Will a Robot Take Your Job? with Tom Taulli, Author of The Robotic Process Automation Handbook

(Intro Narrator at 00:00:02) Hello, my friends. Today, Joel is talking to Tom, author of the books Artificial Intelligence Basics and The Robotic Process Automation Handbook. And they discuss what will happen when jobs are automated away at massive scale, and why many advanced professions may not be safe from automation. All of this right here, right now, on the Modern CTO Podcast.

(Joel Beasley at 00:00:32) Here we go. This is the Modern CTO Podcast. What is your backstory? Like, why did you decide to write this book?

(Tom Taulli at 00:00:49) Yeah. Yeah. The irony here is that we have pretty much almost exact backgrounds, except I'm a little older than you, probably a lot older than you. But my background is similar, but my dad was not an engineer. He was really into technology just as a passion. So he read a lot of science fiction and just loved technology. And this is back in the early eighties. So he bought a comp— I didn't even know he was gonna buy me a computer. He bought me a computer and just put it on my desk and then just walked out of my room. And that was his way of saying, you know, you figure it out.

(Tom Taulli at 00:01:24) So I looked at it for a couple days, you know, just staring at it, thinking what should I do? And so I just started using it and kind of like you, you know, started simple, learning the basic— you know, no one really talks about the basic computer language anymore. I learned the basic computer language, started creating my own programs. And then from there, I learned Pascal. That was a really popular language back in the eighties.

(Tom Taulli at 00:01:47) And then C, this is before Internet. So the way I made money is there were computer magazines that would have the code in the magazines itself printed in the pages. So I would write programs and publish them in these magazines and make money that way in high school just like you made money in high school. And then when I got into college, I was never really good at taking exams. So I thought about creating— but I saw students come in the classroom with computers.

(Tom Taulli at 00:02:17) And I thought, well, probably a lot of other students are not good at exams. So maybe I can create software to help them take and be better at exams. And so that's what I did. When Windows came out, I developed Windows software for that. Ironically, one of the first exams I developed for my software for was for the real estate exam.

(Tom Taulli at 00:02:37) So people in the real estate industry could get licensed quicker and then get that process going so they could make money faster. Then that company evolved into the Internet came around and that company evolved. And I started another company that was kind of like a Shopify for 1997 where you can build your own website and do some fun things on it. And we sold that to a company called Infospace and kind of rode the wave for the rest of the nineties. But along the way, I continued my writing, and I started writing for forbes.com and different other publications.

(Tom Taulli at 00:03:10) And so I was kind of always, you know, talking to a lot— just like you talking to a lot of interesting people. We didn't have podcasts in those days, but, you know, we had— I wrote for the online portion of Forbes and in those days they didn't even want to talk to me because they thought that the magazine was better. And, but I still got to talk to a lot of interesting people. And so the way the, you know, I've written different books over the years. Probably the one that I'm most associated with is the AI Basics book and that came about a lot of it had to do with forbes.com because I was talking to a lot of founders, and AI just kept coming up over and over again. And I thought, you know, there's probably a lot of people that wanna know what this is about, but they don't have a technical background.

(Tom Taulli at 00:03:57) So my idea was to write a book where you can learn about deep learning, but not have to, you know, download TensorFlow and, you know, have an engineering background to know what it's all about. So that's kind of a little winded way of saying, you know, I kind of start off self taught like you, different startups, but I did some writing along the way and education along the way and came on this idea for the AI book. And AI has always been very interesting, you know, to me back when I started creating my computer programs in the eighties. You know, that there's a lot of movies about AI, and I thought it was really cool. But there was no way I could in those days— they didn't have TensorFlow or open source or any of those types of things where you could or even data that you could create these small models.

(Tom Taulli at 00:04:44) So it really wasn't practical. So, that's why I focused on these other areas, you know, like the exam prep, Internet, and those types of things.

(Joel Beasley at 00:04:54) What other books have you written?

