AA111 - Exploring AI in Health Care (with Stormy Dickson)

Stormy is back on the arguing Agile podcast. Stormy is back. Howdy. Howdy , and also we have Om Patel Enterprise Agility Coach, and also The Tsar of Tom Foolery. Tom Foolery. I love Tom Foolery. All right, so this today we're, don't talk about whether AI has impacted your job. I think from what I'm reading, it's not so much about has AI impacted your job or cause job loss, but will it? So more fear and anticipation of that happening than it actually happened. So lots of layoffs, obviously that have happened recently. Yeah. But having nothing to do whatsoever with. Ai, but now this kind of fear, of what does this mean for my job? Yeah. I mean, maybe currently we don't have thousands of people laid off because of AI directly, but that, that's really the concern that a lot of people have is, am I gonna lose my job to ai? is the parallel to the industrial revolution here that like the way of doing work is changing significantly and therefore we're at the end of an era in the beginning of another one. Is that, the parallel? I believe that's exactly what's happening. So when I started thinking about this and kind of equating, Artificial intelligence in the manner in which potentially it could be utilized and how that's encroaching on our knowledge workers. So the people that currently would be responsible for writing code, writing tests for code were so much of that in our current state right now. Even in the infancy that we are in, um, you know, AIA has been around for a while, but it's blowing up because we are finding valid use cases for this to be utilized in a good way alternatively, you have things that cannot be replaced by a human. I would say that quality to some extent, certainly user, the user acceptance. When you put this in front of a user, you need to have a human be able to evaluate that. At least that would be my hypothesis. Oh, maybe, maybe, maybe there that human's robot could evaluate it in the future using ai. I'm thinking about like a future u a t, right? Or what business acceptance has to use your acceptance test, whatever I'm thinking of like a future one is like, Hey, do you like this feature? And you just hit the thumbs down and as soon as you hit the thumbs down, it goes back to home office. And home office starts modifying code to get pushed out. Like all AI starts modifying code to get pushed out based off of x number of people have hit the thumbs down for this feature. Now I'm gonna modify this feature based on whatever the model has been trained to do. Mm-hmm. And you know what I mean? Your dislikes are training the model that processes and creates the software and spits it out and deploys it. Mm-hmm. I'm like, testing. Ooh, ooh. It's a very, very product minded thought process there. Product I, I would say, but I would say a QA minded QA mindset. Qa. QA as well. So qa. But what I would say is, isn't that already being done? Very inefficiently, yes. Thank, yeah. Okay. Yes, I would, I agree with you. It's being done extremely dis, like dis efficiently, uh, inefficiently, not inefficiently. That's not a, that's not even a word. I'm making up terms AI can't stop me. Yeah. I think in general, abstracting that, you know, AI's gonna do things quicker, faster. Because currently what happens is using just that example of testing it's perspectives. Yeah. A tester might say, good, enough's good enough. Another tester might say, Hey, now it doesn't meet exactly what I'm looking for. so there's different perspectives, right? AI may be programmed for the lowest common denominator, depending on what your objective is to get things out there. Good, enough's, good enough, or no, it isn't. It needs to be a hundred percent and it's probably not gonna be one over the other because it's context again, right? Mm-hmm. So where you have, situations where it can impact lives mm-hmm. You don't want good enough, good enough. You want absolutes probably, because even one life loss is too much. I feel we didn't, we didn't give the industrial Revolution a fair shake because Right. The industrial revolution wasn't really about putting people out of jobs. The industrial Revolution was about taking an, agrarian culture and moving them to an industrial culture where a lot of people worked in fields and families had to have a lot of kids cuz that was the free labor to work in the fields and stuff like that. And now people are sending their kids into the cities to work in, industry and work in factories and stuff like that. Mm-hmm. And it was it was a transformative time. Transformative, is that a real word? Transformative. Transformative, yeah. That's, that's a better word. I kind of wanted to touch on those, but I, I just don't, I don't see how we're gonna shove all this into the podcast on the, like, at the intro especially where people are like, oh God. So socioeconomic implications of becoming an industrialized nation. Yeah. It's a little too heavyweight per, yeah. It is a bit heavy so I'm trying