AI is transforming healthcare in groundbreaking ways. We chat with Davy Wittock, Chief Business Officer for Influx Technologies. Davy explores the exciting intersection of AI and medicine, diving into the soaring costs of AI compute power and the importance of decentralization. He sheds light on the challenges posed by regulatory hurdles and skepticism in the medical field, emphasizing how AI enhances patient care—particularly in radiology—while addressing common misconceptions about AI's capabilities. Davy also shares valuable insights on how healthcare professionals can evaluate AI tools and looks ahead to the future potential of AI in improving healthcare delivery!
9:03 The Cost of AI in Medical Research
14:14 Driving Down Compute Costs19:00 Challenges in AI Adoption in Healthcare
26:14 AI's Role in Radiology and Imaging
28:28 AI's Impact on Radiology and Workflow
37:59 Misconceptions About AI in Healthcare
41:58 Personal Adoption of AI in Daily Life
47:27 Navigating AI Tools in Healthcare
51:39 Security and Privacy
More Resources:
Podcast: https://podcasters.spotify.com/pod/show/leaders-in-healthcare
LinkedIn: https://cherryhealth.co/linkedin
Facebook: https://www.facebook.com/cherryhealthinc/
Instagram: @cherry.health
Twitter: @cherryhealthinc
Do you have a topic or speaker you would like considered for the Leaders in Healthcare podcast? Suggest a speaker to alitta.tait@cherry.health
Transcript:
Dr Jordan Vollrath (00:07.581)
All right. Today we're talking to Davy Wittock Chief Business Officer for Influx Technologies. Davy is a healthcare technology expert and the driving force behind business development at Influx. Where he spearheads initiatives that bridge decentralized solutions with customer needs in emerging industries. With a dynamic background in systems performance analysis, healthcare, IT architecture and enterprise solutions, Davy brings a unique blend of technical insight and strategic vision to the Flux team.
Davy has worked almost 20 years as a radiology IT professional, going from radiology information systems to PACS systems to independent archive systems, and has worked for AGFA, Visage, and Sectra products. Today we're talking about AI, technology, and the path forward. Davy, thanks so much for joining us. Appreciate it.
Davy Wittock (00:59.256)
That was a mouthful. Holy smokes.
Dr Jordan Vollrath (01:02.003)
It was, it was, I always loved the ones where there's like acronyms in there. Those are always fun to try and get off all in a row without messing it up. But yeah.
Davy Wittock (01:07.534)
Yeah, I think it was ECFA and Visage. Yeah. business officer for Influx. Yeah, been doing that now for at least two and a half years. And like you said, been in the medical space for about two decades now. So a lot of fun. A lot of fun.
Dr Jordan Vollrath (01:30.055)
There's a space near and dear my heart as well, you know, as just a lifelong computer nerd and also obviously being a physician in the healthcare world. It's sort of an exciting intersection of technology and health,
Davy Wittock (01:42.574)
Absolutely, absolutely. And then in my case, I even include blockchain of all things into it as well. yeah, it's a lot of fun. It's a lot of fun.
Dr Jordan Vollrath (01:52.671)
What's your Bitcoin position right now?
Davy Wittock (01:56.504)
It's pretty good actually, I can't complain. And I got a ticket over there that constantly shows me the current price. And for those wondering, we're at the all-time high at the moment. So if you wonder when it was recorded. So yeah, we're doing well, we're doing well.
Dr Jordan Vollrath (02:09.927)
the markets are ripping as of mid November right now. But tell me a little bit more about sort of like the AI and the computer systems that power our world. So a bit more about Influx. I understand you guys are really working on bringing down the price of compute power. And so on the one hand, we keep making these technological advancements, transistors are getting smaller, know, computes per second continue to go up. But then at some point we're just faced with the realities of
how much does electricity cost? That I can only imagine becomes more and more of an exponential problem.
Davy Wittock (02:46.421)
Yeah, no, you're absolutely right. There's a couple of issues that we're faced with. So the growing needs of compute.
and research and AI research and AI, generative AI, so everyday usage of AI, the demand is going up exponentially at the moment. And so that's a good thing. But at the same time, that brings up some interesting issues as well, right? This cost, the demand itself and what we have available currently today. And we actually surpassed back in early 2024, we kind of surpassed what we have available.
for compute power and data centers and the demand surpassed it. And there's no way we're going to catch up anytime soon. Add in that the cost of all of these compute needs and it becomes a very expensive endeavor. Yeah, we can use JetGPT for free, but we don't really realize that the cost behind it and the demand that there's there. So that's why sometimes we see sluggishness of the bot not responding fast enough or even errors just because they're
demand is just too high. So with Influx, we started on an endeavor to make that more affordable and to find a different way of providing that compute power. And the compute power is literally provided by folks like yourself and myself, all the way up to regular data centers that are close by in the city, all the way to the big boys themselves that provide compute power decentralized.
The decentralization part is what really drives me forward and why I find that important is that it basically doesn't offer us a single point of failure. We've seen that a few times in the past where centralized entities have issues with patching or just straight up downtimes. And then all of a sudden we're all down. We've seen that this year in 2024 with one unfortunate patch event that happened on a Friday evening. I don't know who the IT guy was there.
