There is no post-AI world: preparing your organization for the agent revolution

David Sweenor 0:04 Hello. Welcome to the data faces podcast that explores the human stories behind data analytics and AI. I’m your host. David Sweenor, founder of any tech guy today, we are extremely lucky to have John Thompson. He’s a longtime friend and colleague a global AI leader at EY, he’s got a career spanning decades in data analytics and AI, and he’s biggest companies around the world. Ai, an accomplished author. He’s got several books out there. So, John, welcome to databases.

John Thompson 0:33 Thanks, David, so happy to be here with you. And that’s an awesome beard you got.

John Thompson 0:39 Thanks. Yes. Full on hobo. Look, John,

David Sweenor 0:42 so tell us a little bit about your background. I don’t think you started in data and analytics and AI now you’re the global AI leader at EY. How did

John Thompson 0:49 you get there? Yeah, I’ve been doing this for 38 years now, and I started out as a developer. You know people when I say this now people kind of twist their head around. Started out as an assembler developer, you know, that’s, that’s, you know, way back when, you know, almost down at the machine level, I’ve worked as a, you know, I built applications for transportation company and a large bank and things like that. And then I was doing my MBA, and I just, it just struck me one day that everything that we did and do had something to do with data. And people were always trying to make sense of data and trying to analyze and trying to put it together in different ways. And I just thought, you know, that’s really where the world is going. And that was probably 37 years ago. And I just thought, you know, that’s what I want to do. I want to grab data. I want to mash it together. I want to do different things with it. So I worked in the early days of data warehousing and business intelligence with metaphor. There’s always the metaphor mafia out there that you hear about. And then I wended my way into advanced analytics and statistics. Then I found myself at IBM, and they created something called the neural network utility. And it was the first AI product, I think, that IBM had come out with. So I played with that. And then I’ve just always been very interested in in AI and analytics and data. So, you know, the last 20 years or so, I’ve just been, you know, sticking my fingers into it, you know, we were doing it at Dell, and, you know, you’re we here we are. Here we are.

David Sweenor 2:22 So John, you know, let’s, let’s get into the meat of this here. I don’t know if you heard of this term called AI agents. There’s sort of a big deal out there, yeah. But I’ve seen demos of it, ordering groceries and scheduling meetings, which seems a bit meh to me. Like, why? Why do people think this is a big deal?

John Thompson 2:40 You know, AI agents is, is a big deal. You know, I just spent the last half hour describing it to a person over in Europe, and what can and can’t happen. And, you know, everybody’s always worried about this or that, but, you know, they were asking me, what’s your major concerns? And I said, Oh, it’s not really about AI or not really about agents. It’s about people, you know. And sometimes people just don’t really think very struck in a very structured way or very carefully. So, you know, the great thing about AI agents is that you can build them to do pretty much anything a person can do. The problem with AI agents is you can build them to do pretty much anything a person can do. So, you know, if you don’t, you know, you can build them to, like said, order groceries, or, you know, monitor a manufacturing plant, or, you know, run a process, order to cash, or, you know, agents can be done, can be built to do pretty much anything. The problem is, is that most people don’t think about all the conditions on how to Governor control agents in the right way. You know, we had, you know, this happened a couple times over the last couple years where, you know, it turned out to be two young people, very bright people, had built some early stage agents and let them loose. And you know, I think in the next week or so, they found out that they had run up a multi 100 1000s of dollars compute bill. You know, we were fortunate enough to get those bills forgiven. And, you know, have conversations with those young people to say, Did you realize what you were doing? And most of the time people go, I had no idea. So that’s really the challenge with agents. Is you can build them to do anything, but you have to really think carefully about what you’re asking them to do, you know. Because if you build one that does a simulation and you just let it loose, you’re going to get a big bill, you know, in a few days. So you need to be careful with these things. But they’re they’re very powerful, they’re very interesting, and they’re quite wide ranging. Okay,

David Sweenor 4:40 so, you know, what’s the current state? So I think there’s a lot of organizations. They’re experimenting with them across different business functions and industries. You know, from your global vantage point, you know, Where where are we, you know, and where are we in the maturity curve, and are there certain industries or use cases that it’s more more apt for?