(Tom Taulli at 00:04:56) Another book on robotic process automation or RPA. It's a technology that automates the user interface, say, ERP systems or CRM systems. You know, the irony here is that companies have spent so much money on technology, and they create their own workflows that can just bog down and get, you know, repetitive. So, you know, you go into a typical company, and there's always employees doing cut and paste all day long, you know, or, you know, taking a look at a PDF and, you know, doing a couple functions on it, but the same functions over and over again. RPA is— you can create these bots that can automate, kind of like macros, that can automate those processes.

(Tom Taulli at 00:05:45) A lot of people confuse us with with AI. It really isn't AI. But companies like UiPath and Automation Anywhere, the leaders have been able to leverage data to identify these processes that are not as optimal and provide suggestions of better processes and can create bots for you. You don't have to create your own bots. Yeah.

(Tom Taulli at 00:06:06) So there's a lot of innovation occurring. It's one of the fastest growing areas in enterprise software, especially in the last few years with with COVID. Companies have had little choice but to automate these systems because they have these huge spikes in activity. And so, they just couldn't hire people fast enough to handle the volume. So they started relying on this RPA technology.

(Tom Taulli at 00:06:32) And UiPath is the leader in this space. So my book on RPA goes through the, you know, the industry, what the technology can do, how to build a bot, you know, how to implement, how to manage, and things of that sort. So that's another book. Yeah.

(Joel Beasley at 00:06:51) So for your for your AI book, what is the audience like the just the average person or investor, like or who?

(Tom Taulli at 00:07:01) Yeah. Well, and the funny thing is, you know, I would talk to investors, VCs, and a lot of times I listen to them, and they weren't quite right about certain things about AI themselves and even founders of companies. You know? And then, you know, there's probably a million AI companies out there because 999 thousand of them, you know, just claim their AI and they're really not AI. And it's—

(Joel Beasley at 00:07:24) So you—

(Tom Taulli at 00:07:24) Give them the money.

(Joel Beasley at 00:07:26) Exactly. You gotta pitch it. You know? I mean, I used to be in the trenches where I had to pitch my ventures, and you gotta you gotta tell them what they wanna hear. And even if you don't have it, you know, I mean, are they really gonna know?

(Tom Taulli at 00:07:39) You know? So, you know, you gotta fake it till you make it. So, there's a lot of that. So I see a lot of that. So, yeah, it could be for a founder. It can just be for someone who works at a technology company, is curious, PR person. You know, could be someone in HR. You know? They're evaluating software for their, you know, to handle the onboarding and talent management. Some really tricky issues with that with privacy and things that can go wrong and, you know, discrimination and things of that sort.

(Tom Taulli at 00:08:11) So, you know, you don't expect your HR person to have a data science background and things of that sort. But, you know, with this book, at least they get the general understanding of what this technology can do, some of the risk factors, some of the things we anticipate, you know, how to implement this technology, or how not to spend a lot of money on something that's not gonna get you anywhere.

(Joel Beasley at 00:08:35) Yeah. And I read the book. It does a great job, like, just to cover a couple of— you've got data, machine learning, deep learning, you know, a little bit. One chapter on RPA, natural language processing, robots, implementation, all of those things. So when I was looking at the book, I was like, oh, you know, I know about all this stuff. I wonder what this book's gonna be like. Is this gonna be a boring book? Is it gonna be interesting? And then I was thinking about, like, how it would bring value to the audience. Right?

(Joel Beasley at 00:09:02) Because we're mostly CTOs, VP of Engineering, it's next generation tech managers and leaders. And one of the things I came up with was it's really, it's really good because it provides you— like the way you succinctly describe and explain things is better than the explanation I have in my head to give to a non technical person. So it's like education for me on like, you know, if you want to explain it to someone who's new at it, use Tom's words.

(Tom Taulli at 00:09:33) Yeah. Well, you know, I've been at it for thirty years of writing, taking technical topics and making them understandable, you know, to those people that don't have that background. You know? So, I mean, I was doing that in high school with these computer programs and doing it while I was in college with Forbes and things like that. So a lot of experience of that and just trying to be in the shoes because, you know, sometimes you really know that you you wanna just throw a lot of information out there, you know, and that's usually not the best approach.