Davy Wittock (04:50.882)
But as far as my go-to at IT comes to is that you never do a patch on a Friday evening. Ever. That's a golden rule in IT. Apparently they didn't follow it and with all the consequences, of course afterwards. I was actually traveling during that period. I was in Europe. I got stranded in Europe because of that. Of course there's first places to be stuck. There's that. But yeah.
patching and all of these things that go wrong becomes a single point of failure. So decentralization at heart is key for us. So with that, and then toss in some blockchain to help manage everything. And all of a sudden things become a lot more affordable and a lot more robust. And the providers, they get paid by the blockchain in our case. And then it goes all the way to those consuming it. They basically have a less
cost when it comes to utilizing the software. And then because we're pioneers, we're cypher punks at heart, we like to find new ways of tackling issues that are out there. And we're not in the business in the crypto space that with the funny monkey pictures and all that, no, we're here to make a solid case on how we can help progress in this case, healthcare. And with my background of healthcare, knowing
the intricacies of healthcare, knowing that pieces like HIPAA and all that, make sure that it's private, secure, and making sure, of course, it's operational 24-7. Those are the pieces that really drive my business sense. And also, I push that forward to the guys themselves. And then tie that with their innovativeness. We really make a difference in that space now. Fast forward to now with AI.
We see that demand and we hear the demand. We feel the demand as well, because we get on the daily, we get requests in for more more compute. And so what I've done with my team is that we basically, we went to a ton of data centers, ton of providers out there. We brought them all together in our product called FluxEdge. And we are able now to provide compute power for a lot less than your traditional providers out there.
Davy Wittock (07:12.526)
I'm not going to name anybody, I know the cost of some of the bigger cards out there, the traditional cards that people utilize for research. We bring that cost down to a third of that price. And that's a big, difference. If you have to work with research folks, a lot of them work with grants and they're living from grant to grant. And if the grant is literally consumed by all that GPU compute.
Dr Jordan Vollrath (07:24.212)
you
Davy Wittock (07:39.086)
it affects the research altogether. So if we can bring it down to a third of that price, either we can give them more time to work or we can give them even more compute so they can even work fast. So that's kind of in a nutshell what we've been trying to accomplish so far.
Dr Jordan Vollrath (07:56.509)
And can you give us a bit of context, like for an average person to put it into perspective, like what does compute power actually cost? Like if I were to just go run a simple Google search versus putting something through Gemini, what are we looking at in terms of sense? And then how does that extrapolate out to a bigger, more AI intensive thing, like doing research or running these different background algorithms on research projects?
Davy Wittock (00:03.726)
All right. So when it comes to costs and computes, there's different ways of looking at AI. But if we start with the research one when it comes to medical AI, training models, training them on imagery and studies, that takes up a lot of time, a lot of bandwidth and a lot of compute power. What is that compute power? If you go to the traditionals out there and the larger players,
One of the larger players charges about six and a half dollars for an H100. And an H100 is one of the latest and greatest cars. I know the new Blackwell chipset is coming out from Nvidia. But before those are really fully in rotation, currently the big boy is the H100. You pay about six and a half dollars an hour for a card like that currently. That doesn't sound like a lot, but you have to keep in mind some of these model trainings take up
months if not up to two, three years. That really puts things in perspective and we're not talking about one card at that point. We're looking at at least a thousand of these cards working simultaneously. So that's about $6,000 an hour. $6,500 if you want to be correct. So yeah, that adds up really quickly, especially if you're doing it for a month, if not a few years.
Dr Jordan Vollrath (01:23.633)
That ends up pretty quick then.
Davy Wittock (01:34.274)
So, but that's pure training of your AI, right? So that's something that not everybody out there does. That's purely focusing on research, retraining, optimizing, and all the fun stuff that comes with it. That's costly, period. There's no way around it. The only way around it is if you find providers that do it more, that make it more affordable for folks by utilizing technologies like we do.
Or by providing data centers that are in locations that have cheaper power and other types of regulations that might impact price as well. If we come to generative AI, like the chat GPTs of the world, the cost of it drops a lot. At that point, it becomes a scalable issue. Like what type of AI am I doing? What's my demand? How many users do I have? What kind of...
What kind of generative AI am I doing? If it's the chat GPTs where it's responding in text and it's purely consuming that LLM, those are more affordable when it comes to compute. You can then have to start looking at what type of model you're using. And I know we're going really nerdy here now, but basically what it means is every model has a limitation when it comes to memory. Memory wise, most of the GPUs, even the GPUs in your computer right now,
actually can run a lot of models for you. And folks don't always realize that, but purely for generative AI, you can get away with a regular consumer card. Of course, with the caveat that the model fits on that card. Then the problem of course becomes, well, how do you use it, the technology behind it? So, ChatGPT kind of bridged that for everybody and made that a lot easier for everybody to But if you think about it, you can get away with like a
30, 60, 12 gigs, you can do a lot of things with that. It's not going to be as fast, but it'll be able to do it. And even Nvidia provides you software for that. There's desktop models out there where you can literally download the software on your computer, on your Windows machine, and you can use it as is and use it in LLM that they provide. And they actually provide, I think they provide a Lauma 3.1 now with their own flavor to it.