to align here. So with AI and the Industrial Revolution. The Industrial Revolution, which kind of pulled a human workforce into these factories. At the same time there was actually. there was a couple of things that followed. And that was building machines that then replaced the humans in the factories, along with machines ultimately, that would go back and replace the humans from an agricultural perspective. So we have less, less need for humans in the agricultural sector, so there was this, very gradual transition, right? So something really sudden that, that, that happened and then a gradual transition and adaptation to that. And that's kind of, I see where, where we are now, we're not really sure where this ultimately is gonna go. There's a lot of fear. I hypothesize that this is going to disrupt our lives as we know it for many years to come. We're going to be adapting and adopting to this. I have the opposite opinion. I think, humanity is in the information age now. Mm-hmm. And I don't think this breaks us out of the information age. Mm-hmm. I don't think the disruption is like the marker of the end of any era in particular, I'm willing to be wrong 10 years from now on YouTube, I don't think that we're gonna break outta the information age. I think the next there I'm some distance away. We're expanding. You don't see this as a, the foundation of a huge paradigm shift. No, no, no. See, I kind of do. That's good. And I, and I see this, in my mind, not the AI itself being the big shift, but being the, the catalyst for that shift. Right. Okay. Uh, no, that, that's, that's good. I'm, I'm glad that we have differing opinions because, because we're doing agile. Let's talk about disruption, market disruption now, and when you were talking, I, like, I couldn't stop to think about an example of, well, if we had microphones every time a doctor talk to a patient in a professional medical setting if, if they were micd up And we fed all that dictation back to the model so it can be trained. I mean, obviously labeling would still have to happen. You'd have to feed it back the medical records after the fact to find out if the diagnosis was right or wrong or whatever. But if you're feeding the conversations that are happening and the diagnosis is after the fact, diagnosis is the diagnosis. Diagnoses, diagnoses, diagnoses, yeah. After the fact, to the model. And the model can learn from it. So if you're willing to expose it to, to enough input, records before, after type of stuff, the model will eventually learn basically to be a doctor. Uh, I disagree. Okay. But that's, lemme tell you why. That's the point of the model though. No, , no. But what you described is already in play and has been for a long time. Okay. It all, I feel like you've had, you've teed me up so, um, so just, are you arguing against AI now? Is that what you No, no, no, I'm not. But I mean, so to replace a doctor or a nurse. So I think there's a difference. Um, uh, and maybe, maybe it's not as simple as, as simplistic as qualitative and quantitative, but that, that would be a good, a good dividing line. Right, so, just to think about this, but your large language models didn't happen when ChatGPT released. Its, its l l m in November. That's, you know, these have been being built for a very long time. And exactly the way that you just described from a medical perspective, Nuance has been working on this. For, a very long time, and I'll give you some examples. So for instance, when you think about how a doctor or anyone in medicine speaks, it's a, it's a different language. Like coding is a different lang, you know, different code. Yeah. It's a different language, right? Some they can't, when they speak to someone else and they're speaking in their medical code, but Dragon, Nuance's application was able to utilize that and gather that data from many, many, many, many different voices. And then separate it out because not only does a medical vocabulary differ, but different genres of medicine within medicine also have completely different languages. A cardiologist language is very different from a gastroenterologist is very different from a gynecologist, so on and so forth. Now we start capturing slices of these. Mm-hmm. And going like, okay, if you're a gynecologist or if you're a cardiologist, these are the words that you're gonna be speaking in context with something else. Okay. So now we can start making assumptions. Right. So I'm, I'm, I'm kind of explaining all this because it's something that I've been working on for a lot of my career. Now we apply this to what you were saying. We, we could just replace doctors and nurses. No, we can't. What we can do is alleviate an enormous burden on them in regards to all of the documentation and the regulatory requirements and the coding that they need to do and allow an AI to manipulate that data. A doctor needs to be able to hear, examine a, a computer isn't going to be able to do a physical examination. Right. From my perspective, you can