107: Doug Rose - Opportunities in Artificial Intelligence!

Are you thinking about how to grow your business with near and offshore staffing, but you're also wanting to make sure that it's reliable, cost effective and really well done. I think docsa talent is the way to go. Here's why. They don't just fill your jobs. They bring integrity, top talent and a people first approach that's hard to find. I've known the people at Doxa talent for over 20 years now and I say that you can count on them. I have talent. Talk to Doxatalant and you won't be disappointed. Go to doxatalent.com Bill G. Again, that's doxatalent.com Bill g. One more time. Docsatalant.com Bill G. Check it out today. Well, Mark Cuban has said that the world's first trillionaires will come from the field of artificial intelligence. They will be the ones who figured AI out and did well with it. Today's show will be about what you can do with the AI tools of today, the mistakes to avoid and the opportunities of that. So we'll keep looking at artificial intelligence by artificial intelligence. We're really talking here about prediction, machine learning kinds of things. The weak AI, not the scary sci fi, general AI type stuff. So listen in and see what this means, what you can do with your business in the world of AI. Hello and welcome again to the Scaling Up Business podcast.

Guest I'm your host and growth coach Bill Gallagher. On the show with me today is Doug Rose.

Doug is the author of several books.

On artificial intelligence, data science, management, software development.

He's also a teacher at the University.

Of Chicago and through places like Lynda.com and LinkedIn learning. We're talking to Doug today in Florida where he has begun his summer break with his family and he's going to.

Do his talk from outside while I'm.

In the office studio here doing that.

So I'm jealous of where you are today.

But after you talk, I'm actually going to take my dogs for a walk. So I feel you starting to move into summer.

Yeah. Well, thank you very much, Bill for having me. Yeah, I mean I'm in Florida here, so you might hear some random wildlife. I'm going to try and stay close. Yeah.

So Doug's an upcoming speaker at our Scaling Up Summits. That next summit will be in Denver, Colorado in October. 1617 in Denver. You should register now. If you are a coach with if you are coach or work with our company, we'll list you and have you give you a discount. Have you be part of that group, but he'll be a speaker there. I think we're going to get a taste of some of that. We're going to talk today about artificial intelligence for business. The things that you need to know, things that you should think about as you bring AI tools into your business. Right?

Yes.

So tell us about, you know, give it. Probably most of our folks don't know you. Give us a little bit of your background. How do you come to writing a book on AI for business?

Yeah, thanks. I started out really kind of in the software space. I was originally went to the University of Wisconsin, Madison. I was English major because I was planning on being an English professor. And but then when I graduated in 1992, I found out that one of my hobbies of sort of working on computers and things like that, actually employers were much more interested in that than they were in.

When did that hobby start? I find most folks who have any computer background, like I started playing with computer stuff when I was a small kid.

Yeah, I had a Vic 20 or something when I was a small kid. But I didn't actually become interested in computers until, strangely enough, my freshman year. I didn't lock up my stuff very well and it all got stolen. And so I got a check from the insurance company and bought a Macintosh LC with it, which I kept in my empty apartment. And I kind of learned on my first Mac. And so I became sort of a technology person. And so my first jobs were really around technology. I was never a software developer, but I was more of a network person. And then later on that kind of turned into me being almost like an early project manager. And I started working with rapid application development and then became agile. And I sort of worked into becoming an agile coach right around the time when there wasn't very many agile coaches. And yeah, so then I ended up working kind of on these really interesting database projects, sort of like these data science agile projects. And I worked for companies that were using large data sets and they wanted to kind of use the agile mindset with that. And a lot of these organizations were just starting out using machine learning. You know, they had these massive data sets and so they wanted to get deeper insights and so they started working with machine learning algorithms to kind of figure out more interesting new things about their customers. And that led to me kind of becoming very interested in artificial intelligence and specifically how managers and people who are not engineers can kind of understand these AI tools and technology in a way so that they can kind of work to develop products.

So let Me just break it down for our listeners and you can correct me or if you've got a better one. But I think a lot of people.

Sort of glaze over.

There's the science fiction notion of AI and generally we're talking there about a general AI, something like an aware machine that is either utopian or dystopian. Right. It's either out to kill us or to help us. It could be the girlfriend, like in her, or it could be a devious robot like in Terminator or iRobot or like that. But that's not what we're talking about here. And for most business where AI are powerful tools, but it's in the world of machine learning. And for that we're distinguishing between the.

