stuff, like the simple things that wellness exam, I think these algorithms do amazing.
I think when it gets more complicated, you have to treat it with caution. but what happens is that now they're focusing on time value, right? Before, I have to enter all this information and type, right? No one says, you know what I love to do?
I love to enter text in fields. Like, no one says that. That's not said. But, when they have manually typed it in, they're not necessarily reviewing it because they know what they said. When they do these text things and conversion, they should [00:11:00] go through all those manual checks, right? Because these machines are imperfect.
They have high error rates. They're not for us. They're not trained on us. you know, dogs are not humans with a tail, right? All the type of fun stuff that we used to say, and so, but their time value, even if they spend more time, and they've shown that in human medicine, that even voice dictation will, you know, they may not save as much time as they think, but the time value that they spend on correcting Is more acceptable to the user, that is where I'm seeing a lot of shift in acceptance because people say, I don't have to type it saying, oh, I can record it and then just quickly edit. You can quickly edit is longer So my favorite thing is, have you ever tried use your voice to text on like any of your Siri or Google or whatever, my Siri just went off.
I've had some very entertaining, corrections to my words and I'm like, Oh, backspace, backspace. And I wind up backspacing it when I realized I could just have typed it out the message faster, but it's okay. Cause you know, when it gets it right, it's how many times you have to correct it and can I catch it when it gets [00:12:00] it wrong?
And because the technology has become so easy to use that now people say, Oh, well, I just opened the site, start dictating and it does a great job. So even if I have to spend more time, I may not have to spend more time as often. And so I found that particularly in the informatics field, the language models have allowed people to start saying, Huh, maybe I can present this to users in a way that they'd accept that the AI generated from them.
That's the other thing is that now people expect it to be right, right or wrong. People expect it to be right. And that was something before it was like every AI algorithm is wrong. I'm the doctor now that we're seeing it as less and less threatening. In some cases, I think there's more acceptance. So an informatician that proposes AI to be introduced into workflow may not have as much resistance because, oh yeah, no, I use AI at my house.
Those type of things are now becoming commonplace that people now accept it and so they can focus on, [00:13:00] well, the results are wrong or I can improve it as opposed to why we're using AI for this.
And so that's been a huge paradigm shift. I think that's echoing all communities from software design to informatics to everywhere.
Megan Sprinkle: Yeah, so, you're right, and I think the speed of which it is improving is also interesting as well. It used to be wrong a lot, and it's getting wrong a little less and a little less. At least when I'm using it for certain things, but you're right, it still can get wrong. And it sounds very convincing when it's wrong.
At least, At least when I'm using like chat GPT and even Claude. I love Claude too, and so it is something that we need to watch for. So that's a good call out. Are there other opportunities that you see that we can improve or how you're foreseeing this getting more prevalent for, again, I'm going to stay with the veterinary team and then I want to ask you about the pet owner too, [00:14:00] but for the veterinary team, what are you kind of foreseeing this year and maybe the next year?
Jonathan Lustgarten: So I think our AI is actually gonna hit a wall. We're gonna hit some sort of asymptote, mainly because the data that we are collecting we have is not as robust and available as a lot of the other data in the area. So we're we're going to improve, improve, improve. And then , we're already seeing it the rate of actual true improvement is more of how do we use the systems in a more intelligent way, as opposed to here's a brand new system. So there's a lot of that. I think for the system, the real, the real next step is that there's going to be lots of tools. How do I use it? When do I use it? Who's responsible for its use? And how do I ensure that, we're not hurting our pets?
first off, people are asking for these tools. talking with a lot of veterinarians, you know, at VMX and other areas, I see people say, well, I'm actually not sure how I did it. So they're going to try, right. And a lot of these companies, they can't do implementation at every single [00:15:00] hospital to make sure that you're using it.
It's just, it's, you know, there's too much variability But I think what's going to be next for the veterinary team is like, okay. It's no longer, is there a tool is what tool should we use? How do we evaluate it? And then how do we implement it?