use an AI to take data that's been gathered by a human, a physician in this case that does a physical examination, because that needs to be done. We're talking about a woman having a baby. If physical examination needs to be done, there's additional data points potentially from like imaging studies or maybe, you know, or laboratory results and, and, and, and different. You know, kind of, uh, bodily measurements. Mm-hmm. Those kind of things. Those could be digitized as well and in input into the consideration of an ai, and also certainly mapped to, to create the codes that we need for billing and so on and so forth. So all of those things that are just so burdensome from a regulatory perspective for the providers to alleviate them. But you can't ever have, , I can't say I ever, I don't want talk in absolutes, but at this point or no point , in the near future, are we going to have a way for, a computer to do, to do Leopold's maneuvers where we can actually start actually feeling a fetus inside of a mother that has to be done with physical hands by someone who knows how to do it. And then they need to get that information along with all, all this other potentially digitized information. For instance, maybe an ultrasound and some other things. All of that thing can come together and now an AI could potentially make some, some diagnoses, observations, things like that, but you realize like at the same time or an examination, but at the same time you're having this, this, this discussion with me. I was gonna say this, this spirited disagreement, it's not really, we haven't really, really disagreed yet cuz, machines can't do that job. Right? Like, you just can't flat out can't. if, if we're gonna take it back to the, to, to, to our, our little swim lane. Let's talk about team coaching. I don't, I don't know how deep you can disrupt the market of team coaching. Like, can you get an AI to tell you like, Hey, yeah, probably your team's probably getting too big. Like, can, what? Can like Google meet automatically? Just start reading your meetings and listening to all your meeting back and forth in your meetings and start saying like, you probably should split this team. They seem to be working on different, like, is that the level of AI we're talking about? Uh, not in the near future, I don't think, because again, what's underlying all this is context. I think it goes all the way back to the, the medical example too, right? Data and a pool of historical data is one thing. Taking that and then making inferences is another right. But you need a human being to interpret stuff because there's that depth. So you don't always say, given these, comorbidities for this patient given, you know, given their ailment currently this is the recommended diagnosis, this is the recommended prescription to go forward. Because there's other things at play, which a trained medical professional needs to interpret? Mm-hmm. I don't think you're gonna replace that anytime soon, which is where the disruption lies, in my opinion. So there's no disruption really. Mm-hmm. What's happening in that example is, The doctors, medical workers, I'd say generally the broad broaden the category they are getting the information they need and the analysis of that information they need much faster. Yes, that's it. That's the extent of the disruption. There's really no disruption here. And it's happening currently. This isn't so and has been for a long time. So your clinical documentation, basically they have been able to already for many years now have, or since certainly we've, we've adopted electronic health records, being able to take those data points and make suggestions as to, diagnoses, medications, alerts alerts for things that's nothing new. But it, it, it always then came down to the physician provider's, expertise in discretion as to whether they understood or, or, or, or how they should move forward on those recommendations. And many times as a check, check, check, yes. Awesome. Okay. Check box, check, all implement orders are done. You know, CPTQs are done, DRGs done all my billings coded. Cool, cool, cool. And other times it's like, yeah, yeah, no, just whatever. This doesn't apply to this patient. So again, that's all data based, that's all electronic data based, that is being fed to, to a human that evaluates that in the context to your point of another human to determine if that's appropriate or not. But when we're talking about creating those data points mm-hmm. Human to human, that must happen. Human to human. Yeah. But that's the, to the, to the model that that is just labeling. That's just labeling. It is labeling. The, the model is just the AI assist looking over your shoulder until you've done enough and it's studying you when you're manually labeling things God, how did I get on the side of ai? Each patient is unique. Yes, but like, also like every question that you throw into ChatGPT is unique, but it's dealt with so many questions cuz it's, it's been over the shoulder of