Kind of simple instructions that could be.

As basic as you create a long formula in Excel or you write a basic program in, in a computer language, to teaching a set of rules and outcomes to a machine that then teaches itself by looking at a much bigger data set. Right. So instead of teaching the specific things, we're giving some broad parameters and letting the computer teach itself to recognize and adapt and predict things. Right?

Yeah, exactly. Right, yeah. Some people refer to it as general AI, or more commonly just strong and weak AI. And strong AI is kind of Commander Data if you're a Trek fan, or C3PO if you're a Star wars fan. You know, that's just kind of like a robot that has a sense of humor that knows why it's doing things. You know, I mean, when we grab our phone and we talk to Siri or Cortana or something like that, and we say, how are you doing? She responds because she has a data set which kind of gives her an appropriate response. She says, I'm fine. But when Siri says she's fine, it doesn't really mean she's fine. You know, just. I mean, it doesn't mean that she's feeling good or that her data centers are running smoothly. It's just sort of her responding with kind of the. She's kind of flipping through a massive book of responses and then responding with something that is sort of most commonly used. Well, not pre programmed, but yeah, but sort of that connected, you know, so that's. So we're kind of, we're just starting out with Wiki really. And weak AI is really, when you're going to Google or something like that and you type something in and it finishes your search or something like that.

Or even it's making prediction based on.

Yeah, or it's predicting or it's finding patterns in these data sets. Most of the tools that we're talking about, not really all the tools we're talking about now are around weak AI. And so it's in my book, I kind of talk a little bit about some of the challenges in these weak AI projects and how some managers should sort of look at these challenges, some of the things that they should kind of understand when they start building out these AI products and some of the things, the language they need to understand in some of the, the, you know, ethical considerations.

So let's talk about that. Let's dive into what are the big ethical management challenges or the, the top ones that you think people need to think about as they go in. Because I think everybody's now who isn't already. If you're already involved. Look, this is maybe not, you're already going to know a lot of this stuff, but, but if you're thinking about what does this mean for my business, where do I begin? This is really addressing that piece of it. Right?

Right. Yeah. And it's kind of a lot of companies, when they're just starting out with machine learning, they kind of leave a lot of these decisions to engineers that are locked away in a room and they don't really kind of understand the language as well as they could. And so a lot of these ethical challenges are being decided by engineers, which is not, you know, ideal kind of long term. So if you, you know, if you have a self driving car and just.

You don't, you're not saying engineers lack.

Ethics, you're just saying that more people.

Should be involved in the conversation.

Right, exactly. I'm not saying engineers aren't ethical, I'm just saying that that's not consistent with their training. You know, if you, if you're not, if you're deciding, if you've been trained to kind of figure out how to implement something with technology, you might not be the best person to decide whether or not that technology should be implemented. And so if you think about a self driving car, they're programming self driving cars so that they do a certain thing. If they're in danger of being in a car crash, they might decide to run into somebody else or you might have to just kind of as a passenger, just take a hit for the team. And these decisions are kind of being made by engineers and they're not really being made necessarily by managers who kind of understand a little bit about what's going on, sort of can get insight into what kind of decisions are being made. And you know, and there's sometimes when they, when they approached Google about what were some of the algorithms that they had to decide how a self driving car would crash, they weren't very forthcoming about what it was going to do. And so these kind of things are I think, a big challenge. And I think managers and business people, they kind of need to be in the conversation a little bit more.

Yeah, so what? Well, why don't you tell us some of the examples? You had some great examples from the book in your talk about places where folks have got it wrong or early mistakes. Why don't you share some of that with us?

Yeah, I mean, since I'm in Florida, a pretty good example is what happened here on the other coast in Broward County. There was a company called North Point and they were, they came up with statistical machine learning algorithms that tried to figure out if someone was going to be a repeat offender. So, you know, someone goes and they, you know, if Bill, if you and I decided that we were going to go rob a 711 here at 2am Here in Florida, then, you know, after we were booked, they would run us through this program and see what are the odds that we're going to kind of Rob another 7:11 once they let us out. And so they created this machine learning algorithm. They used statistics and it turned out.

Minority Report, right?

Yeah, it is actually. But the problem is that they went through the massive data set that they went through to train the machine was the result of jury trials. And it turns out that, you know, jurors can be really biased. And so if it turned out having a real. And this was something where the judges would look at the report just before sentencing. So it wasn't just kind of something interesting, it was something actually that was wrapped up in our judicial system. And it found out that when they looked at it, it turned out that if someone was African American here in Florida, there were 77% more likely to be tagged as someone who would commit the crime again, even though they committed the exact same crime as someone who's not African American. And they may actually have an identical criminal record.