(Tom Taulli at 00:10:05) You know, with this book, are you gonna learn everything about deep learning? Absolutely. Just scratching the surface. But you know what? It's actually just helpful to know what the difference is between machine learning and deep learning is because I read a lot of articles, and it's pretty clear a lot of people don't understand what that, you know, the difference is. So—

(Joel Beasley at 00:10:27) Can you can you tell me what that difference is?

(Tom Taulli at 00:10:30) Yeah. Yeah. So, think— you look at circles. This I use the circles analogy. So artificial intelligence is a— is the circle the biggest circle. It encompasses everything that has to do with a machine acting like a human with intelligence, taking data and coming up with insights. You know? So I mean, Siri is obviously kind of artificial intelligence. It takes in data. You know, it takes in words.

(Tom Taulli at 00:11:03) It analyzes those words, and it, you know, parrots back something and answer for you. Right? So, yeah, it's pretty basic idea. So that's the big circle's AI. There's another circle called machine learning, and machine learning is more about using traditional statistics to come up with conclusions.

(Tom Taulli at 00:11:27) So if you went to, you know, in high school, regression analysis, you know, where you would look at a couple of variables and find the correlations of those variables and what kind of outcomes. So, you know, if you have a certain number of bedrooms and a swimming pool, and then you live in this neighborhood, your house will be worth roughly this amount of money. That's a regression. That's machine learning. So then what is deep learning?

(Tom Taulli at 00:11:51) Well, deep learning is a circle within the machine learning. It's a smaller part of machine learning. So it's a circle there. And what is it? It's very similar. It takes in data. It analyzes that data and comes up with the conclusion. The thing is it's way much more sophisticated. So you'll have these things like hidden layers. There'll be a lot of these hidden layers, and it's taking in this information and going through these hidden layers and allocating certain weights to this data.

(Tom Taulli at 00:12:25) And then it tests it and saying, is that accurate or not? If it's not accurate, it'll then go right back to where it began and then try again and find— and it's just going through that and just— and these models are just huge. I mean, Microsoft has these models that are, like, a hundred billion parameters that just crunch numbers as deep learning. So it's just so much more sophisticated in terms of the analysis. So that's— so in a way, they're similar, but it's more of a difference in the huge degree and differences in terms of the amount of data that's being processed and how it's being processed.

(Tom Taulli at 00:13:03) And then there you'll hear things like backpropagation and all these different really complicated concepts. And that's— you don't have that machine learning. Again, machine learning, a lot of times, kind of what you learned in high school. Yeah. It's more sophist— yeah. Obviously, more sophisticated.

(Tom Taulli at 00:13:18) But this basic statistics, machine learning just goes to to a whole another level of analysis. And that's where you need these hugely sophisticated, you know, GPUs, the sophisticated hardware systems that can process just a large amount of information. Humans can't do it. And a lot of times, it comes up with patterns that humans would never even think of. And that's where the real power comes from this AI.

(Tom Taulli at 00:13:45) I mean, it tells you something you never— you know, it comes up with conclusions that, you know, you couldn't figure out. So it could be something like you have a oil rig, and it'll tell you, you know, a certain part is going to fail within the next week. It may be a $50 part, but if your rig fails because of a $50 part and it fails for, like, twenty minutes, or you have to take a helicopter and that costs you $30,000 to get a helicopter to get that $50 part and bring it back on the rig, that's a big deal. So those kinds of deep learning systems can be hugely beneficial to companies. And that's where in the last ten years, we've seen so much of the innovation in AI is in deep learning.

(Tom Taulli at 00:14:30) And we've had that because we have the GPUs, Cloud computing, access to data, and then also all these sophisticated theoretical concepts that have come from these geniuses like Geoffrey Hinton and all these brilliant people from Facebook or Google who've come up with a lot of the theoretical analysis for that. So the last ten years has been really a new age for for AI, and largely that has been due to deep learning.

(Joel Beasley at 00:14:57) And so when are they gonna take all of our jobs?