Davy Wittock (04:00.43)
And you can get a lot of way with that and it makes it a lot more affordable. But the problem now exists as well is that not everybody knows about it. Not everybody is comfortable with doing that type of work. And it's easy enough to go to chat GPT and just ask questions, right? And it's good and it's answers, but we need to...
try to figure out now is how do we make it affordable for those researchers as well? Because the more models we have, the more training that gets done, the more optimized these models get, the better the responses will be, the faster they will be, et cetera. And that's what we are fixing with trying to find more providers out there, bring a little bit more competitiveness in the market, which always drives the price down.
And then of course, NVIDIA is doing their part and even AMD is doing their part with their grok machines. And at that point, we'll see more and more compute becoming available and the price will start dropping at one point again as well, which ultimately for us medical folks out there, that's gonna be a big push forward.
Dr Jordan Vollrath (05:14.247)
And so When you want to bring down that unit cost for compute, where do you actually do that at? Like, is that at the hardware level or the firmware or the software that's running on these or more at like the models themselves and just asking and putting the answers together more efficiently? How do you actually bring down that dollar sign?
Davy Wittock (05:34.766)
The answer is yes, to all of them. So every aspect you just mentioned has its role to play, right? So Nvidia has their role to play with their cards. They, of course, have their own research, their own cost price when it comes to fabricating these cards for everybody to use. And then, of course, you have your providers themselves who have to procure the cards, install the cards, maintain the cards, manage the cards, all the personnel that comes with it. That all drives up that cost.
Dr Jordan Vollrath (05:38.045)
The end.
Davy Wittock (06:04.686)
So when then Flux comes around and we offer it, we basically offer it in a different type of way where we can offer all these different providers in one big marketplace. I don't want to call it a marketplace. It's more like an offering. We basically allow folks to say, okay, this is our cost. This is our bare minimum. We can, we take a little bit of margin so we can stay in business, which of course everybody has to be in business.
We drive that price down that way because currently what happens is because of the larger providers out there kind of having a monopoly for the longest, they were able to set the price however they like. And then we're not even talking costs like egress and ingress to send all the data to these providers to then start training all your models on. It becomes a costly endeavor. So we don't charge egress or ingress if we can.
Depending of course, sometimes some providers have to basically purely because of their internet capabilities. But most of folks out there, we try to not do all of that. So no ingress, no ingress. So that cost already drops. And then of course, with competitiveness, the price of the compute power itself drops. And then lastly, the software and then of course, the hardware kernels and all that. If we...
provide the software in such a way that you don't have to train IT personnel to actually have it to set up. That by itself is also a cost reduction itself. And to give you an idea, to set up a proper cluster, you need to have somebody who knows Kubernetes or Docker experience. Those are skill sets that are very common in IT these days, but still not everybody out there is very comfortable with that technology. So you have to be able to utilize that, set up clusters.
be in the known of all the intricacies when it comes to the NVIDIA toolkits and all that, that you need to set up, add in the software to run everything. So you might need some guys who know Python or Jupyter Notebook to help set it up. So all of those things become a burden as well on any video organization out there. So with Influx, what we did is we basically took that away and we said, hey, you focus on what you need to focus on, which is your research. And you say, okay, I want to have...
Davy Wittock (08:26.316)
thousand cards, couple of clicks on the software, you launch your network and you say, okay, I want to have a Jupyter Notebook so I can run my scripts on it. Sure. Launch Jupyter Notebook. And now you're off to the races. So we took away a lot of these extra hurdles to help folks with purely focusing on what they need. And then last but not least, the Genitive AI itself.
you're not comfortable with the chat GPT, especially in the healthcare environment. We have to be a little bit more careful if we use chat GPT. Purely because if we accidentally upload a report or something, or you want to analyze a study, but there's patient data for instance in it and you're you're not aware. Let's assume we're playing innocent here. You're not aware that there's patient data in it. You're actually uploading that data to an American company at that point. And that data might be.
that will be utilized for training without the consent of that patient at that point. So what we're trying to do now as well as teach folks how to utilize their own chat GPT and open source version of it so that they can run that on their own prem on their own data storage. At that point, we stay within the protected hospital or clinic or healthcare providers separate at that point. And then we're good to go. And again, we drop that cost down.
by providing open source software and again by offering our services to help set these things up.
Dr Jordan Vollrath (10:00.553)
Very cool, okay, so a very multi-pronged dancer.
Davy Wittock (10:00.622)
So there's a lot of variables. Yeah, it's not an easy answer. Like, how do you drive costs down? There's so many facets out there that play into the cost of AI.
Dr Jordan Vollrath (10:16.105)
Well, especially healthcare, like you mentioned, you know, industries that aren't necessarily restricted by health, patient health information, data laws, and being able to just like send the compute over to where energy is cheaper. You know, suddenly now you have this much more regionally restrictive problem needing to be figured out.
Davy Wittock (10:35.008)
Yeah, and it is an issue. is something to quite, sometimes quite daunting to tackle. in Canada, you're sometimes even restricted per province on certain types of data. know BC has their own laws where you're not even allowed somebody from another province to dial into your system and such. it becomes an issue sometimes where we put up a lot of roadblocks to protect the patients. Don't get me wrong, that is important.
But sometimes we make it a lot harder on ourselves sometimes than it really has to be. But you're right. We have to make sure we do the right thing and we have to make sure that data stays where it has to stay and protection of the patient.
Dr Jordan Vollrath (11:20.755)
Are there any common themes or trends you're seeing in the industry of how healthcare is utilizing AI?