a human labeling for so long and it's learned how to respond to them. Now. It knows how to respond. It doesn't know how to respond to them. Cuz it can't, like, it can't tell me not to be the, the czar of sassiness or whatever the asked it to do. The maharaja of magnificence. The maharaja of magnificence, that's what it was. Uh, but I will assume the car Indy! um, to your point though, so if that's how it learns, so, but if it, so at some point there may be a crescendo, where we actually get to the point where an AI is able to, based on the input of a specific individual and the context of the patient and patient types that they work with. Sure. And that's where we would have like real-time learning, real-time input. Sure. so it's constantly learning not just from the internet and other practitioners, but from this specific practitioner and their specific cohort, cohort of patients. Sure, sure. And at some point there may be this, be this crescendo where they, where an AI actually becomes more accurate in diagnosing and more efficient and with less errors than the actual human. I would say that's possible. Nowhere on the near horizon. But yes, I would say that's possible. I mean, that, that would be the beneficial place for it to get to is like, I would like if my doctor had access to whatever my phone tracks about my health, like whatever, whatever Google keeps on file. Like, okay, my doctor has realtime access to that now. Like if I wear like a smartwatch or whatever, like my doctor has real time access to like my stats from that now, like, I'm o I'm a hundred percent okay with that and it feeds their system, which tells them, you probably should tell your pa like your patient is doing whatever. Your patient's not like, not sleeping well, not doing whatever. You probably should check in, and then I get a personalized check-in from my doctor. Like that's, that's, that's, that is enabling better quality of service. Yes. That's already there. The adjunct, that's not what I'm talking about. I know. I'm not going there, but that's where we're at now. I'm going, I'm going with a flyby net, like I'm, I'm going now. That's where we're at, the South Florida Pill Mill version of doctors. Enabled by ais, we're gonna put your, your, your family practitioner out of business by phoning it in with this model that's trained on so much or whatever that it, it can, like 98% of the time it gets it right. But the, it's the quality of the information that's being provided. Right. So WebMD is BS because. But the people who are feeding it information are unqualified and biased and Right. So it's like Wikipedia basically. It is, it's Wikipedia for medicine, right. That said, once, if we can get qualified, experienced, personnel to be feeding that ai, information based on their education and experience that AI will additionally learn and become better and better. That said, it still does not ever get hands off of patients well, so that stuff already is out there now. Like, look at the, the human, the human interchange or interface, where it's called thi.Ai. Right? Basically these things really help medical professionals today. Mm-hmm. We're nowhere near the discussion of, uh, instead of. A person interpreting the data, it's going to do the interpretation for you. Mm-hmm. And making deterministic analysis. We're not there. I don't think we're gonna get there anytime soon. No. But that, those tools exist today. Yeah. Right. Well, they'll actually, I mean, they are evaluating and making deterministic, uh, ev, Eva, you know, it, they're not making decisions, but Exactly. But that falls back onto the provider to decide whether that is appropriate or not. And then I'll throw this one in here too. Go ahead. Our litigious society, could we make more money off of the big data company that is providing the AI that has incorrectly diagnosed and and, and suggested a treatment that was incorrect? Who or and will a doctor be like, yeah, I'm hands off, let them take the fall forward instead of me. No, the AI will automatically start suing people. That's, that's right. I mean, that'll be the but that'll be the lawyer. Yes. So the be as best That'll be the lawyers AI trained. Yeah. They'll be looking for those. No more ambulance chasers. No, no. They're just going to be plugging in all this data. And, you know, you, you, you're going into a, the ethics topic. What does the lawyer do? Does the lawyer subpoena for the model and all the labeling information that the model is built off of when someone gets killed from a model. You know what I mean? From an ai, like what does a lawyer do in that? I, I'm very interested actually. Um, but also, I don't want to go into this. Uh, it's scary and um, it makes me wanna cry. we were talking about disruption. We didn't talk anything about what happens when you eliminate an entire workforce. I know there's probably companies out there that already are thinking of what I'm talking about as a product is like, you don't need a doctor, you just need a model