(Tom Taulli at 00:15:03) Yeah. I got a really deep discussion about this with the— well, actually, my first in person presentation was two weeks ago. And I did a presentation on AI, and I brought this topic up, and we had a really deep conversation. I, you know, you talk to companies and they'll say, oh, it's— you know, AI is not gonna take jobs. It's going to enhance our jobs. It's gonna make our jobs better. Or it's gonna take— you know, it's gonna take things that we do. It's repetitive and tedious, and we don't wanna deal with that. We humans wanna do sophisticated things. The problem is this AI is gonna do more and more human judgment type abilities.

(Tom Taulli at 00:15:44) I mean, driving a car is, you know, that takes a lot of intelligence. I mean, that's not something that's, you know, tedious. I mean, that that's, takes some skill to do that. You know? So I think that I think the next ten year— you know, right now, we have a job, a labor shortage.

(Tom Taulli at 00:16:05) But I think in ten years, I think there'll be some professions that will be automated away. I mean, I think like like a CPA, even an attorney, you know— I mean, I talked to the CEO at Automation Anywhere, and, I mean, he told me, like, in five or six years, he thinks that, you know, the AI will be sophisticated enough to understand any contract. You know? So, you know, I do think that, and then the thing is that, you know, people say that, well, you know, if it gets automated away, you know, new opportunities will come up. But how do we know that's going to happen?

(Tom Taulli at 00:16:48) So I think we're in for some interesting changes, and, you know, your career path may be radically changed because of AI maybe within the next decade or so, because of that.

(Joel Beasley at 00:17:02) Yeah. Well, I mean, I fully agree. But that's been happening since the beginning of time. Right? The jobs we have today are not the jobs that we had a thousand years ago.

(Joel Beasley at 00:17:11) So I like both sides of the argument, to be honest with you. I think that's the most fun because I agree with parts on both sides. And when I started to research this, it led me to a man named Ray Dalio.

(Tom Taulli at 00:17:28) Yeah, sure.

(Joel Beasley at 00:17:29) He talks—

(Tom Taulli at 00:17:29) Bridgewater Capital.

(Joel Beasley at 00:17:30) Yeah. He was the one that taught me about how the economy works and gave me a real good mental image of how things work. And so I was trying to sort of reconcile these two ideas of, okay, let's say that we automate everything. Like, we automate everything.

(Joel Beasley at 00:17:48) Even all the human stuff. Like, there's machines making music and stuff. Well, the way that currency works is it's an exchange between humans for agreed upon value. Right? So we're not trading money with chipmunks. They don't care about our dollars, right?

(Joel Beasley at 00:18:06) So we're exchanging things of value with each other, and that's what the economy essentially is. And we will—our values change. The things we value change over time, and I think they'll always change. And so let's say that everything was automated. Well, I don't know what it'll look like exactly, but there will still be people trading things for other things. And even if 99% of all humans are some various form of artist, right, you would want to pay for the authenticity, right? And also, we always tend to leave out the fact that our population grows and will become interplanetary and will expand, and there's more people to trade with. And so I think there is a lot of pain along the way.

(Joel Beasley at 00:19:00) I think we're more connected than ever, and people now can talk about that pain, and it's more clear that it's going to happen, and there's actually things we could try to do about it. So it's definitely taking a lot of our attention, but I am an optimist long term. I think it'll all work out, but yeah, there's going to be a lot of people displaced along the way.

(Tom Taulli at 00:19:23) Yeah. And, you know, I think that's—retraining, reskilling are very important. We've seen that already with IT. You know, with a huge shortage in qualified developers and coders, I mean, you've seen the emergence of these online platforms, these schools, these boot camps to get people trained for that. So I think that's an important part of that. If we can get better at retraining people.

(Tom Taulli at 00:19:55) But yeah, change is always not fun for a lot of people. For the entrepreneurs, you know, disrupting the industry is fun. But if you're the one being disrupted, it's really not a fun place to be.

(Joel Beasley at 00:20:07) Yeah. Oh, tell me. My father-in-law is about to retire. He's, for 30 years, driven semi trucks for UPS, right?