Cause like I, as you said, when it gets it wrong, it can be, I call it hysterically wrong because it's like, it comes up with stuff like, wait, what? I'm not even talking about that. Just brings in some other random facts I think having our team set up and understanding it and making sure we're not burdening the team is going to be a huge fact because all these tools, they You know, speed up some areas, slow down some others, but the net benefit of time spent means that we have to figure out, okay, are we actually saving time?
Are we, you know, sometimes when we implement new processes, it's our technical staff that actually wind up being impacted. Are we making their job harder? Right? You know, some [00:16:00] people say, Oh, well, I can do it. Or what if the technician starts dictating? Into the record, right? You know, during the history, like, you know, people talk about ambient listeners, which I think a lot of people really like unless you're next to room with me because my voice reverberates through the walls.
And these are not the most soundproof walls that we have. So, like, you'll wind up being recorded on someone else's conversation. so there's a lot of things to solve. But. You know how you do it and we do I think that's what the vet team is going to have to really focus on and that's going to be really hard because human medicine hasn't figured this out and they have a multi multi multi multi multi billion dollar companies helping them solve it, I was talking to some deans over the last week veterinarians aren't are not trained, you know, I wouldn't say all MDs are trained, but you know the AMA says that medical MDs need to be trained.
And DOs at least exposed to the informatics, and that's basically what this is. It's when technology, how it integrates, how the data comes in and out, [00:17:00] just at least a cursory understanding. Some schools have implemented more than others. I went to PITT, so obviously more integrated, not that I'm biased towards PITT at all.
But I think the veterinary industry as a whole, we're not being taught this. I'm not being taught how to, you know, evaluate this technology. We're not being taught of what questions to ask. Now, luckily we have incredibly intelligent individuals across the spectrum, so people are figuring out, but that means every single person has to figure it out.
That's different than when I'm coming into veterinary medicine and having a baseline understanding, like when I use this technology. I need to make sure I understand how to measure success.
So I think that's going to be a very interesting trend over the next couple of years because there are people like me who kind of combine two different degrees to figure it out. But all these veterinarians who are really trying, they're going to, they're going to be technology centric, but not necessarily understanding technology outcomes or how to evaluate those outcomes.
And I think that's where over the next Even the next few years, [00:18:00] we're going to see that pop up a lot, you know, responsible use, you hear that word. What does responsible use mean? Does it mean that I know when I should use it? Does it mean that when I use it, I know why it gave me the answer. All these questions are going to need to be figured out.
And as you said from, you know, Adam, I love listening and talk, uh, but like from a lot of these speakers, it is the question of What does it mean to use it? you know, so that means we're already behind the times, right?
You know, one person is going to take a decade to, I'm like, yes, but if we never start, it's always going to take more. Right. So I think that's going to be an interesting push because these tools are going to be out there. They're going to be everywhere. They're going to be integrated into almost everything.
And then I'm going to be like, wait, why did it do that? my favorite is when you search Google, right. Been around since. Forever. 20 plus years, right? When they get a search result that you type in, you're like, wait, why did it get that? Right? Because the search algorithm did blah, blah, blah, blah.
So that's great when I'm searching. But if if I'm trying to document [00:19:00] on a patient and it got it wrong, can I catch it and do I understand how I got it wrong so I don't do it again later? These are things I think not just the veterinarians, but the veterinary staff,
uh, all the associates, all the clinics, all these even larger companies, everyone's got to figure this out and, you know, AVMA introduced a, new, committee, you kind of a task force towards it. But these are very hard questions to ask because there are no regulations, you know, a lot of these things are not being regulated like a medical device.
I'm not advocating that it should. I'm saying now, but now we have to figure out how are these people understand how to do it so they don't hurt . We're going to see that type of dynamic play out. That's what we saw in the early thousands in human medicine, the same dynamic, new tools that are going to be used.