So wait, you're saying that. So they developed an AI that was used in sentencing to say, look, is this person likely to commit a crime again? And that report was given to a judge for sentencing, but the data set that it based on was an already biased thing that had inherent racial bias bias built into it. So it's kind of like confirmation bias with AI tools now accelerating and expanding.

Is that exactly right? Yeah. And when the judge is handed this report, you know, from a, you know, from a AI machine. It's got a lot of weight because people tend to think of AI and the data and the patterns it finds as math. They seem to think of it as facts.

The computer said this.

It must be exactly, you know, the same thing that delivers my email said this. And so it's got to be right. And so, you know, that's. Those are some of the challenges, you know, and there was a. In a company that I do a lot of work with, I work with Microsoft, LinkedIn, and they ran into a very similar challenge where they came up with kind of this, this teenage chat bot. Yeah. And the engineers in a room decided that they were going to scrape a bunch of Twitter feeds and use it as the training set to respond and have a conversation with the Internet. And if you think about it, I mean, if you're scraping random Twitter feeds on the Internet, that you're gonna run into a lot of stuff that's kind of a little unsavory. You know, people are always the best selves on the Internet.

Wait, wait, wait, wait. You're saying there's racism and sort of base humanity on Twitter?

I am, I am. I know we're trying to break new ground here in this podcast. And so horrifying. I'm gonna go out there. Yeah.

So, yes, I will stay sarcasm alert for everyone. I'm being sarcastic.

Right.

Twitter is filled with just absurd conversations, right?

Yeah, it's just like raw ID in text form. And so it's. So they turned out that once they took that training set and fed it into this AI bot that was supposed to have the personality of a 13 or 14 year old girl, I mean, she turned into like a Nazi or something within 24 hours. Started questioning whether the holocaust exists really happened. And so they pretty quickly took it down after that and then actually brought her back up again and she did the same thing. And so these kinds of challenges and questions shouldn't be just the sole struggle of engineers sitting in a room trying to figure out if it can be done. Managers and business people kind of need to be in the room to be asking tough questions, understand what tough questions to ask and to understand these challenges. Because, you know, I mean, Microsoft obviously took a pretty big hit for that.

Yeah, I think in both cases. Right. We talk about things that deploy and then have big impact rapidly. So technology is a tool. Right. And a tool gives you advantage, scale, speed, things like that. And so you get something and it begins to have unintended consequences on a very large scale rapidly.

Yeah, And I think that that's, you know, these consequences aren't really being addressed early enough. I mean, they kind of pick up after something happens, but there's not really enough people in the room right now. And I think, and this is one of my. One of the reasons that I wrote the book is that it's because I think a lot of managers and business people don't feel comfortable with the language and don't feel comfortable with the concepts. And so I wanted to kind of create a guide that both introduced people to AI concepts, but also kind of got them percolating about these challenges.

So from this I hear probably the need to test and iterate and scale carefully, be ready to pull down your AI tool. I hear the potential for big mistakes on a big basis. But it seems to me that all of these things are using machine learning, predictive tools. So we have painful examples like we shared, and then there's more comical examples like Siri or Alexa picking the wrong music for me or doing things like that that are annoying, amusing, but they're not necessarily massive business cost problems. Right?

Yeah, I think that's right. And I think that another thing that businesses have to keep in mind, sort of, and again, I think this is more of a business decision than a technology decision, is that I don't see enough business people kind of talking and thinking about what's going to happen with the massive amount of automation that AI is going to lead to. I mean, you have places like PricewaterhouseCoopers or something like that, and they say in 15 years, 38% of jobs are at high risk for being automated out. And you've got to employ research firms like McKinsey that says by 2030 that 50% of occupations can already be automated by the current tools we have, and that there could be as much as 60% with new tools that are coming out. And so this has the potential of just having massive numbers of jobs just disappear. I mean, the same report said there could be, and this is kind of a wide guess, but 400 to 800 million jobs kind of worldwide just kind of disappearing.

So when AI learns how to do something that you do now or have somebody doing now, that job potentially goes away and then maybe some other jobs arise at the same time. Right. But not necessarily. Maybe declining number. Right.