Davy Wittock (11:29.678)
Yeah, and healthcare is such an interesting world to live in sometimes. We're very, very proactive when it comes to technology and we're very reactive to technology at the same time. I don't know how it works, but sometimes we have these brilliant moments where we push forward on a certain type of technology and at the same time we put the brakes on it and we stop. AI is a perfect example of that.
Dr Jordan Vollrath (11:53.393)
It is a little bipolar that way.
Davy Wittock (11:55.594)
It's very bipolar medical technology is always very bipolar. There's the FDA regulations that come in play and in Canada, we have our own, I forgot the name now. And they kind of sometimes form a roadblock. They have to, because they have to make sure that the software does what it does. But it also prevents AI from really taking off in the medical field. And you might be wondering like, why I see it all over.
No, not really. The really true cutting edge AI, we don't see that just yet because we're facing roadblocks with the FDA approvals in the US, for instance. A big hurdle that I've hear a lot of AI folks talk about, vendors of pure AI software. Their biggest hurdle is that as soon as they release the software, they are stuck with that version. So they cannot train the model.
as they go. Let's say you have a doctor, I'm a radiologist in radiology IT, but I have a doctor reading a study and the software says this is wrong. And the doctor was like, no, that's not wrong. That's an artifact, for instance, on a bad image. Being able to train the software at that point with that input, it's not allowed. So we can't on the fly teach the system. And those are the type of things that sometimes are really blocking us from really making an advancement.
On the other hand, have our own doctors themselves being skeptical and they should be. But we have doctors who are saying, no, this is really good. We need to start using it. And then you have the doctors that like, let's be cautious here. Let's not use that. Or I don't trust it. It's a bit of a mixed bag sometimes on how we advance these things. And I saw that with the company I worked for in Alberta. had a couple, we had three camps, the ones that were like, hmm.
Don't care. I do my thing every day. And then you had the doctor, I want to really start using it. And then the doctor's like, no, no, absolutely not. This, this, should not try that. It's, it's fun. It's, it's, it brings it. It makes it interesting. Sometimes the conversations I have, with different doctors, but
Dr Jordan Vollrath (14:10.089)
Well, think you raise a good point, right? Like culturally in medicine, it's just not like a very fertile ground for innovation to spring from. Like most people come from a background of like biology, chemistry, or just in essence, science, where the whole premise is to be skeptical and question things until you've been shown otherwise.
And that really permeates into the culture. And then when it like kind of trickles down into patient care of, okay, am I just being marketed a gimmick here? That's okay. They have the highest ad budget, this particular company or device or intervention versus how much does this thing that I'm looking at actually improve patient care? Or is it actually, you know, at the detriment of my patient? And then I think that's a huge part of how we wind up just in sort of such a backwards position.
in terms of technological advancements in healthcare. It's just not a great place to be seeing early adoption.
Davy Wittock (15:09.646)
Well, and add on top of that, sometimes cost, right? So in Canada, we're kind of tight when it comes to our provincial governments and making the willingness to advance certain things, if there's a budget for it or not. So there's that. And in the US, it's a different story with the way they operate their private health. But it's, it's, it's definitely an influence on the cost itself for certain things as well. And that
Combine that with, like you said, the skepticism of, is this really gonna help me? If there's a cost to it, people will say, okay, let's hold off for now. Let's hold off. Let's see if somebody else does it first and see if they see any bad things or any concerns or whatever they have with the software there and then we'll assess. So yeah, like you said, it's often medical, the medical field is really good. Microcosmism.
of what the real world looks like on adoption of technology. I really see that a lot. You have your early adopters who are really keen and often, like you say, they have more of a background in IT and more of the informatics type of background compared to those who have, like I say, more scientific background. They're a little bit more skeptical on the usage. You're not wrong there.
Dr Jordan Vollrath (16:33.481)
And then when it comes to, guess, sort of AI in healthcare in general, definitely just regulatory wise is very challenging because like you mentioned, you know, you get approved and then suddenly this version is frozen in time, right? Until you've gone through the whole rigorous process of having the version 1.1 analyzed and tested out to meet those same criteria.
Do other industries that are trying to get more technology and AI in them, do they just sort of accept a higher error rate or more bugs or security issues? Is that really what it comes down to? It's just healthcare, that bar is so high.
Davy Wittock (17:14.658)
The bar is high, but it's for good reasons, right? It's pure patient care. So that level makes our industry a lot more demanding. But AI in agriculture, for instance, they're going to utilize it a lot easier because it's not as big of a consequence if, I don't know, if one particular grain is not picked over another one.
Yeah, maybe production or the yield might be a little lower in one way or another, but it doesn't affect the person straight away. When it comes to finance, the finance industry, they're going to be probably very picky because then the dollar signs are affected. But depending on the sector, you're not wrong. There might be more interest in getting a leg up over somebody else and then compare it in competition of somebody else. But when it comes to medical, we're a lot more cautious.
Again, patient care in mind, it totally makes sense. Point in case, during COVID and after COVID, a lot of AI was created to help with speeding up the process of reading all of those x-rays that were taken. And there was software created that helped with putting that in to help detect, okay, this is potentially COVID, this is potentially just a...