that's trained on real doctors and then you don't need a real doctor anymore. That's terrible. And they will advertise that with like a a 0.0 point disclaimer at the bottom that says, certainly, you know, not FB approved, blah, blah, blah. The longer, this is what we're talking about of like workforce slash market slash opportunity Disruption is, let's fast forward this dystopian hellscape to say, what is the average age of a doctor in America right now? I don't know. It's 40. It's be over 40 so we, uh, like I checked on the internet and the internet doesn't lie. It says the average, average age of a doctor is 53 years old. But that doesn't, that doesn't dig into specialities. I like saying special. I don't like saying specialty. I like saying specialty cuz it makes me feel special. So like what happens when those people are replaced by ais? And, uh, or like, uh, like age out of the workforce and less people come in that's already happening, mid-level folks are now raising up. Right? So you got, you know, anesthesia nurses now actually administering anesthesia without the anthes being around the helicopter doctors. Yeah. The reason for that has been the reduction in medication errors, like the, so there's a huge reduction in medication errors when you are able to take the human error out of it. So where all of a sudden I'm not mixing, I'm not mixing drips anymore. I'm not calculating drips anymore. Right. That is done by a computer. I know that you can, but in general, these are things that are done by a computer. It eliminates that, that human error factor. Mm-hmm, there's some real good that. Yeah. There are benefits to automation. There's no denying that. Right? Well, uh, okay, AI can't replace all the doctors. Yeah. But AI can replace like a, a certain number of doctors. Like there will be a certain part of the market that, like this, this flyby night version of a doctor, this, this cyberpunk version of a doctor, replaces. And maybe that market share gets bigger over year, over year. Like much like automation does, automation, gets better and the data models that are powering this get better year over year. the average age, I don't know if it would get younger or older, because the people are leaving the market and not being replaced. Those opportunities are not there anymore. Mm-hmm. That's, that's the, the purpose of talking about disruption. You, when we're talking about it in this hyper-specific category, that there's all kinds of pushback, you're ascribing a level of difficulty where I am looking at it, I, I would hope in the similar, light as a data scientist would look at it to say, everything you do is just manual labels. Everything you do is just more records and more input. And the then if you could feed it everything you, if the AI could be sitting on your shoulder through your entire workday for the next 10 years, like it would learn your job with appropriate modeling and labeling. No, I I think you have a point about data labels though, right? given enough of a large data set. Could be years, I don't know, enough of a large dataset. It could have enough, variance that it can learn about through the dataset. There's always the possibility that it doesn't know this one instance. Right. Right. Just such just like a person, the model learns like a person. And, and, and if we're not talking about basic model building and labeling. We can go into neural networks, which are way more complicated, and now we can, we can knock out a whole bunch of extra, like, that's a different podcast, I feel. Yeah, yeah. Because now you're into, now you're into experiential learning. Right? And that's, yeah, it's getting a little heavy. So just as far as medicine versus, so, so I think there's a real differentiation here where we're talking about, let's just use the example of utilizing AI. And how that would disrupt, for instance, the development, the application development and coding industry, as opposed to how it will disrupt, much more human involved industries like healthcare, I believe there's disruption that will happen in both of them. I believe there's actually probably already potentially more baseline, I don't know this, this is me hypothesizing more of that already being implemented in the healthcare space for a long time than potentially being applied to that development space so far. so I feel like from a, from a disrupting perspective, that, in our, information technology arena, That there's far more opportunity for disruption, as these, to your point, ai, if they're sitting not just on the shoulder of a doctor, but sitting on the shoulder of a developer who staff they could be Yeah, exactly. Yeah. Sitting on the shoulder of a developer and, and then also taking into account what bugs, you know, what bugs came, came about afterwards. So now we can start, you know, making sure to, mitigate those issues. So taking all that data into place. Mm-hmm. So I think it would be, you know, like I said, it's a specific language model that doesn't