We may not understand how to use it all because a lot of people weren't trained on it. So then what happens? And so I mean, it's already happening, but it's going to accelerate, as you said, because there's so many tools are out [00:20:00] there that now people can some just go ahead and pick up,
Megan Sprinkle: Yeah, actually, you started to hit on some of my thoughts with that. So on the education piece, I mean, you were saying that human physicians were being taught informatics two and a half years ago, right? That was already part of their curriculum. it's been a little while since I've been in school, but I still don't think we have any type of curriculum that goes over informatics, AI or anything. And then on top of it. where are the rules like I know you said, you know, regulation that's that might get a sensitive topic. You don't want to go too detailed. So there's a task force. But how do we support veterinarians in all of this? Legally and all that.
that
Jonathan Lustgarten: as an individual, you know, we all have our thoughts. I don't know that answer, like, as a definitive answer. I do know that exposure education is critical. So, like, I've had, four years, since actually [00:21:00] was an intern at Red Bank under Tony DiCarlo. I have taught people about informatics.
I've taught people about analytics. you as a veterinarian, every single veterinary staff member does informatics. They may not, it may not be called that because informatics was a niche term, but everyone does it right. But I know a couple of people in the task force and brilliant people that are very experienced, you know, and luckily veterinarians and the, veterinary staff are brilliant.
But there's the familiarity of practice and all that stuff. It's going to be really hard to keep up with and really hard without that baseline. I don't know of a way to be successful longterm without there being some sort of educational program. You know, there's people like me, everyone has these types of platforms for learning. but these are like adjunct, these are like, Oh, you can go here. As opposed to saying, you know, we should really try to familiarize with what does it mean to use AI? What are your responsibilities? Because that's the other side.
As you said, legal. There are legal ramifications of using some of this technology. HIPAA compliant does not mean that your [00:22:00] information you give them is not being used in a way, or if something happens, that you're not legally responsible, right? You know, these tools are imperfect. My favorite fallacy is everyone expects, automated driving to be better than humans.
The humans are not very good drivers overall, right? Some people feel that they're better than others and that stuff like that. But when we say, you know, automated driving, the AI that drives the car needs to be better, what is your metric?
They must never have an accident. I'm like, that's that's like you can only be so much better than you know what you're trained as a baseline. And if they're using people who drive as their baseline, then they're seeing lots of people who make mistakes, right? So these are things like, okay, what is an acceptable level of performance?
How do we measure it? And that may be as a population. As you know, a pet is not a population. So when I have a reference range, that may not be, you know, the pet should be within it, but maybe that pet is different, right? Or maybe there's subtle changes within the reference range of the value that I had measured that [00:23:00] signifies some sort of disease, but it's still normal, right?
These are things that I think we're gonna have to figure out with AI. And I it's gonna require a lot of people, a lot of time, which is why, you know, things got on hold in human medicine a long time ago. But I do not know of a way to do it without education programs that support people in this and take the veterinarians and give them time to focus and learn these type of patterns.
Right now we have, luckily, very smart people that are figuring out how many 120, 000 veterinarians in the United States, let alone, you know, how many more hundreds of thousands around the world. A lot of them are not spending time, but everyone that uses software that has AI is going to be impacted
Megan Sprinkle: Yeah. And I think you, you pulled out an excellent point too, is for every patient is an individual. And that is where the human element of all the schooling that we went for comes into play. It's taking the data, including AI data that gives us and looking at the patient that's in front of us and saying, does this make sense for this [00:24:00] patient?
And so it's putting it all together pulling in why we're here and what's important and and also kind of going in that direction of the individual patient one of the trends that some people are talking about is this client centric shift in veterinary medicine. How is AI and all the things that we've been talking about going to impact the client?
How are they How is it going to touch them? How are they going to feel about it? Do you have any information so far from that population?
Jonathan Lustgarten: So there's already stuff like, you know, all the wearables, all that stuff is AI. I mean, there's lots of really cool wearable technology out there that is like, here's a alert, your dog may have an issue, right? False positives, false negatives, this stuff like that, none of that stuff is generally published, right?