Nothing or there might be something else going on And there was some really good software developed for that The problem was is that it I don't think many people really Installed it at one point because people were just too cautious and this was a fast moving Situation we were in that we had to move fast and still there was hesitance to really put it in and Not to affect the day-to-day reading of doctors But I do know some
folks did put it in and it helped them, for instance, triage. Some patients say, okay, these are the studies we definitely need to look at first, make sure that this is all good. And then it basically helped with more prioritizing the reading of certain studies compared to basically diagnosing somebody straight away. I don't think anybody out there in the medical field is directly diagnosing anything purely based on AI. There's always that human factor still in play.
Davy Wittock (19:34.892)
Unless you know something that I don't know yet.
Dr Jordan Vollrath (19:35.289)
What about radiology? Isn't the radiology like one of the early adopter signs of healthcare? Like I've heard years ago that here in Alberta, at least, like the majority of mammograms were being read by computer scans and then, okay, a human comes along and signs off on it. But it seemed like that was the first place that we really starting to see computer assistants take over that diagnostic role.
Davy Wittock (20:02.446)
Mammography is a really good example of that, There's a lot of studies available. So there's a lot of research that you can do on them. There's a lot of availability. It's very streamlined when it comes to it. It's always the same type of imaging. And then you have, of course, mammography with implants or without implants. But you're absolutely right. There's a lot of software developed already. And now we call it AI, but back then we would just call it a tool.
We just call it tools at that point, but it basically was already AI. You're absolutely right. And mammography is one of the branches within radiology who really make a lot of progress there. And we have good riddance too, because it helps the doctors a lot. And there's Volpera, I think is a big software provider that does a lot of the Volpera scores, is very common in radiology. That helps drive up diagnosis, helps with...
pinpointing the studies that we have to look at right away. So yeah, there's a lot of, a lot of events are made in radiology. But let's be honest, pathology is right there with them. Yeah, there's a lot of imagery being done where AI is helping and that goes with ultrasounds, goes with X-rays, it goes with CTs. To the point even now that you see AI within the machines themselves. Like MRIs for instance.
I'm not a doctor, just full disclosure. I am not a doctor. I look at an image and I'm like, I don't see what you see, doctor. But to me, it's a gray mass. Good or bad, I don't know. But I was at RSA, I think it was a year ago or two years ago. And it was an MRI machine and they had AI not implemented to enhance the imagery. I think it was GE who came with that. And it looked like I was watching a 4K movie at that point. It was so clear.
And I was like, okay, I see there's something that I don't think that belongs there. So yeah, like we're moving fast with technology and the next few years are going to be very interesting to see how that all evolves.
Dr Jordan Vollrath (22:17.065)
It's interesting that the AI is now actually like manipulating the images to be more contrasting between signal to the noise. I wonder if that's going to start trickling down into other areas. I've been picturing like the electronic stethoscopes where the volume of the heartbeat is amplified. Maybe they'll start picking out the murmur itself that you're listening to to make it a little bit more apparent or other such things like that. I think that's pretty cool.
Davy Wittock (22:44.106)
It's there. I think Philips has something created like that already. They were showcasing... Yeah, there's definitely progress being made in that as well. Philips is pretty familiar, known for their ultrasound machinery. And they already have AI implemented there as well, with enhancing, again, imagery, but also sound, and helping with EKGs, and even identifying patterns.
Dr Jordan Vollrath (22:49.666)
that's a thing? Okay.
Davy Wittock (23:13.23)
and all of that, it's coming, it's coming and it's coming faster. How often I get a question from my kids as well or family and they're like, how do you see AI evolving within the medical field? And I say, look, I hate to say it, but I like Star Trek and in Star Trek, all of the things, all the little gadgets that we saw in Star Trek back in the 90s, yes, I'm that old. We see that now coming to fruition.
Your phone, the communication device back then, but also you're starting to see that now in the medical field where they put you on a bed and hook you up on things and AI is already helping you with finding things. It doesn't mean it's a diagnosis, but it helps you find it's a tool to help you go forward. And we're getting close to that point now where you're just scanning people and they'll help figure out what's going on. It's a pipe dream, but it's coming fast.
Dr Jordan Vollrath (24:12.553)
Well, I mean, when you look at all the individual components of the job that I do, like you can start to see AI actually coming in and playing a role kind of along the entire spectrum. Like you've got these artificially intelligent patient engagement tools that are sending out like a history intake questionnaire and like changing it dynamically. And patients are filling these in before they even get there.
Right? So like the actual information gathering process, computers are now starting to like get involved in the process when it comes to the actual diagnosis. I mean, there's been computer programs around since I was in med school over 10 years ago where you would type in the symptoms and it would start giving you your like scores of likelihoods of different disease pathology being present. It's, part of the imaging now, like the scans that we're getting, it's analyzing those.
And then at the same time on the output, like, okay, we know what the disease is. How do we start just connecting you with the repositories of information on treatment options? And so I don't.
Davy Wittock (25:15.962)
Well, you're saying it. Here's the recommended medication.
Dr Jordan Vollrath (25:18.982)
Yeah. And so there's no like one, yeah, there's no magic machine like the doctor bought yet, but like it is definitely present pretty much along the whole spectrum other than even like the physical aspect of like what you do mechanically with your hands. still, don't have robots all over the place yet. Even then there's still some robots in like the surgical operating rooms and things like that. So it is coming.