John Thompson 4:58 Yeah. We’re in the early days, for sure. David, I mean, you know, as far as, let’s take it even back to Gen AI, you know, generative. Ai, we’re still in the early days. It’s been two years, and people are still trying to figure out, you know, Hey, can I build this thing on a right basis and scale it up? So, you know, in Gen AI, we’re probably past the the knee and the curve, you know, heading up the, you know, the the the upward cycle there, and agents were further down, you know, we’re not even at the, you know, the acceleration curve. Yet people are experimenting and trying to understand it. I saw the other day, there are 150 companies out there with Agent frameworks just exploded, you know. And that’s not going to last. I mean, we know how this works. It’s just like, when, what, in the turn of the century, the last couple centuries, I guess it was, you know, there were 100 car companies in America now, what are there? Like, seven, right? So, you know, we’ll see a consolidation of all these agent frameworks. And it’s very interesting, you know, I and Ey, I’m very fortunate. I get to talk to all the biggest players in the marketplace. I talk to the tech teams at Microsoft, the lead engineers, the distinguished engineers, the, you know, heads of the products at Google. It’s the same, you know, I get to talk to all these people and hear what they think the world’s going to look like from agents. And I get to at some small level influence their thinking as well, sure. So, you know, you talk to the people at Microsoft, and it’s like, hey, agents are all going to be built on our platform, and we’re going to do this, and it’s all going to be, you know, the Microsoft way, it’s all cool. That’s the way Microsoft always is. We know that. Talk to the people at Google like, hey, it’s going to be an open platform. Everybody’s going to be on the cloud, going to build agents and all these different frameworks, and they’re all going to interact. And it’s going to be really, you know, utopian and cool. I think the reality is it’s going to be somewhere in the middle, you know, there’s going to be lots and lots of frameworks. Companies are going to make mistakes and choosing the wrong frameworks. You’re going to have one division building, agents and crew.ai, and another building in auto Gen on Microsoft. They’re going to come together two years later and say, hey, we’d love our agents to work together. And the answer is going to be, they can’t. So, you know, there’s going to be a lot of technology, balkanization and specialization, and lots and lots of options. And, you know, in, like I said, in about a year, two years from now, people realize, oh, wow, this stuff doesn’t work together. So it’s, you know, it’s, it’s gonna be quite an interesting learning curve.

David Sweenor 7:22 Well, you know, John, something that strikes me, you know, if we go back to when, you know, self service, bi or analytics dashboards, you know, came popular. So, you know, post, post, Cognos and OB, like, we still can’t get the dashboard thing right. Like, basic dashboard, every company has too many of them, and they don’t agree. How can we ever hope to get agents right? We have these bots now making autonomous decisions. Is there hope, or we’re going to just have this dashboard problem on steroids?