So you can't really evaluate it, which leads to skeptics in the veterinary community, as well as skeptics on the client side, right? So I think [00:25:00] there's a lot of, this stuff is already out there. I think in, from like AI driven, you know, interactions with clients. I think what's going to be interesting is the new chatbots that are coming out.
I think that's going to be a huge area. One thing I do think for client is the ability to relay information, in a different way. One of my favorite things about like these, you know, large language models like ChatGPT and stuff like that, if you write something like, you know, it doesn't read right.
You can actually ask these language models to rewrite it, and that's actually where it excels. It does a really, really, good job at changing your tone, at changing the language you use. write this for a high schooler, right? Write this for a college student. Write this for a veterinary student.
I've actually tried that. It's quite entertaining, what they think a veterinary student level is. flaws in there. But, I think some of that ability to explain things, the ability to review, if I've written a paper or written a response, you can ask, are there areas in this [00:26:00] email that look unclear? Still veterinary side, but impacting client communication is going to be very quick. Imagine all of us have bad days, right? you know, I write a terse text or I, you know, say, hey, sorry, I can't, can't help you, right?
The client may not be as forgiving on those days. we're all human with all things, but a lot of times the, you know, we see veterinary burnout because there's a lot of angst and tension between it. Imagine a system that could flag, this tone may be a little bit more harsh. So that's still on the veterinary side, but it's interacting for the client communication. I think on the client side, the understanding what was said, That being able to book appointments through virtual and digital, I call it like automated appointment booking.
I think there's a lot of people that still want that. I think there's, that's an interesting dynamic that probably has a whole episode dedicated to it. but I do think a lot of clients are like, oh, should I go? I actually don't know if it's still there, but there was an Alexa app that was [00:27:00] built to Help the client understand is this a serious thing? Should I call my vet or not? Those are the types of things I see that are going to be developed for the pet parent. the wearables are already kind of, are kind of approaching that, right? When the wearable says, Hey, your pet may be in pain. We've seen them, you know, struggle to sit over three times in the past hour.
You know, you should call your vet. That is a AI device detects that about the pet and advising the client to contact the veterinarian. That type of question answer, that type of, I call it like pocket vet. going to be more and more there because on the veterinary side, and I'm sure you've had the experience.
you may not say it to the client, but you're like, Oh man, I wish you'd come in like three days ago. And your pet's been in pain. You may not say it because you don't want to make them bad and they may not have it. Right. That's part of the why wearables has such a good attraction.
Well, the client is not a trained veterinarian. They may not know that. Oh, an older dog that is struggling to get up, maybe [00:28:00] in pain versus, oh, that's just being older dog, right? You know, those are two very different statements They may not know.
And so having these types of systems to help them assess it is going to be a huge demand. I think. But again, the line of practicing medicine and giving medical advice for saying, Hey, and so what do most people do? Oh, anything we see, you should go talk to your vet. I think that defeats the purpose of what we're trying to do to give comfort.
The other thing is, I was actually just talking about record exchange. Being able to have your own records, imagine a system being able to query your own records on your phone and pull out the information that maybe you're at an emergency event because something happened. All that, I think that is also going to be more and more expected but it's again, it has to be done very carefully because these are, you know, a dog is not a small human, right? People can only relate to what they know. And so the software is going to be developed to improve that, but I think it's how they [00:29:00] do it and the way it's trained.
And as I said, we're going to hit a barrier with our data to train it because all these systems are not set up. So there's really smart people that spend a lot of money and time doing it. But that's because they have to compensate for a lot of the information that we're storing and using. Because this is going to be a need and a desire.