(Joel Beasley at 00:20:17) And, you know, I've been with my wife for about eight years now, and every year I was giving him updates on the automation of the semi truck industry. And he was telling me the whole time, it's not possible. It's not possible. There's so much to it. It's very complex. It is a hard test. You know, because you wrote the real estate test—the trucking test is actually a very difficult test.

(Joel Beasley at 00:20:39) It's right up there with, like, real estate's really tough, insurance is tough, trucking stuff. But I was, you know, every holiday meal, we're sitting down, and I'm just showing him the updates until, I think two years ago, they actually made the first delivery with an autonomous semi truck. And he's like, it's not possible. I'm like, well, dude, they did it. Here's a video. Here's a video of it happening. And, you know, one of the things that made me think about was, like, there's this—in different industries and different jobs, there's this weird undercurrent that we don't have as much of in technology.

(Joel Beasley at 00:21:20) So in technology, I'm self-taught, you're self-taught, right? So we understand that, like, okay, there's a new language, there's a new framework, there's a new style. I've got to learn these different skills. We have to—you know, things change, new technology comes out, we have to learn new things. But for somebody who's been driving a semi truck, I mean, very little has changed in the semi trucks. Like, you get really small changes over time. You might get a new brake system this year and a new navigation system. So they're like these really small changes not requiring them to unload huge amounts of mental effort.

(Joel Beasley at 00:21:53) So there's this feeling in a lot of industries that, like, I should just have to learn how to do my one job, and I should have this right to do that job forever. And so I've been trying to better understand that because that's contrary to how the free market works. Like, or how humans progress, even in non-free markets. How humans progress. That's just not how it works.

(Tom Taulli at 00:22:14) Yeah, yeah. Well, when you get older, you kind of get set in your ways, you know? And then you complain about, you know, or, you know, like you said, you know, oh, it's not going to work.

(Tom Taulli at 00:22:30) And as you get older, you get kind of into your own echo chamber, you know? It's hard to change, and it's hard to see things coming at you until it's too late. And even then, you're kind of dragged, kicking and screaming, and saying that it hasn't changed, you know? But there is so much money being plowed and poured into these technologies. I mean, Amazon is spending huge amounts of money to automate everything.

(Tom Taulli at 00:22:57) You know, Google is doing the same thing. Microsoft is doing it. You know, it's just only a matter of time.

(Tom Taulli at 00:23:05) I mean, this can all be replicated. I mean, driving a truck is probably easier than automating a regular car, except when it comes to—you know, there's things where, you know, making those turns and, you know, finding those—you know, getting—you know, backing up the truck and things like that. But a computer probably could do better at that at some point than a human. So, you know, yeah. I mean, and some of these industries, like trucking.

(Tom Taulli at 00:23:38) I mean, trucking is millions of jobs. It's a huge industry. People's families depend on that. And it could be very much an overnight thing because, one, wages are going up. So what's popular for automation is, you know, is it high-cost labor?

(Tom Taulli at 00:23:57) Is it high-cost labor? Well, trucking really is. So there's a lot of incentive to replace that labor. So if companies have an opportunity to do it, they're going to do it because they can justify the ROI. There's some areas where it's tougher.

(Tom Taulli at 00:24:14) You know, because I know there's robots—I've seen robots in fast food companies and so forth. You know, now a couple years ago, it didn't make a lot of sense because the wages were pretty low. But now, you know, it's like $20, $25 an hour to hire someone at McDonald's.

(Tom Taulli at 00:24:33) So it's getting to the point where maybe that robot does make a lot of sense to bring in. It's all there. I mean, the technology is all there to flip burgers.

(Joel Beasley at 00:24:43) And the cost is going down with technology.

(Tom Taulli at 00:24:44) And the cost is always going down too. Yeah, yeah. So that cost curve is always going down, and then wages usually don't go down. They usually go up.