Yeah, we've always been lucky, really, in terms of having good job replacement. You know, jobs disappear and then new jobs come in. But here we really have the potential of losing kind of a Massive number of jobs. But the jobs themselves kind of don't disappear, they just become automated. So you'll have companies that basically hold these jobs, but they just hold them with automated tools. So I mean like there was a huge uproar in the US election here about all the coal miners and things like that losing jobs, coal mining 10, 20,000 jobs in West Virginia. But once they perfect the automated truck, you're talking about three and a half million truck drivers.

Right.

And it's not like all those truck drivers are going to become personal trainers.

So from a business, right, so our folks are entrepreneurs, business leaders. I hear a couple of different things. For one, this is not merely labor. I mean there are business decisions, economic prediction, market driven trading, there are a variety of business decisions and sort of general business work that may get changed with AI tools, significantly reducing the headcount processing of things and so on. At the same time though, my company, if I don't embrace and use AI tools, my competitor may, and they may end up with significant advantage, cost, speed, quality, etc. Over me if I don't figure out how to apply these machine learning tools to my, to my own business. Right. At the same time, if I do it before them or better than them or then I may gain some advantage.

Yeah, And I think that's exactly right. And I think that there's going to be, and there's like a, there's a lot of entrepreneurs that say that there's going to be massive opportunity. I forgot someone said, was it God? I forgot the name of the person who owns the Texas team, but it said that the first trillionaire is going to be for, from AI Mark Cuban. Yes, thank you. And so, you know, there's no doubt that this is going to be, there's going to be massive opportunities here, not just for large businesses like Google pretty much say that this is their primary business now, but also for kind of small niche entrepreneurs to come in and perfect kind of a machine learning algorithm, machine learning tools to, to sort of take care of some niche. And I think that that's, you know, I think that's going to bring a lot of prosperity. But I also think that businesses have to kind of weigh that against what their responsibility is to their employees and to each other. And so it'll be kind of an interesting struggle kind of going forward is that, you know, what's, what's your responsibility to your workforce? What's your responsibility to your shareholders? As I mentioned, kind of, I'm from southeastern Michigan, I grew up outside of Detroit. And So in the 80s there was these massive exodus of jobs from that area and a lot of people felt that employers did not have a huge responsibility to their employees. And so that may have changed by now. So sometimes you have to kind of weigh those against your responsibilities to innovating, but at the same time your responsibilities to your employees and to society.

Well, setting aside the political and social implication of the whole thing, there's an immediate business opportunity, right, to not get left behind or to lead the way with doing something. So figuring out how to use this to give your company an unfair advantage, to move yourself ahead, to put your solution. And that's to figure out, so what are the things that we do right now where maybe it's buying and selling something? Around the world of the scaling up coaches, we've worked with a couple of companies that did work with AI driven ad tech. So people were making better ad buys as a result of the tools that were making better predictions about which buy would produce better conversion and so on. So there's one example, but think about in any category, whether you're a small manufacturer, so it might look at buying things, at defects, at spotting and quality management of your lines, or managing inventory levels and purchasing and things like that. You could predict things in staffing and people related things, what people are needed when getting smarter about that, leading that, finding them. Online recruiting is a huge challenge for everybody right now given the, the current unemployment rate as we speak today is not favorable to the employer right now. But in any realm, as we think about the different companies, equipment rental that we deal with, specialty contractors, technology firms, software services, a whole range of things. In all of those, there are things that, where you might be able to use AI tools to improve the way your business runs.

Yeah, I mean I personally think that the sort of the free model is going to be the road to a lot of businesses doing very well in this environment. I think that the currency that you want to accumulate is data. And so if you can come out with tools that give value to an organization and then return for that value, collects massive data sets from them, then over time you're going to be able to do something really interesting with that data set that your competitors aren't. And so it's kind of like the accumulation of data. If you can get the more of it that you get, the more likely you'll be able to succeed using these tools kind of going forward just with the technology we have now.

So building or obtaining large data sets is a valuable thing that you could do something with and the tools can help you to find the patterns within it that are useful, right?

Yeah. And even building is sometimes better than buying because you can, because then you have something that's truly novel and something that only you have access to. And so if you can create free tools, you know, I mean, it's. If you think about it, when Google was giving out Gmail, you know, free email to everybody, it didn't seem to just seemed kind of like a freebie, it was like something nice that they did or when they do Google voice or something like that, it just seems like sort of a freebie. But the amount of data that they collected from our voicemail messages and from our email messages are the things that allowed them to kind of do that. I don't know if you saw the duplex that they did, this demonstration of an AI agent kind of calling up a bunch of businesses and making appointments.