And that's where that tension of having veterinarians who understand and help with why do we record the data, I think is going to be very important for us to succeed. But I do think the client app is where. I think there's a lot of benefits that we can have and a lot of things that we can be able to use and do, but the last thing we all want to do is say, you know, the pet seemed fine, and then something, you know, tragic happens, right? And that, but, it may have happened no matter what you did, but everyone's going to point to the AI machine as the one that did it, even though you ask a hundred vets.
it's 101st. I would have said, Oh, no, actually, you know, maybe that's a sign. The other 100 vets maybe say, Oh, that's fine. Or the technician that they spoke to said, [00:30:00] No, well, you can monitor at home and, you know, come in. Seems reasonable based on what they said. you have the systems?
They spoke to a person. They documented it. Do these chat bots or whatever they have to talk about? Do they record that interaction? My favorite thing is you have all these chat things. You have all these people text. Are they going into your system? Do you have that record of that interaction? Is that interaction reviewed?
Because these things are going to be, even though they're not technically offered as advice. No matter what you put the disclaimer, people are going to take it as advice, right? That's so can we do something, make sure it's safe. And that's where the knowledge of understanding of what these systems are doing really comes into play.
Megan Sprinkle: so what I hear is a lot of communication improvement from a lot of different angles, 1 from connecting to potentially something that could be wrong with the pet and communicating it to the veterinary team. But also, some of the things that, you know, a handful of years ago was such a hot topic for us to talk about on [00:31:00] improving our communication with clients. Now we have tools that help us do that. Whether one, we're in a bad mood and we can make it sound better. Or because I think when you know a lot of things, it can be hard to communicate that in a understandable way. And one of my favorite examples that I heard actually at a veterinary innovation summit was a veterinarian used it.
to explain a disease process in the terms of basketball because she knew the client loved basketball and it did it and he got it. And so like something like that, like that's a really cool tool that we can use. And like you said, that's something that it does very well. again, there are some challenges and yeah, I think sometimes our, our expectations of these tools is high. I like the car example because we, if you look at the percent of accidents that have people driving extremely high, [00:32:00] and we expect that a self driving car should be zero. but I understand it because it's scary. It's, it feels out of our control. And when it's out of our control, we need it to be perfect. And so again, it goes back to being a tool that we can use to improve our overall flow and hopefully health of the pet. One of the things that you had mentioned in our last episode was trying to make sure that we are not data rich and information poor.
I liked that phrase. Do you feel that we are getting to a point where we are more rich in the information side? Is AI helping us there? Or do you still have this concern?
Jonathan Lustgarten: I think it's starting to help. looking at the number of systems that have AI to help extract information and find you what you want is still very low. So [00:33:00] I think we're still on the inverted side of having a lot of information in our system, just not able to recall it. The three R's right information to the right person at the right time, that's how you become information rich. I think our systems are helping pull some out. AI and general AI and extraction and all the fancy things making it easy to consume. All that still hasn't been applied to veterinary medicine, right?
So they haven't been applied to human medicine either. So it's not like we're we're behind. It's just it's Probably a little bit away, but I think we're getting there because I think their tools are now more centric on presentation of information, not just getting information. So I think some of the AI stuff that we're seeing is helping that get that information.
But I still think we're inverted. I still think we have a lot of data and When I say, you know, and data rich can also be applied. Is it good data or bad data? Right? But we're going to skip that discussion. Just avoid other topics. But I do think that [00:34:00] the data rich information for is going to be there until the system can help feedback.
You know, clinical decision support systems are still very rare and they are hard. I mean, don't get me wrong. There's people who spent their whole lives building clinical decision support off of the data in the E. M. R. S. And on human medicine, that struggle, right? A lot of it comes down to, oh, we'll just build a lot of rules, right?
And, you know, you can talk to human physicians about how much they love the rule systems in their, EMRs, right? So we have to do it right. So I don't think we have flipped it, but I think the tools that are being developed offer a larger potential now than ever before. Just as you said, because I can convert from one thing to another.
Right. I can convert from maybe the what a clinical pathologist would write as a long thing, and I can help it summarize it as long as that summary doesn't mislead. And that's the part where it's like, we have those tools that can do it, but can they do it correctly? Can they [00:35:00] do it accurately? And for some people, you know, the good people who say, oh, I look at like, that doesn't seem like what a clinical pathologist say and goes back and reads the original.