necessarily need to,, to, for instance, in medicine, examine a patient. There's enormous amounts of data already out there that they can pull from. Yep. so there's, there's, there's a differentiation there. It's almost thinking about which of these industries are reliant on the, communication and and needing to, work with other people as opposed to someone that doesn't necessarily, I have, so we have development teams that never ever talk to customers or, and, you know, and that's not uncommon. I will tell you as the product, manager, that's something that I push against and try desperately to get them involved with. But I think that's the exception to the norm. Yeah. When you don't have to actually into, when, when interact was the word I was looking for, interact with people, all of a sudden you have it from, from the way that I'm reading the tea leaves here, a far more, ripe opportunity for AI to come in, learn and, um, and, and disrupt. So there are AI models out there in different countries that basically have a direct interaction with people. Like, for example, there are robots. I don't wanna say manning, but that's not the right gender, accurate term, what the, the manning the booth in a mall. It's not as the information provider, right. You go in there and you ask this person, they look like a person. They blink, they smile, they look at you, they answer your questions, but they're not human. Mm-hmm. That's already out there. How's that? Than anyone that works in a mall. That's true. Yeah. Nevermind. That's true. Just thinking, you know what I, just thinking about a specific healthcare example. So for instance, and I was thinking about Boston Dynamics, which I, I like love some, like, I just, I followed them in, in the things that they are doing with robots and, and how love robot dogs. Yeah, Boston Dynamics, who I followed and just am so intrigued by. Okay, so take the robot that the human, literally robot that can. See and evaluate a judge, distance, run, jump land, do flips all of those things and integrate that with AI and the knowledge that, for instance, a doctor can use. So to diagnose appendicitis, there's a place called McBurney's Point. So McBurney's point is the gold standard. If someone coming in and they have abdominal pain and you push down in this very specific area, the pushing isn't, the, isn't the important part. It's the quick release, and they have S more, they have severe pain with that quick release that is kind of this indication, this, this, this pretty clear indication that. There's a good chance we've got appendicitis going on here, right? Right. Now that needs to be done by a human. But when I think about Boston dynamics and think about AI and some of the, you know, being able to understand what symptoms are you coming in with, what symptoms do I actually need to go ahead and evaluate you for? And if appendicitis is a potential on that list, having an actual robot be able to identify Mc Bernie's point, understand how far to push in and how fast to release and measure a patient's, uh, you know, reaction to that. Then on a scale your body, like seriously, I'm not saying like, that's not tomorrow, that's not next year, but, think about how those things could be aligned and ultimately. There are ways to kind of solve some of these problems that, that the company that, you know about where, you know, they, they're solving problems like pipeline issues in whatever oil pipe pipelines or, or sewage systems. Right. You got robots going traversing through these pipes, miles and miles and miles of it. Yeah. And there's nothing wrong with the pipes, but what they're doing is they're measuring the thickness of the pipe and they're comparing to the levels that are within tolerance all the time really fast. Mm-hmm. So wherever there is even a slight indication of weakness, they would identify that and map it, they geo map it and go that. And then the next time around they say, oh, this is getting worse. Okay. And that's when today what's happening is they would relay that. You could see that. It is a, it's like people could see that on the screen and they would dispatch a vehicle out there to go fix it ahead of time before it becomes an issue. Yeah. So with ai, right. So that, that's already, now, that's been going on for a while, but with AI they wouldn't do that. What they would do is they'd be going along and going, yep, that's a problem. Spot market, geotag it, move on, and there's another robot going behind them that's gonna fix it up. Mm-hmm. So you would not even know about it. You would know about it post-event. Oh yes, that was fixed. That's the, that's the next level. Do you know what with healthcare, with that, Okay. All we are is some plumbing. So plumbing and electrics, right? So, so same exact thing when you're talking about arterial sclerosis. Osis. Yeah, same thing. So nanotechnologies be there, right? You'll swallow a pill and it would go map your arteries are, and you'll say, so utilize these blockage in your