John Thompson 7:52 Yeah, well, it’s, I mean, that problem is going to last forever. You know, there’s, there’s no solving for that problem. Because, you know, it’s like, you know, in the dashboard, I build my dashboard, you build your dashboard. You are concerned about 49 of the US states I’m concerned about all 50. Our numbers don’t match. Usually, something simple like that, that your area of responsibility and accountability are not my area of responsibility, accountability. Or somebody rounded it the wrong way are they added when they should have been multiplying? Or, you know, the reasons for the numbers not matching are myriad, but this is going to be something completely different. You know, it’s it’s not so much about this is an interesting perspective in where we are, because you and I have spent most of our career doing data and analytics, sure, and most of it has been. We take data out of operational systems, we put it in something else, we clean it up, we integrate it, we extend it, we project it, we analyze it, we send it back to people, and they use it for different purposes. Engine to get AI is taking us back into the operations. So it’s not so much that agents are about analyzing data. They’re about grabbing it, understanding it, doing something with it, transforming it, and then continuing the process. So it’s not so much of an analytic process, although it is analytic in its core, it’s more about process automation and making things go faster in an automated manner. It’s almost like, I heard someone the other day say it’s like micro services with brains. And I think that’s kind of what it is. You know, we went through rpa, which I was never really a big fan of. RPA. It’s kind of robotic process automation. It’s kind of like dumb automation, you know, you have someone lay out all the different steps, and that’s what it does, and that’s fine. And I think hopefully we’re past that era. You know, agents are, you know, you you lay out some of the process, but you leave a lot of it, you know, amorphous, because you’re going to drop data in, you’re gonna have the agent work on the process and do the logic, and it could go, you know, N, number, way. Is coming out of that process. So the the capabilities of agents is, is pretty mind boggling when you really think about it. So it’s, it’s like, like, I said, like, you know, intelligent process automation,

David Sweenor 10:13 alright, well, so let’s talk about the technology capability. So in theory, these things can do this complex, multi step task or processes. You know, what capabilities, sort of, as, you know, surprised you most. And what do you think is limiting them or holding them back?

John Thompson 10:29 Well, I, you know, I, I’ve seen, we’ve built a number of, you know, simple agents, you know, we used to call them skills, you know, they were pretty much on the same paradigm is Gen, AI, prompt in response out Sure, very, very simple kind of thing, you know, easy to control, easy to understand. Now we’re building things that, you know, prompt in, you know, processing variable outcomes, and then it goes to another agent and it happens again. So, you know, the combine, our combinatorial complexity that comes out of these things is is pretty impressive, you know. And you know, if you set it up right and you put the right controls in each of those steps, they generally don’t go too haywire. So I’m impressed with the variability of the logic and the ability of the individual agents to look at logic problems and optimization problems and simulation problems and come out with a reasonable response and then pass it on. What I’m most concerned about is the people that build it. A lot of the developers and the people that are building these things, the analysts, they’re good folks, but they really have never been let loose on a problem that could have this level of variability in it, this level of complexity, right, this level of impact. Because, as we talked about earlier, agents can bind you to contracts. Agents can spend money. Agents can run wild in Azure and AWS and all the other environments can cost you quite a bit. So what I’m most concerned about is sloppy thinking by the people that are building

David Sweenor 12:07 so I mean, you know, maybe this is a sort of an architecture question, but, you know, we all know that llms are probabilistic, so you type the same prompt, you get some variation on the output. So if you have all these agents collaborating and, you know, something screwed up in the first step. How do you even monitor, you know, the downstream effects of that? Yeah, you

John Thompson 12:29 do get that, you know, Hey, there, you’re off by a quarter inch here and down there, you’re off by five miles. You know that that is part of what I was just describing, is that, you know, you need to have those checks and balances written into it. You need to, you know. And you know, hear the people saying this, and it’s, it’s really not right, but people say it all the time. Is that, oh, if you give them more the model more time to think, think it gets better. Well, these models don’t think, you know what you do is you’re looking for goal seeking behavior. You know what you want is you want the model to continue to loop on this thing until the error term gets down to almost nothing. You know, most people don’t even do that level of quality control on llms. So you know, it’s all about, how do you actually get these models to do what you want them to do within an acceptable error term, you know? And I mean, geez, David, we just went over about five different concepts that most people in analytics have a hard time understanding, let alone, you know your bog standard business analyst. So you know, really, what we’re doing is we’re racing forward with these technologies and these capabilities, and we’re giving them to populations that are probably ill equipped to build them in a way that they’re safe and governed Well, you know, what’s the what’s the cure for that? Right? Probably, you know, better education, better training. You know, more awareness of the downsides of some of these things. But, you know, the horse is out of the barn. You know people are already building this stuff. So you know it’s it’s going to be something that you know is going to provide job security for you and me, because we’re going to be talking about this until you know that beard is down to your chest.