I and how many people are doing it? And that's where you get to responsible. But I do think the baseline technology is here. Now, I think it has to be customized and built up, but I do see that we are developing the tools, which is why I'm so pro education because we need as many people who know how to help develop the tools as possible.
I think we're having developing baseline tools, and now it's how can refine it and apply it, personally, I don't think we're there yet, but the past two years has seen more progress towards tools that are more generally accepted than I've seen in the prior five. And so I think that's super exciting for me because that means that we're starting to have people who want to say, how can we use it instead of trying to convince people, should we use it?
I think that that paradigm shift is what will drive more of the data rich to [00:36:00] information rich.
Megan Sprinkle: And one of the things you had mentioned early on is It can be a shiny object and people want to use it for everything. Is there something common you've seen people try to use it for? And you're like, no, no.
Jonathan Lustgarten: So for me, I've actually I like it when it summarizes a document to be fair, bias. I've been trained in this. I've been doing this for two decades, right? So I have a different perspective, . But to me, I think the use of documents. I'm going to use a transcription can be done. It's almost like they, when they take a yellow and black books, the criminal books, like when,
Megan Sprinkle: Oh, cliff notes and spark notes. All those. Yeah.
Jonathan Lustgarten: but those were written by human, but even then they messed up. This one is so, as you said, speak so confidently that I worry that when it's really sensitive or really technical, people are using it and not realizing that [00:37:00] it doesn't actually know anything about that area to the level of, like, understanding so when anything technical, highly technical, or highly specific to a discipline, and they're not using a model for it, I, I always say I would not use it.
I, or if you do say, hmm, use it as a high level outline, but do not copy anything in, make sure you read the source. And I feel like. Until we have a better hand on it. I think we're going to get in a lot of trouble. So I tell people like, if it's something that you are an expert in, don't use the language model because unless they downloaded your brain and we're like in the Johnny mnemonic type phase, then the tools are not ready for you for what you want to do.
I think it's great for high level. I think it's great for like partial bullet points, but anything super technical, I always pull back. And same thing with transcriptions. If you're dealing with a complicated case. if I have to read the record more than once, these systems may not understand it because we've been trained in it and these systems are not trained.[00:38:00]
And so more simple stuff, I may be more aware, but if it's something that you're questioning, the system is not going to necessarily be better. So don't, don't just use it and don't just objectively use it. And then also one thing I say, do not use the system if it has someone's information in it there are places like Azure, you know, the AWS Amazon has their own things where it's a private and there's all sorts of stuff with that going on the chat.
Gpt. com and putting in someone's personal record, saying, summarize it for you, you are giving, it's not chatgpt's fault. You are giving that company personal information on your client. Most of the time it gets sensored and stuff like that. They don't see it, but they have no thing. And they have shown that people using chatgpt will have another person have that information pulled up for them with the same query, not even in the same company.
So you have to be very, very careful. So if you have anything that's sensitive. Unless you have specifically set up privacy and, you know, siloing, do not use these [00:39:00] public tools that that is a way for you to get into a lot of trouble.
Megan Sprinkle: Yeah, actually that was a tip I learned at VMX just this past week is that there is a way in chat GPT you can actually go in and say, do not train on my data. And so it's a way to. have some more security on it. I don't I still wouldn't put sensitive information, but at least it gives you an extra level. So that's nice. But yeah, basically, everything is free for them.
What did they used to say is Nothing is free. If something's free, you're the product or something
Jonathan Lustgarten: Yeah, no, exactly.
Megan Sprinkle: And so, yeah, I think that education piece is probably what we're going to keep coming back to, because there's so many different tools you could use, you know, some are better for other purposes. the ones that I know we've got some out there that, that are. They're specifically fed data from veterinary textbooks or something like that. Um, so knowing what tool you're using and how it's been trained, I think is important. any [00:40:00] thoughts there before I, I go slightly adjacent,
Jonathan Lustgarten: Well, no, I think you said it best. And I think it's the understanding, you know, the impacts and problem is education. Like, what do we what do we educate? And I think that requires some discussion. But yeah, I don't I don't see us going forward. without an ability to do education as we go.