(Tom Taulli at 00:24:53) So there's a point where it just makes so much economic sense for these companies to buy this technology. It was always tough because, how do I justify it? You know, am I going to spend $100,000 on this robot? You know, how long is that going to take? You know, I don't—yeah. A lot of times, they just don't want to make those upfront expenses, but it's getting to the point where it's starting to make sense to do that. And that's going to drive—

(Tom Taulli at 00:25:18) It's going to accelerate the change, and it's going to change the world in a big way. And I don't think we've really thought about it. Yeah.

(Joel Beasley at 00:25:28) I think about it a lot. That's why I have this podcast. I was like, I need relationships at scale. I need to know a lot of people. I need to know the smartest people, and then that way it'll position me best for wherever, whichever way it goes. So relationships are vital, man. I was just—I was talking with Kyle yesterday. He's not the CTO of Verizon anymore. He's, like, executive VP of technology innovation. I don't know, some long title. But he's a real bright guy, was an engineer, and he's super smart. But, you know, we were talking about the future and technology and all of this stuff, and we didn't really get into automation.

(Joel Beasley at 00:26:06) But one of the reasons that I was excited to talk with you is that, you know, sometimes when I get to talk with these more executive people, they won't be able to speak as freely about certain things because they've got PR teams and whatnot, which I understand. But if I were to make you, like, leader of the free world, how would you even begin to approach, like, the questions of people saying, like, what are we going to do about automation?

(Tom Taulli at 00:26:32) Well, first thing is you can't really do anything about it because it's just going to happen. It's happening already, you know? Trying to stop the train is not going to be helpful because companies are going to find a way to—and we live in a free market system. So if they want to buy a robot to replace somebody, they're going to do that.

(Tom Taulli at 00:26:57) If it's cost-effective for them to do it, they certainly will do that. And I agree that, you know, you mentioned about the PR thing. In my book, I have a story about IBM. Because in the early days, IBM was investing heavily in AI. And then there emerged this talk in the '50s about replacing jobs, and IBM wanted nothing to do with that.

(Tom Taulli at 00:27:22) So they started backing off. And most of the research in AI in the '50s and the '60s and even the '70s was primarily from government institutions and educational institutions, not companies, because they didn't want to be associated with that because of the bad PR. So, yeah, so if a company tells you, oh yeah, this is—you know, don't worry, well, then you probably should worry.

(Tom Taulli at 00:27:46) And it's going to start automating jobs and jobs that you don't think would be automated. So we always think it's, you know, again, the flipping of the hamburgers or things like that. But, you know, I talked to someone who's pretty smart with technology, and, you know, one thought experiment was if you had a nurse and you had a doctor, which one would be automated first? And we thought that probably the doctor, not the nurse. Because there's certain skills that the nurse has that computers probably will have a hard time—you know, bedside manner, you know, just dealing with people, things of that sort.

(Tom Taulli at 00:28:25) And some people just want to deal with people. And for those kinds of things, they don't want to deal with a doctor. But, you know, a lot of times, what does the doctor do? They ask you questions, they give you tests, and then they prescribe you something. It's pretty much the typical workflow.

(Tom Taulli at 00:28:42) Now it's a lot different if, you know, heart surgery, cancer treatments, and all that. But I think for a general medical, which takes up a lot of our costs anyway, I mean, it's pretty straightforward for medicine. And we've already seen this, you know, to some extent, with the use of, like, Teladoc, you know, through the COVID. You don't have to go to a doctor's office. You know, there's these ways of remote treatment, things like that.

(Tom Taulli at 00:29:08) So I would say that, you know, my recommendation would be, you know, let's find ways to retrain people. You have a certain path for them to take or different paths for them to take. Because I do think that, you know, it's very—you know, we've grown up in a society that, you know, we feel that we have to work because it almost shows—it's how we value ourselves. Like, we feel valuable. If we work, we earn money, support a family.

(Tom Taulli at 00:29:43) And if you can't do that, it's really no fault of your own. Psychologically, I think that's very damaging. And so, yeah, maybe over generations that changes. But for those generations that have grown up that way, it's very difficult. So I think that transition is not easy. So I think it's just trying to find ways—maybe it's some new deal of automation. I don't know. But, you know, there has to be different paths, different education, and things, you know, things like that. But I think there's certain industries that, you know, a lot of it will probably get automated.