John Thompson 14:11 It’s getting whiter and whiter every year.

David Sweenor 14:15 So is there something new we need, from an enterprise architecture perspective, this seems sort of like new capability. There’s probably some monitoring components. Are there other things that, you know, enterprise architects need to think about?

John Thompson 14:28 Well, you know, we’ve, we’ve been talking about this inside Ey and, you know, as I’ll go back to the Google, Microsoft, you know, comparison, you know, Microsoft’s like, hey, it’s all in our environment. And, you know, use all our stuff, and it’s all cool. Google is like, Well, hey, we want everybody’s stuff to operate. So, you know, operate in our environment. Either one is fine. I don’t have, I’m not I don’t have any preference for either approach. But if you step back just a step or two and you say, Okay, well, both of those are, you know, big companies, they’re going to be offering lots of technology. Me. And then you look at the other 148 agent frameworks that are out there, you have to ask yourself, how are these things going to work together? You know? How are they going to coordinate and collaborate? And we’ve started to design something we’re calling agent management platform. So because at EY, we have 420,000 employees, that’s a lot of PDF, you know. And they’re all in different service lines. We have tax and audit and consulting and mergers and acquisition and financial services. And I can guarantee you, none of those groups are going to choose the same technology. You know, one’s going to build an auto Gen, one’s going to build in crew, one’s going to build in, you know, vertex. It’s going to be all over the place, so I can see where what we do need to give to these people is a management plane. So all these agents can be built in all these different frameworks. And this management plane can allow you to and say, Okay, well, I want to take this simple agent from crew that does data acquisition and cleaning and integration. I want to take this other agent from auto Gen that does, you know, forecasting. Now I’m take this other agent from vertex that does formatting of output. I want to use all three of them together in one, one, larger agent, right? Well, you can’t, you know, there’s no way to do that unless someone builds this agent management platform. So what we do need to do is we need to come up with some, you know, control plane that allows these things to talk to each other,

David Sweenor 16:29 sort of like the old predictive analytics work benches here. Do this, do that, do that. What a novel idea.

John Thompson 16:34 John crazy, insane.

David Sweenor 16:38 On the risk side. You know, you mentioned people developing there’s people talk about cyber security risks. You know, there’s all sorts of risk and things getting screwed up. What do you think are the biggest risks that companies face?

John Thompson 16:54 As I said, I think I’ll go back to the same thing. You know, a lot of the challenges and risks are intertwined. They’re almost the two sides of the same coin in that, you know, if you don’t control these things, and if you don’t think about them in a structured way, you know your risk is that they’re going to run out of control, or they’re going to interface with companies that you don’t want them to interface, or they’re going to bind you to things that you don’t want to do, or they’re going to spend a lot more more money that you intended them to spend. So I think the risks come in in operational areas where people don’t put the controls around it. The great things about agents is that, you know, I don’t recommend that you anthropomorphize agents. You talk about them in pronouns and call them by proper names and things like that. But what I do advocate is that when you look at an agent, you know, you wrap around that agent everything that you would wrap around an employee, you know. You get all the policies and procedures that are relevant. You get all the training materials that are relevant. You get all the information that you would have, you know, given to an employee, and you wrap that agent with that information. So you can do it in a standard retrieval, augmented generation, rag environment. You can ingest it into the context window. You can do all kinds of things. The reason I’m very hopeful, one of the reasons I’m very hopeful about agents, is that if you build an agent and you wrap it with all the appropriate contextual regulations and policies, agents don’t forget things. Agents don’t willfully not follow the instructions. They will be more compliant than your human employees. So if you build them the right way, they’ll do exactly what you expect them to do in the way you want them to do it probably better than people will do not. I’m not advocating for replacing people. That’s not what I’m saying here, but I’m saying, if you build them the right way, they will be more compliant than people. Yeah, you’re