So we're going to hit a wall because it's going to be the people who understand and the people who are veterinary and practitioners until we start building that bridge with people who are cross trained, maybe bumpy, maybe a little bit more bumpier than we want it to be.
And, but I think the tools can do a lot. So what was it? What were you thinking about going in?
Megan Sprinkle: adjacent and actually even maybe even the education part, maybe part of this answer is, one of the first questions I asked you is how has all of this maybe changed your career. How do you think it might change careers in veterinary medicine?
Jonathan Lustgarten: it's very interesting. So first [00:41:00] off primary pathways is one thing. And then the second thing is what do people do after they've had a career in veterinary medicine? I think both of those are going to be huge because you know what people have been practicing for years, right? A lot of expertise are going to be able to transition to this type of advisory role or this type of situation.
that becomes a core to the product. You know, sometimes you have like, oh, here's a veterinary advice. They work for pharmaceuticals, whatever things like that. I think this type of technology as advisors is going to be a huge new pathway for veterinarians, either to start with their career. Cause you know, some people want to become vets then say, actually, you I like the research part or like the MLAI part, right.
Application. And I think that for me, what's going to be very interesting that there's going to be a lot of people who are 10, 15, 20 years in, they're going to be Well, I can actually help this now. You know, I practice, I really love my clinic, but now I can do this. I think that's going to be very exciting for, [00:42:00] our career and our veterinary industry, because there's going to be such a need for people who understand what the information is telling you and what quote, unquote, should be the output of the systems it's going to open up a whole new career post medicine that I don't think existed before that.
Megan Sprinkle: Yeah. And I think on so many levels, not only the medicine, but learning how to teach. Leadership, you know, like all of these things that each of these roles could go in and start helping training models that can, amplify what they know, I think is really exciting. You are far too easy to talk to.
I think our hour
is up. Any final words you'd want to leave people about 2025 and technology.
Jonathan Lustgarten: one thing about 2025 is that I think the new and shiny is going to be awesome. But I also think don't be overwhelmed by the shiny. Make sure that when you look at the product, don't just rely on the sales pitch. [00:43:00] Try to understand what it's doing and focus on whether you can adopt it. Because one thing I worry about is people pulling all the software and then realizing it doesn't help.
Not because the software or the technology is bad, but because the way the hospital operates or the way they want to implement this just doesn't jive. So for me, it's like, explore, keep reading, go to conferences like VMX, go to conferences like VIS if you can, I love VIS, it's one of my favorite conferences, go to all these conferences where these technologies and these companies are presenting, but also make sure to understand that when someone says they can do this, that's great, and you may say, oh, we can use that, but how you use it, where you use it, make sure you don't get too distracted by the shiny, because it, a lot of these things, A, they're not zero cost, And also they can impact you in ways you're not expecting if you're talking about, Oh, it can do this.
, there's so many new things, so many products coming out there, so much innovation and technology space that I think 2025 and the next couple of years are going to be huge, [00:44:00] but make sure that when you see something, you understand how you use it.
Megan Sprinkle: So don't jump full force into anything without, uh, A little education. I like it.
I hope you enjoyed this fascinating veterinary story. We can make an impact in so many places. Check out the show notes for lots of resources. Please make sure you are subscribed on your podcast app, subscribe on the YouTube channel and follow me on LinkedIn, where I hang out the most. You can contact me on LinkedIn, on the website at vetlifereimagined.
com. And brand new is that you can text me to send me a text message. Find the link at the top of the show notes below that says, send us a text message. I want to thank our longtime sponsors, fire consulting, and we'll use who support the podcast over on our hosting platform. Buzzsprout. You can support the podcast to just check out the show notes for a link.
And I hope to see you next time on Vet Life Reimagined..