l a d over here. Right? I mean, I mean, I'm no, I'm no data scientist. But I'm just wondering about the collaboration between these mechanical type, innovations along with the data and understanding of AI and how those two put together. Could potentially from a medical, medical perspective be, really disruptive, well, well, uh, disruptive maybe, but beneficial? Yes, definitely. Because again, if all these people retire, who you gonna get to help, fill the gap. Like, you, you, you need some help. Yeah. It, it's like trying to get, uh, a contractor to come, uh, like fix up your house or whatever. In, in the heat of covid when everyone was doing like home repairs or whatever, you know, th those people were basically impossible to get on the phone. Sure. Because they had so much business that they could just look and say, well, if it's not under this amount of money, I'm just not gonna even take your call. Basically, the market will get to that point. The point here is, is it when there is nobody else left that picks up the phone, when the prices spike so high, this market will be ripe for disruption. When, when all of the people that are doing this job, age out of the job and retire and whatever, and there's basically, you can't get anyone on the phone and when you need to go see an expert, there might be a prolonged waiting time. Because there's only so many experts in the system. Mm-hmm. That, that type of market is what I'm talking about here, though. That that's the one that's gonna be targeted by one of these Silicon Valley companies who has, you know, very low empathy, extremely high ego. They don't care if they're gonna kill, it's only one per 1%. And the industry probably responded by saying like, ah, the price of progress. That's right. Exactly. It's the price of progress, dude. It's gonna be the same going forward. Yeah. I think that's exactly gonna be the same. People are gonna say it's the price you pay for, you know, Moving civilization along. Right? This is ethics in ai. This is what we're talking about now, ethics in what we're talking about. I, I do think though, for in this specific instance, what's gonna happen is yes, you're gonna have those, those, companies that are only subservient to the almighty dollar. They don't care right about these, right? There's a certain number. A lot of those companies, yes, there's a certain number of companies that will just chase that. The trick here is gonna be. To what extent will we regulate these things when it comes to, you know, not necessarily Widgets and Boston Podcast I'm talking about, uh, yeah, right? Yes. I'm talking about healthcare and other industries. Like, listen, listen. Stormy just looked at me and gave me that. Oh, like, like that, that, that Wall Street, 2008 bailout, like look the problem here is like, I'm still not a hundred percent convinced that government knows the internet exists. So, like, even though it was invented for them, right? Yeah. Right. Yeah. Well, they, I mean, they listen, they have so much money, they don't know what people are inventing for them. Sure. They have no idea. what we didn't talk about in this podcast was, creating new jobs, creating a new market, from the ashes, I guess, of the ones that you're eliminating. I was using testing at the early, at the early stages of this podcast. Yeah. If you are using testing, and I'm gonna say, well, what, what new roles are you creating in testing? Well, when you have AI models that are trained off of, off of tester activities, you still need expert testers to label those activities. So your expert testers are still, they're still around. They're, they're are, are just less jobs for them doing labeling based off the expertise they have in their career field. Yeah. This, which is the same thing as medical is like there, there are still doctors and nurses around. Sure. They're just spending, instead of spending their day talking to people, now they're spending their day doing labeling, treating the model to the machine. Right. Yeah. Yeah, it's different jobs, right? That's not the same job cuz your job changes, doesn't it? Right. Like in the testers example, right. Your job is not to now, you know, do test scenarios and execute those, right? Right. Your job is to basically arm the machine with Right nuances and whatever else, right? So yeah, you're still gonna have that, but that's not creating new jobs necessarily, is it? No. Well, well that, that particular example is just moving the existing jobs over into different areas. I mean, the, the, the three main jobs that come with, building, models are the, data scientists, the data analysts, and the engineer, right? Who is taking your data models and, and programming it into actual applications that serve an end user. Yeah. I agree with that. I don't think that those roles go away. you know, I think that in the, in the category of creating new jobs, it, it's gonna be those people that find innovative ways to adapt the models. Right, right. To different industries. Right. That