(Joel Beasley at 00:30:17) Yeah. And different industries are at different levels. Like, when the whole port backing up thing happened, I started researching into that. And there's fully autonomous ports all over the world. We just—the unions got in the way and said, we're going to make it manual, we're going to have slow ports.

(Joel Beasley at 00:30:35) And when I see stuff like that, I'm like, I get it, right? Like, let's say I'm 55 or 60, and I'm about to retire, and I'm a port worker. Like, you don't want to go make that person learn a whole new skill. Like, there—and so there's definitely human considerations for the edge cases or the transition points. Like, universal basic income is something a lot of people talk about. I don't know a whole lot about it other than they're running a couple tests in some different countries, but it's like—some of the things, I'm like, those sound like okay solutions as long as they don't violate principles of a free economy. Because, I mean, you just give everybody money, you'll see runaway inflation. It's crazy.

(Tom Taulli at 00:31:25) We've already seen that.

(Joel Beasley at 00:31:26) We're seeing it.

(Tom Taulli at 00:31:27) In the last—we've seen it in the last two years. I mean, we pumped a lot of money, and we paid people not to work. I know. I mean, and we're now seeing the consequences of that. But, you know, there's certain people that prefer probably not to work, and they'd be just as glad to take money and not work.

(Tom Taulli at 00:31:45) I don't know what that percentage is. I know there's a percentage that believe that or are fine with that. I know there's another percentage that they just love to work. And even if they're not making money, I think they just need to work. So, but in aggregate, if you have a large percentage of the population getting paid not to work, I do think that's a problem.

(Joel Beasley at 00:32:05) I think it's also a problem if the—you know, humans are very impacted by their environment, right? So I'm careful about what's in my environment. You mentioned Geoffrey Hinton earlier. Is that how you say his name?

(Tom Taulli at 00:32:18) Yeah.

(Joel Beasley at 00:32:19) What does he do? I haven't heard of him.

(Tom Taulli at 00:32:21) Oh, yeah. So he's really interesting, uh, background here. So he grew up in England. His—I think his great-grandfather was Darwin. And his mother told him, if you're not a professor, you're pretty much useless.

(Tom Taulli at 00:32:40) Not that bad, but pretty much, like, you have to become a professor. And so he grew up in a very academic family in England. And so he pursued that path for himself of academia. And so he got into college probably in the late '60s, early '70s, and his focus was he loved AI. The thing is, during the history, during the last 30, 40 years, there have been these AI winters where AI gets shunned.

(Tom Taulli at 00:33:14) It just—everyone doesn't want anything to do with AI anymore because it's—and by the time Hinton got into the University of Oxford or Cambridge, I forget which one, he came in right smack when the first AI winter hit. And, you know, governments were pulling back funding because of inflation, and they said, you know, who cares? This technology is, like, it's interesting, but it doesn't help us in any way. So he got in at the worst time, and he didn't care. He thought that this technology was the way of the future, and he was in a winter for decades.

(Tom Taulli at 00:33:57) He was in a wilderness. I mean, no one wanted... Some of these times, some of these professors would come up with other names for artificial intelligence just because artificial intelligence had such a bad name because of the AI winters. But he had none of that. He was, "I'm totally into AI," and he kept at it, just kept at it.

(Tom Taulli at 00:34:19) So during the eighties, he came up with this concept of back propagation, which was a core part of deep learning. And the thing is, deep learning, the origins of deep learning go back to the late fifties. And someone named Frank Rosenblatt who came up with the core concepts, and it was called the perceptron. And it tried to emulate the neuron of a human brain, and he had the different hidden levels and layers and things like that, but not as sophisticated because he didn't have the sophisticated computers. And then there's another group, the symbolic group, Marvin Minsky and the folks at MIT, who were totally against the perceptron and thought you could create these expert systems and do if-then statements and solve the world problems.