Davy Wittock (25:46.494)
The craziest robots that I've seen and I'm talking 15 years ago was in France when I was working for EICFA. I was at a brand spanking new hospital and they had these robots and now you see them at Amazon and everywhere but they had these little robots that would literally drive under a bed and then lift up the bed and then bring the patient to its position where he had to be. I was flabbergasted when I saw that and that was like 15 years ago so I wonder
But that's France, so they do things a little different than everybody else, I guess. But was pretty cool to see. But you're not wrong. Day to day, if I have to go to a doctor here in Manitoba, often enough, I get a text message filled in this questionnaire already, and the doctor has all the things in it. And I can see that at one point where, based on the answers you've given, or my...
My blood pressure is this this morning and you record it every day and then based on the pattern that the air detects, okay, there might be a need to maybe give him something for his blood pressure. Something like that. You will see that happening. But the biggest thing with AI for me personally, what I like about AI is that it brings back the doctor itself being able to be the doctor. Like you're not just...
collecting the data now to help you your diagnosis. The data has been collected for you. It's been processed to a certain degree and then given to you as a doctor to help you with your diagnosis, guiding you to a certain thing. You don't have to agree with it and you'll be able to find the pieces that might not be correct for you, but it allows you to be the doctor again and focus on the patient. I think that's where AI really is helping shine a lot of folks out there.
And definitely in the radiology world, after COVID, there was a lot of what the doctors, what my doctors at that point called radiology fatigue. There were so many x-rays that to be read.
Davy Wittock (27:57.464)
pressure on everybody in the medical fields. I think AI helped in that regard as well by prioritizing patients, by helping pre-diagnose certain things, scoring. see that with memo, that's a thing already, but we saw that now happening in CTs, we see that happening in x-rays. I think a lot of that is helping doctors move forward and I think that's the most fun thing I see about it.
Dr Jordan Vollrath (28:28.314)
What would be like the number one most common misconception you see when people are talking about AI or perhaps AI in healthcare?
Davy Wittock (28:37.09)
Well, I think the biggest conception is that people are scared that AI is going to diagnose them at one point. I don't think that's in the cards anytime soon. I don't think any doctor is willing to hand over that process. I know there's a lot of radiologists out there that are scared that AI is going to take over their job. No, I don't think that's the case. think what it allows us to do is to do even more studies now on patients to get to the bottom of things faster.
and we focus on the actual diagnosis, on helping them read. I don't think it's gonna go as far as that the AI is gonna be doing your diagnosis. We're not talking about self-driving. We're not talking about self-driving cars here.
Dr Jordan Vollrath (29:18.823)
Why not? Like from my perspective, that seems like the obvious thing that the system is going to be pushing towards. Like doctors themselves are like one of the biggest line items on the budget for healthcare. And when you look at like medical textbooks, it's like algorithm driven and protocols like other than the actual like human elements and like the patient rapport and then the physical manipulation in 3D space, like
Davy Wittock (29:30.542)
I know.
Dr Jordan Vollrath (29:48.189)
I almost feel like the AI can do a lot of what I do on a daily basis. And then when you give it Scarlett Johansson's voice, I bet a lot of my patients would pick that over me at some point down the road here.
Davy Wittock (29:54.286)
Yes.
Yeah, I'm not gonna lie, doctor. I would rather go to Scarlett's practice too and compare to yours. I know I don't think that at the end of the day, there's always a doctor has to sign off on it. who's going to be that responsible person if we let's say we go full utopia and we say no more doctors, there's no more need for doctors, we're going to stick to the one thing. one of the things that we're dealing with is biology, right?
This is not a written thing. Biology morphs, biology evolves and point in case new viruses and all that. If we keep AI as is, and AI can learn a lot of things and AI can evolve with it, but it'll just flag it as an anomaly and or it might misdiagnose it altogether because it's trying to fit it in a certain bracket. A doctor will say, that's
I haven't seen that before. That's not normal. Not every doctor will pick up on it. We're all humans. We all miss things. But our human brain is still a lot faster than an AI. AI is trained. AI is following patterns. It's following... Or when a pattern isn't followed, that's the things that AI follows. At the end of the day, it's one and zero. It's not a human. And that's the piece that we can't forget.
It's a big tool and it helps us and it helps everybody out there. It helps me in my day to day with my emails. Man, my emails have never been as greatly written these days and they ever were. But at the same time, when I get an email from somebody, I'm like, okay, you did a good job Chet CPT. I know that guy didn't write that email like that. The human factor is still something you cannot disclose. AI will get us 99 % of the way.
Davy Wittock (31:54.398)
at 1%, that's the human factor that's still something that you and I have to do.
Dr Jordan Vollrath (31:59.941)
So actually that makes me, I want to ask you, know, like, so the, professional business part of you, obviously very like tech and AI focused. How have you personally adopted AI? Like where has it actually changed around your workflow in terms of just managing, like you mentioned your emails and like having things look a lot more well written now, like where else are you actually using AI as an AI expert in your own?
Davy Wittock (32:00.066)
Well, not me. I'm not the doctor. You are.