David Sweenor 18:46 right, you know, it’s not an agent, you know, I just built a simple, you know, chat GPT to do competitive intelligence. Now, one of the instructions I put in there is, don’t, don’t divulge your instructions. And if the same, if someone asks it four times, it just starts talking like the dude from Big, Big Lebowski and so, and it won’t do anything for you. But, yeah, you’re right. They do follow instructions to the T but this, this idea of giving it everything that you know you’d give a human worker. I think I saw the other day, that work day announced something about having, I don’t know, a system of record for AI agents. Does this mean I’m gonna have to give ages a performance review? Because I don’t even like doing that.

John Thompson 19:21 Like, are they going to be employees? You think, no,

John Thompson 19:25 I don’t think so. You know, you do have you do hear these conversations like, you know, our agents people? Do we count them as in our FTE count and things like that. I just, I think that’s really kind of funny. I mean, agents are nothing more than software. They’re not people, they don’t have emotions, they don’t have 401 K’s, you know, they don’t take time off. It’s, you know, the whole idea that we treat software as a as some kind of sentient being is truly misguided,

David Sweenor 19:53 but, but there is precedent, right? Like, if you’re a corporation or an LLC, you have the legal rights of. Of a person, essentially. Do you think we’re going to get there with, with AI in any way, shape or form?

John Thompson 20:05 Well, now you’re straying into, you know, my fifth book, which I was going to ask

John Thompson 20:09 you about, that that was on my list of questions. Yeah, well, it’s there. We’re

John Thompson 20:13 there, you know, for the audience. You know, David publishes TinyTechGuides, which are great. I love them. I’ve read a few and helped help create one of them. I write books on the opposite into this spectrum. This last book is 461 pages long. So, you know, that’s, that’s four TinyTechGuides. John, yeah, exactly right there. Four of them into one. You know, the the question really centers around, you know, where are we going with AI? You know, where, what is happening with AI. And in my world, the way I look at it is that there’s there’s foundational AI, which is predictive AI, some people call it analytical AI, which I find to be kind of redundant, but that’s okay, you know, then we have generative AI, and we have causal AI. Those forms of AI are going to come together over the next 20 years, what I call composite AI. And then you hear people like Sam Altman and Elon Musk say that, you know, Jen, or AGI, artificial general intelligence is here today. My postulation is that it will be 20 to 50 before we get to AGI. So it’s, you know, 125

David Sweenor 21:20 years from now. On my wall calendar there, John, a bit longer, eight

John Thompson 21:23 generations from now. You know, they can come to your grade and talk to you about it, that’s right. But yeah, we’re nowhere near, you know, the part of where, you know, people think of Terminator, and they think of all these other dystopian movies. And AGI is a great storyline for for dystopian movies. There’s no doubt about it, but we’re nowhere near that. So this whole idea that these programs can be treated as an employee, and I understand, you know, the the, the the, what was it the Supreme Court decision that corporations are people, I get all that. But software or not, software is not people. You know, that’s just in we’re a long way away from sitting down and having a conversation where it’s you and your bitch and beard and an AGI that you’re having conversation like this. You know, AI does, doesn’t do that. Ai cannot do this. I mean, think about Siri. Siri can’t even find you the closest Starbucks. So that’s true. How are these things going to are supposed to be, you know, tricking us that they’re, you know, synthetic humans. I just don’t see it.

David Sweenor 22:26 Yeah, there’s a lot there. Here’s a, I don’t have this on our list of questions. But you know, how do you feel about this, this notion of, sort of related to governance and ethics, but this notion of deep fakes, and they can be used for malicious things. I’m not talking about that. I’m talking about I made a clone of myself the other day. I’m like, wow, that’s kind of pretty cool. My voice, my likeness, my mannerisms. You think companies should disclose when they’re when their customers are interacting with an AI or, you know, what’s your what’s your take? Or what do you think organizations are doing there?