(Tom Taulli at 00:35:05) And Minsky wrote a book that said the perceptron was terrible, and then Frank a few years later died in an accident, a boating accident. And the perceptron went away until Hinton took a look at it and said, "No, this is it. This is it." And so he became really the godfather of deep learning. And it was about 2010 or 2011, 2012 when we started getting these huge datasets.

(Tom Taulli at 00:35:34) And they started to apply Hinton's theories against these datasets, and the accuracy rates were just off the charts. Just off the charts. You know, so the recognition of... It was like cats, you know, cats versus dogs, and it's kind of simple. But the recognition was just so much higher. And then, all of a sudden, Hinton became the guy in artificial intelligence. But it took him, like, thirty or forty years to get there. So he's still around. Still alive. He cannot sit down because he has a major back problem. And I've seen interviews with him, and he never sits down. And he's just a brilliant person. And Google bought his company. He started a little company, and works for Google now.

(Joel Beasley at 00:36:28) Do you know him? Do you have a relationship with him?

(Tom Taulli at 00:36:30) I've not met him. No. But done a lot of research on him. Yeah. If you could interview him, then you should reach out to him. Fascinating story.

(Joel Beasley at 00:36:40) I'll have a standing desk company sponsor the episode.

(Tom Taulli at 00:36:43) Exactly. Exactly.

(Joel Beasley at 00:36:46) I want to talk with you about bias in AI. Right? And specifically, in the AI data, because data is one of the most important parts, you know, the whole garbage in, garbage out thing. Right? So this bias in AI data, you see it kind of emerge publicly as things like, "Oh, they weren't able to recognize this individual's face on the video because of the skin tone color." Right? And other different ways or people are represented in different groups have different angles on it. But my thought on it is, it's slightly different. Like, when we're talking about it, how can data not be biased? I'm curious to know this. Like, how can you have a dataset that's not biased?

(Tom Taulli at 00:37:36) Well, data reflects our society and our real world. So, and we're a biased species. I mean, we don't like to think we're biased, but we are. And we do discriminate. We're, from time to time, not very fair in the way we do things. So the data picks up on that. So, you know, in a way, you're just reflecting the real world, but you don't want to have decisions that you make using data that will result in unfairness and discrimination. So, you know, if someone is going to get a loan, we want to make that as fair as possible. You know? If someone wants to get hired, let's make that fair. We don't want to lock someone out because the data goes against them. So I think, you know, big move in AI is responsible AI, you know, that we just can't create a model and just let the model make all our decisions, especially in these sensitive areas, you know, about finance, you know, whether to... how many years we put a person in prison. You know? Are there computer programs that do that? So we've got to be very careful with some of these things. And I think companies are being more responsible, and I think that's a good thing. But there are a lot of disaster stories of... And the other thing too, it's not like the data scientist is there plotting, you know, these diabolical... Yeah. Exactly. You know? It could, but you then you go and look at a room full of data scientists, and they're all white, and they're all male, and they all went to the same schools. So what do you think? I mean, they're going to take a certain view on the world. So, you know, that's the other thing. I think we have to have more diverse data scientists. You know, we just can't have the same people with the same backgrounds making these decisions and creating these models.

(Joel Beasley at 00:39:44) Let's do a shout out for your book. Where should people go buy the book?

(Tom Taulli at 00:39:48) Just speaking of the world changing, I don't know if there are bookstores anymore. So the best way to get my book is to go to Amazon. Put my name Tom, last name T-A-U-L-L-I. Buy other books like RPA, AI, and so forth. That's the best way to get a copy of my book. And it's both, there's actually an Audible edition to my AI book. And a couple years ago when I published the book, I had people in Silicon Valley talk about it. And they go, "Can it be in Audible? Because I'm on the planes all the time, in the car. I just don't have time, you know, to just sit down and read." So there's an Audible version as well.

(Joel Beasley at 00:40:35) Thank you so much for listening. And if you found this episode useful, please share it with a friend or colleague that you think would get value from it. And if you have topics that you'd like to hear discussed on the podcast, either add me on LinkedIn, or send me an email [email protected]. Every time I get an email or LinkedIn message, it absolutely makes my day and inspires me to keep going.