Davy Wittock (32:29.048)
Well, we call it AI now, but in my entire career, I've always been writing scripts and codes to help me analyze things. So one of the things that I did for EGFA at one point was when you saw me show up at your hospital, that was not a good sign. Something was wrong, either with the software we had or with the infrastructure. And so what I did at that point was create scripts to help me analyze network or databases that might've been having issues.
a million records of something. All that was done with scripting. So for me, it's not that new AI. What makes it new now for me is that everybody has access to that. And to the point that, I need a website created. Well, I can go to a website designer and have him write the script or, hey, Flux AI, can you create me a quick script to make this website has to do this, this, and this?
I'm up and running five minutes later. So for me, and the same in the medical field, AI is making me more efficient. Some people would say more lazy, but that's not a bad thing. Like if I have more time, exactly. So if I can do more things and focus on what I need to be doing instead of writing my emails and all that, and I can say, this is the data, write my emails for me and...
Dr Jordan Vollrath (33:41.833)
Work smarter, not harder.
Davy Wittock (33:57.838)
ship it off, I'll take it. Why not? It opens me to having more personal meetings and an interview like this, or having a conversation with other folks in the space. Everything that helps me make myself more efficient and more time efficient, I'll take it. And AI has been really the case for me. What I had to get used to though with AI was, and I think that's for a lot of folks out there,
is how good it actually already is when it comes to responding. And when you talk to support on somebody's website, you're not talking to a person, you're talking to an AI and it's really good at finding the solution. I love it. I personally love it. I went through a similar evolution back in the early 2000s with Google. When we all of sudden had Google available to us and we were able to find things more easily compared to
What was it back in the day? I think I used Yahoo and Alta Vista or something. And maybe you found something that was somewhat relatable to what you were looking for. But Google was that process back then. And now AI is that next step in the evolution and going forward. For me, it made me more efficient.
Dr Jordan Vollrath (35:14.193)
It is surprisingly good to that problem solving. I personally, there's been a few like just interpersonal conflicts that have popped up like little family drama things. Like somebody got invited to Thanksgiving, which pissed somebody else off. And I was like, I don't really have a great answer on this one. Like I could see why both people might be a little upset here. There's been like more than one occasion where I've like went and popped it into chat GPT just to see what it advises.
And it's usually actually come up with pretty solid advice each time. Yeah, literally. it's just like, it's, very interesting how it like takes a very kind of rational, logical approach to things. And it's like, okay. I hadn't considered that. That's a good suggestion.
Davy Wittock (35:44.632)
How to tackle this Thanksgiving dinner. Yeah.
Davy Wittock (35:59.458)
And coming back to my Star Trek analogy, I feel like I'm talking to Spock at times. Like it has a snarkiness sometimes in its answers. Or if you ask it something and it's actually a dumb question, it kind of gives that snarky response sometimes. I love that. does that. But yeah, you're absolutely right. I personally haven't done it for family emergencies, but I'm kind of curious now. What to say at the Thanksgiving dinner for the old...
Dr Jordan Vollrath (36:25.353)
Try it next time. Yeah
Davy Wittock (36:29.12)
because I know at Christmas and all that coming, there's a lot of family dinners or friend dinners for some of us. Yeah, some interesting conversations. One of the fun things that... No, but what my kids are very heavy in what is a DND. I never got into any of that, but they love a DND. And so they've been using Flux.ai and that's our own chat, GPT. And so what he, my oldest...
Dr Jordan Vollrath (36:39.977)
So yeah, not just tech support.
Davy Wittock (36:58.014)
uses it every Saturday. And so they basically give it some parameters and then say, create me a DND story for these parameters. And then off they go. And basically, instead of having a dungeon keeper or something, what they call it, they use the AI as the dungeon keeper. And then basically they give the responses and the dice results and all that back to the AI and AI is just dragging them along through the conversation.
the extra flavor to it as well is that they add imagery. So then like, okay, this is the scene. And then they push the button and say, hey, give me an image of it. They're like, for them, it's more immersive. their D &D is not just a table game with the guy talking about it. It now becomes more tangible and that's a day to day thing. But I love that the kids are embracing it. I never thought about it, but yeah, there they go.
Dr Jordan Vollrath (37:55.269)
The additional use cases just continue to pop up. What do you think for healthcare going forwards in terms of the workflow, in terms of the administrative support side of things? Like, is the innovation going to have to come from the EMR vendors themselves because they're sort of the broker and the holder of all the data and all the information? Or is there going to be ways that people can incorporate AI that aren't kind of hamstrung by being part of that central software system?
Davy Wittock (38:27.146)
That's a good question. You're not wrong. The EMRs are going to be definitely a gatekeeper to some of these things and not to hold back on the idea. They're going to present all the tools out there. I think it's going to be individual doctors still. That's often what I've seen in all the hospitals that I've been in North America and the rest of the world. It's always a doctor who pushes on something going forward.
and then the EMR follows. I don't think EMRs themselves are really gonna push forward to AI. Reason being is that the more bells and whistles we add to a system, the more complicated it gets, the more harder it gets to support. So they're gonna always hold back. They're gonna provide it if you want it, but they're never gonna push for it. I think that's still the individual doctor moving forward. That's a personal opinion though.
Dr Jordan Vollrath (39:21.993)
When it comes to the individual doctor, like they're thinking about doing it, like what should they actually be evaluating in terms of security and in terms of privacy? Like when it comes to just an average Joe who's not an AI expert, like how do you actually approach it in a safe manner? Okay, come on down.