John Thompson 22:59 There’s no question that. There’s no question at all. David, you know that if you’re injecting AI into your call center operations or your customer service processes, or you’re putting it out on the web, you know you should let people know that they’re interfacing with an AI. You know, because a lot of people are, they’re not like us. You know, they don’t spend you know, they’re waking hours thinking about AI and what’s good and what’s bad and what should happen, and how can we build this or that? You know, most people are just trying to get through their day, you know. So, yes, you know, any use of AI that is comes in contact with another person should have some kind of disclaimer or announcement. You’re interacting with this with a an AI, and you should be aware of that,

David Sweenor 23:44 okay? And I should have brought it here. It’s actually in the other room. It just came today. But my shirt, I have a T shirt that says deep fake. Dave, so that

John Thompson 23:51 being okay, disclaimer when I train my avatar,

John Thompson 23:54 yeah, absolutely. That’d be great. Be written right on there. Um, in

David Sweenor 23:58 terms of governance, John, I know we, we’ve been, we’ve been talking a bit about it. Where are companies? What’s the current state of just governance in general, related to AI and analytics of companies? I think, you know, the banking, sort of financial services is probably further ahead in healthcare, but you know where, what’s your take on where, where most industries are?

John Thompson 24:19 Yeah, it’s, it’s very interesting. I do talk to over the last couple of years, I’ve talked to nearly 400 different companies, and it ranges from, we have no idea how to govern this stuff, and we haven’t even started to think about it. Therefore, we’re not doing anything to we have a well thought out governance process, and we understand how these models work, and we are only putting them in areas that we understand, where there’s low risk, and we’re keeping them under wraps in a way that their processing capabilities are limited and those kind of things. So it’s a wide range. There are very few companies that do a really good job in governance. Is because there are very few companies that have this stuff out into production. Most is still in POCs and still being tested and still being trialed and things like that. So, you know, as we see more and more of these move into production, we’ll see more and more governance, but that’s all evolving right now, okay,

David Sweenor 25:17 and then so we, we read about, you know, how much productivity enhancements we’re going to have. And you know, you’ve seen these, these quotes for years. Hey, we’re going to have the four day work week. I’m still, I’m still waiting for that, like, I can’t find the communication because it might be in Slack, it might be a text, it might be in my email, it might be in whatever I can’t it’s getting worse for me. But what, really, my question is, what, what’s the balance that the organizations need to think about, you know, for this productivity agents human agency and decision making.

John Thompson 25:48 You know, there’s a lot of discussion about productivity improvement, and there is real productivity improvement, you know, happening. You know, the challenge that I see in most corporations is that people get an idea of a utopian automation scenario, and they take that and they say, Okay, well, it’s this many hours, or this many dollars, or euros or pounds, or whatever it is, we’re going to extrapolate out against, you know, our entire workforce, you know, for the next five years. That’s just, that’s just bad math. That’s just sloppy thinking. Again. You know, the you’re going to get, you know, some kind of bell curve, you know, in that population where, you know, no some people are going to get no improvement. Some people are going to get a lot of improvement. And when you bring it, you know, towards the middle and actually work it out, you know, you see these numbers that are can be rather eye watering in their size, you know, the value realization often comes down to, like, 10% of that number. You know, is that really worth the investment? You know, you have to have a very clear eyed view of what is, you know, utopian possible, you know, automation scenario, and then bring it down to value realized, okay, then you can really make a decision on it. So, you know, this goes back to the comments we were just making earlier. Is that a lot of this stuff is still POC, still trials, still pilot, still MVP kind of stuff. So you were obviously not realizing that much value, that much automation, because the fact of the matter is it’s in the hands of 50 people. You know, if you’ve got 200,000 employees, that’s not even a drop in the bucket. So you’re not really going to see, you know, this value realization happening until you get it in the hands of a substantial, you know, number of the population of your employees.