doesn't exist today. Yeah. Uh, or not to a great degree anyway. Yeah. Uh, so that may happen. People training, flight systems, right? Where you can have, no human powered flight basically, or even like, like civilian airplanes not flown by pilots, right? flown by systems people arming those systems, right? so that, that's a different role in that case, right? You're not flying a plane, but you're teaching the model. I like that we've wandered into human powered flight that with an AI assist, because I think that's what it is. Like the AI might be responsible for literally everything while all of the sensors and all of the data and everything goes a hundred percent according to what the model can deal with. The second that the model gets a piece of information or data that it can't deal with, it alerts the human pilot to take over. To take over. Yeah. And, and like I, like I feel all data systems, all these systems, everything we talked about today is the exact same thing. As soon as the model gets something it can't deal with or also in our conversation today, we haven't baked nuance in this conversation. You can train the model to say if it's over 50% confidence, medical, if the model is not over 99.3% confidence, then I'm gonna hand it off to a human. To say, what do you think about this? Yeah. Here's the, here's the inputs I received, here's the outputs that I'm saying. Here's the confidence level. Here's a reason I think that confidence is off human. What do you think? Like, it, it, it extends your capability as a person. Yes. It allows you to handle more patience. Yes. It allows you to do more work. Yes. Just like moving from, industrial revolution, right. Like that, that, that thing we were talking about that we started with, but it can't above a confidence level that is baked into the application. It doesn't wanna make the decision, it wants you to look at it and double check its work. So I don't think you're getting humans out of this period. No, that's true. That's true. It's also worth noting that, you know, as it's doing that in those cases where it doesn't have a, a, a deterministic path just says, I don't know what to do with this over to you. Right. It's learning based on your decisions as a human too. Right. So if it feeds it back in, right. So over time it grows its knowledge of whatever it's doing. Yeah. I'm not sure after the session is over which side I'm on now, like I, I I wanted to start on the side of, ai it's not really a big thing. Like it's, it's, it's like a, a blip on the radar in the information age. And, uh, I feel like I, I ended up supporting it a lot more than I wanted to. Mm-hmm. You know, to say that otherwise, all, all the data scientists and data analysts and people that train the models and whatnot, like they, they're still all to a certain point trained in the industry mm-hmm. By experts. So the, those experts become less and less and less over time. That's true. And, and as, those experts decline in, , just pure numbers over time, it'll drive their salaries up. Mm-hmm. It'll drive their desirability up, , but more jobs will open up for data scientists and data analysts and data engineers and software developers and stuff like that. But, uh, anyway. We'll let, let's, let's cut to everyone for your, for your exit thoughts on this topic. So I'll say that, I still feel like we are in the infancy stages of this and that we will in the next 1, 3, 5, 10 years, look back at this time as a significantly transitional point in the manner in which, AI and machine learning, , I is integrated into our lives and I would, I would highly encourage people to truly learn about it. Do a little research. Sure. It's not hard. Go onto YouTube. Yeah. learn a little bit about how these models actually work. And I think that you, if, if you're not already learning about it, it won't be quite so scary. And, yeah. Educate, learn about it and, and jump on board because it's not going away. The wrap up for me is this, is, don't, don't be afraid of it. It's not coming out to get you first of all. It is going to be a little bit rocky initially, as Stormy said, it's gonna be rocky, but just think about this as a natural evolution of knowledge work, that that's all it is. Mm-hmm. At this stage. Mm-hmm. Does that mean it's gonna stay that way going forward? Who knows, right? I'm not gonna say that cause I don't profess to know the future, but for now, it's not going to be that way. It behooves you to kind of get familiar with it though. That's, that's a wrap for us, well done. Please subscribe and like, and smash that button down below or whatever you do. If you have a question and go to ArguingAgile.com. There's a form on the website and we'd be happy to answer your question. Live on the podcast. And it's all free of charge. And it's all free of charge. Live on the podcast. That's not totally not live at all. But it was live when it was recording. It was live at some point in that way. That's, we were alive at some point in time.