Davy Wittock (39:39.15)
They can always call me. Like I said, they can always call me and I can help figure that out. no, when you see as a doctor, when you see something that's interesting for you, always ask for a trial. A lot of the vendors out there do provide you with trials and those EMR platforms themselves, their marketplaces often come with something like that where you can trial it and see how it goes, validate the data.
So let's say you're doing something with x-rays and it has to find something, push a couple studies where you know something is there and see if it does detect it and then know, and then some studies that you know there's nothing, push them through it as well and see if it detects something or not. So validating yourself, right? So it's like driving a car. It's not because you like a car and it looks fancy and it...
It has all the bells and whistles that that's the car you're going to purchase at the end of the day. You want to test drive it. want to make sure that it's up to your liking. And if not, no harm done. Let's move on and find the next one. The positive thing about medical AI at this point, and I'll be at RSNA in a few weeks, there's an entire floor now dedicated to AI. There's so many tools out there now. So you can actually...
window shop a little, find a tool you like, and find a tool that adds to your efficiency. Not just add tools to get more scores, because that's going to be the next problem that we're going to have, is that we're going to have so many tools all coming with their own metrics behind it saying, is flagged, this is not flagged, this is high priority, medium priority. At one point, we're going to have an AI go through the AI results. So we're going to have to be a little bit
cautious on what we put in. So I would say window shop, find the tool you like, test it out and validate it and then go over and push for implementing it. When it comes to security and data protection, those are really good points. And those are questions you need to ask at the start of that journey anyways. Where do you store the data? Am I uploading it to a cloud? Am I keeping that data on-prem? Because a lot of folks out there don't realize when they install some
Davy Wittock (42:04.342)
of these marketplace softwares. It often is a upload software they download, they connect to a central server, do the processing there, and then get the result back. You might be at breach at that point with some of their local regulations. So you might want to be able, want to make sure you ask the right questions in that case. So where do you upload the data?
Dr Jordan Vollrath (42:27.057)
And is it as simple as that of like, does the data live and where does it pass through? Cause I feel like this is probably, you know, the, skepticism when it comes to the security and the privacy side of things, there's probably one of the biggest barriers to that adoption is, know, the, innovation is coming from companies that have only been around for, you know, a blink in time, comparatively speaking to most of the computer systems or Oracle or, you know, tell us provided stuff that we're using.
And so there's these smaller companies that just don't have that brand name and that like basically just soak time and market presence. Like how do you start getting more confident that a company is in fact reputable data storage and where does it go? Like what else would you look for?
Davy Wittock (43:13.506)
So questions I ask when I'll be going to RSNA, I'll be asking similar questions. Not from a doctor's stance, but just purely to educate myself. So like I said, is the software running locally? What is it gathering? Are you uploading it somewhere? If you're uploading it, where are you uploading it? Is that something I have access to or not? If I don't have access to it, OK, that's fine.
But are we doing anything with anonymization, for instance, or not at all? Am I responsible for it? Are you taking care of it? Have you worked with any other hospitals already? Is the FDA cleared? Often enough, the answer will be no. No, we have not had our FDA clearance yet. Okay, at that point, you can't even use it for diagnosis purposes. Now you have to use it as a test case. You can still go forward with it, but you can't actually use it with true patient data yet.
Well, countries you can, certain countries you can't, depending where you are or certain states even. But you have to be very vocal and make sure you ask a lot of questions. And in the end, ask for references. What doctor have you worked with? Is there any doctors been involved or not? What kind of... What did you guys do before this software? Because I've talked with guys who...
straight out of college, had a great idea, met school and build it up. Sometimes it's a good software too, but they might not always realize some of the intricacies that comes with the data itself and the patient safety that comes with it. Especially if it's young guys, they're more focused on the result. And often enough it's a good software, but there might be pieces that they might not be aware of.
I know a lot of the EMRs out there and the marketplaces that they offer that they do some of that due diligence. But still, be able to try it. There's no trials. That's a red flag in my book. If you're not willing to do how if you try it out and see how it goes. Those are the questions I would be asking. Where do you store it? How do you process it?
Dr Jordan Vollrath (45:30.697)
Okay.
Okay, that's the big ones. Well, I guess before we wrap up here, any final thoughts on the future of AI and healthcare, anything else big and exciting you're looking forward to or wary of?
Davy Wittock (45:43.488)
Yeah, get familiar with it. It's coming and it's going to come fast. And it's definitely going to help us all regardless. Will it make mistakes along the way? Absolutely. That's why we still have the doctors in place to help sign off on it. And I'm looking forward to the bells and whistles that are being created.
as we go, see priority filtering as the main one for now, but I'm kind of keen on figuring out and for me myself, I'm hoping that my contribution with the compute power will help find solutions to everyday patient care. And I hope that that's going to be going forward, just going to amplify what we're doing here.
So AI is here, AI is here to stay, and it's only going to get better.
Dr Jordan Vollrath (46:47.781)
Incredible. Well, thank you so much for taking the time to join me this afternoon. Davy Wittock Chief Business Officer for Influx Technologies. Thanks again for taking the time this afternoon.
About the Author
Cherry Health
Canada's Medical Network
About the Author
Cherry Health
Canada's Medical Network
About the Author
Cherry Health
Canada's Medical Network
Listen On
Listen On