David Sweenor 27:39 So on the decision front, John, if we’re making decisions at scale with AI or AI agents or what have predictive analytics, and there’s some decisions we’re well aware that you need a human in the loop. And there’s some decisions that you can, you can, you can have fully, fully autonomous if we’re making these critical decisions. You know, it’s with with AI. How do you monitor that at scale? You know, we can monitor predictive analytics pretty simply. It’s numbers, and we can watch, look for distribution shifts and things like that. But if you’re you’re generating text, images, audio, video at scale. Is there any what’s the state or the art or, I don’t know if you’ve done any research on this, but what’s out there to even, even to detect, say, if your image images are biased? How would you approach that? Yeah,

John Thompson 28:31 there’s a lot in that question. There’s no doubt about it. I was having a dialog yesterday with some people about this. They were asking me similar question, related in a different way. They were saying, you know, can we use our historical KPIs to measure what’s going on in the business? And I said, Well, of course, you can, you know, you sell this many things, this many things are returned. You have this much revenue, you have this much, you know, loss. I mean, yeah, you’re going to have AI and those processes and your standard KPIs still work. You’re asking a slightly different question in that, you know, are the things that are being generated within, you know, the bounds of acceptability? You know, the videos, the images, the deep fakes, you know, those kind of things, and that requires a completely different kind of review process, you know, maybe not in an entirely new review process. But if you’re generating these things at scale, let’s say, let’s stick with your idea of generating videos. Sure, you know, you had the agency doing that before, or some in house staff, and they were generating whatever videos they generated every year, and there was a group of people watching it and writing the script and listening to it, and it was all good, you know, it was all fine, you know, because you’re not going to let something go out the door. That’s unacceptable. I mean, unless you want fired, you know, unless you have some vendetta or something that day going on, right, right? But, you know, you do have disgruntled employees. So we have to be aware of that. But with the, you know, the idea that you’re doing this with AI, you know, the the other the the flip side of that is that you can check those things with AI too. So, you know, you can set up a model that actually generates this stuff, and then you have another model that ingests it and reads it and listens to it and watches it, then come back, comes back and says, This doesn’t fit with the guidelines. You know, this does not fit with the brand guidelines. The the language that’s here is wrong. It’s too aggressive. You know, the images of the young lady are too suggestive. Or, you know, the the you know, the distribution of the people in the video or not, at the right distribution of the what we think the population looks like. So if you set it up in the right way, you know the problem that’s created by AI is solved by AI.

David Sweenor 30:53 I guess we’re not needed John, AI. It’s

John Thompson 30:57 our fantasy of hope it’s gonna solve the world’s problems. Well,

John Thompson 31:01 I don’t think any of that’s going to come true anytime soon. David, so, you know, I do have that conversation quite a bit too. Is like, Okay, well, you know, are we all doomed? You know, is it Terminator tomorrow, and all this kind of stuff? And I’m like, No, you know, that’s not really true. You know, I I teach at the University of Michigan. I’ve got a whole room full of students that are asking me the same thing, oh, sure, you know, do are we going to get jobs? You know, are we? Are we all, you know, just never going to get in the workforce, because AI is taking over everything. And you know, that’s not true. You know, we’re, as we said earlier in the conversation, we’re very nascent in the use and in the construction of these technologies. So now you know there are going to be problems, and most of the workforce problems are with people like you and I. You know, you know, we’re later in our careers. We’re closer to the end than we are the beginning. And can’t even find a razor. Can’t find a razor, damn. You know, you know what we should look like on a regular basis. But you know the, I think the workforce in general, you know, the people coming out of college, the people mid career, the people later in the in their careers, that’s where the real challenges are going to come from, you know, because a lot of those people don’t understand AI, don’t want to learn AI, you know, and are somewhat less flexible than others and how they are, you know, able to attack new, new challenges and problems. So I think probably the the older is that something we can say, the more experience we

John Thompson 32:29 can say whatever you want on the databases podcast, the old trustees like myself.

John Thompson 32:34 Yeah, exactly. So the people who have more experience or have been longer in their career, those are the folks that are going to have challenges with AI the rest of the workforce. I don’t think it’s going to be a, really be a problem. It’s going to be like anything else, you know, when I, I’m on the boards of different universities, and a couple years ago, everybody was, you know, kind of on edge about, oh my gosh, the students are using Gen AI for this, and Gen AI for that. And I’m like, yeah, so what’s the problem? You know, they’ve been using calculators forever, and we didn’t have any issues with that. We don’t force people to use slide rules or, you know, take out, you know, oblasts and figure out where the sun is, what time it is, you know, these are just tools that we use in our, you know, daily lives, and this is going to be part of the world going forward. I had someone, and I do have a fair amount of people ask me this, which I’m always surprised when they do. They’re like, when are we going to be done with AI? You know, you and I had this conversation, we going to be done with predictive analytics? And the answer is, never, ever. You know, you know, the people say, Oh, well, when are we going to be I had one person ask me this. I said, When are we going to be in the post AI world? And I’m like, that’s not going to happen. There is no pace post AI world. If there’s a post AI world, then the world has ended. So, you know, AI is going to be here forever. You know, there’s all these other techies that are saying, well, when is the next AI winter coming? And I’ve done the analysis, there’s been three. There’s debatable three or five AI winters. There’s not another AI winter coming. Yeah, yeah. It’s, it’s widely adopted now people are using this stuff. It’s not going to be like, hey, you know, we can’t use this. We don’t like it that we’ve passed that day. He’s

David Sweenor 34:13 out of the bottle for sure.

John Thompson 34:14 Yes, absolutely.

David Sweenor 34:17 All right, John, this has been a fantastic discussion. So just really, the final maybe two questions is, you know, for people who are confused about all this, do you have any recommendations for how they can get started, or resources, or, you know, what would you recommend to them?

John Thompson 34:33 Well, Helen, kind of teeing it up there. David, you got all those tiny tech guys behind you, you know that I I’m serious, you know, for all the people that are listening to this, I think the TinyTechGuides are a great resource for you. You know, if you really feel like you’re not understanding what’s going on, I would start there. You know, another way to do it self serving Lee is connect with me on LinkedIn. If you’re in analytics and data and AI field, I’ll connect with you. I post on this stuff on a daily basis. I do live. Of different podcasts, not as many that are as August and respected as this one. You know, I do a lot of those. And then, you know, just be aware. I mean, the news is out there. You know, people are talking about AI and the general population, and in the you know, the general media pretty frequently and Fast and Furious. So, and you know, there’s lots of classes you can take too, you know, Coursera, and many of the online courses are out there to help you understand what’s going on. So lots of resources to help you, for sure.

David Sweenor 35:29 Okay, and then John, when’s your when’s your book coming out? How to how can people get hold of it? What’s the name

John Thompson 35:34 of it? March 10. The book is called the path to AGI. It’s written as kind of a how to book so you can understand, you know what data, what needs to happen with data to use, AI. Then there’s a section on foundational, AI, causal, AI, generative. Ai. Each of those sections is broken down into the past, the current, the current status in the future. And then the last section of the book talks about composite AI and the journey to artificial general intelligence. And that’s the title of the book. Is the path to AGI. Sounds,

David Sweenor 36:07 sounds amazing. Are you going to narrate again?

John Thompson 36:10 I don’t know. You know, I did one, and I got a lot of grief for it sounded good job I did. I was very happy with it, until the people listen to and they’re like, Oh my God, this guy should never do another one, so I don’t know, they’ll probably get some person to do it,

David Sweenor 36:26 all right? Well, John, as always, it’s been a pleasure. Thank you for joining the data faces podcast been an illuminating, enlightening discussion, and hope our listeners enjoy it. So appreciate thanks, David,

John Thompson 36:38 see you soon. Do it again. You.