A lot of our listeners probably heard that it's been such in the news, right. That this simulation, and of course on the news also the likelihood that some parts of it were staged in some way. But even if not all of it, it was a significant advancement in this sort of natural language interaction.

Yeah. And all of that information they got from our voicemail messages, they know that if you listen to millions or billions of voicemail messages, then you're going to get a sense of what a human conversation is sounds like. And you can get the ums and things like that and appropriate pauses so that the person sounds, the AI agent sounds a lot like a human. And so that, you know that kind of giving away something that's valuable, that's free to collect data and then use that data, because if it's free or low cost, then you're more likely to accumulate a massive amount of data. There's also a huge opportunity with IoT Internet connected devices, sort of the Internet of things, where you could put small sensors on things and collect massive amounts of data and then do something interesting with it. There is a lot of opportunity now to kind of collect massive data sets. And I think that the organizations that have the most robust and the sort of the, the best data sets, the highest quality data sets are going to have a much easier time. And I'm making money in the future with these tools.

Yeah. So the more folks you can touch, interact with, get free data from, give something away builds you something that can create some advantage at some other point.

Yeah, I think so.

Yeah. I can think of a lot of different examples that we've seen, but the hybrid Companies that will provide some folks free online content or tool or something like that, where they're touching a lot of folks. So again, they're using Internet plus free to build a big following, trading up something now that creates other sort of ethical and disclosure conversations about what you're trading and how you're providing it. But if you can handle that successfully, then that gives you something that you may combine with other customer data or customer profile data. You might be able to figure something out that is pretty interesting.

Yeah, I think so. You know, I mean, a lot of people with the Facebook controversy, with Cambridge Analytica, you know, and everybody was kind of clutched their pearls. I didn't know all this was happening. But if you look at Facebook's model, I mean, they're very clear on what they're doing. And so they. I mean, they are excellent at creating these massive data sets and understanding relationships between different data points. I mean, that's really the core of their business. And so, I mean, if you've ever done advertising with Facebook, then you can see that they can really tie down exactly who your audience is so that you can kind of target them in a way that's extremely sophisticated. And that comes from sort of this model where they're just collecting massive amounts of data on pretty much everything you do. And Amazon's even sort of. I think they seem to me to be. I mean, I don't have that much internal insight in, but they seem to be catching up a little bit. But now that they collect sort of mass amounts of data, set about everywhere you go on their website, they do, in my opinion, a much better job with their advertising, which is sort of a growing part of their revenue.

So if I want to be on the winning side of this, I'm running a business. I want to collect and build my own data set as big as I can, and then combine that, leverage that into some other world or some other meaning. But I want to be careful not to undermine the brand value, our business reputation. Do it in a way that's appropriately disclosed or appropriate protections for privacy, complying with laws, things like that, so that I don't end up finding myself smeared or talked about in the media. In a way.

Yeah, I think. I think that there still is quite a bit going on that would surprise people just in terms of data collection and how sophisticated these tools are already. But, yeah, I think that if you want to be. Absolutely. If you want to be on the winning side of this, a, don't be a trucker, and B, make sure that you're collecting massive data sets, high quality data sets. And a good way to do that is by providing something of value, whether for free or low cost, and then building it up over time, refining it, using machine learning to sort of perfect the patterns that you see within the data. And this might take time, sometimes years. I mean, when you looked at the Google duplex thing, I mean, that was a result of a decade or more of kind of crunching these kind of voice, this massive amounts of voice data that they have. And then once you've done that, you've sort of perfected your model and you're using the correct machine learning algorithms, then you're going to have something that's extremely valuable.

Yeah, it's interesting. And you may discover that you have data that you didn't realize, depending on the stage that your company is at. When I ran international calling for telecom company years ago, we had a very, very small percentage of that highly lucrative, highly profitable piece of the business. But we handled and terminated calls. So the, the stuff that we build and sent out to the world was a tiny percentage, and we got back a tiny percentage, but that percentage was random and applied against everybody else. And we inferred and guessed correctly that the people that received calls from other countries likely placed calls to other countries. And so we just put all of our traffic data tapes through something and compile lists against each country. We pick the profitable countries, the profitable routes, and then we match those to marketing databases of information. And then we could call you or email you, or target you, or mail you and say, hey, we'll give you a discount on that route or we'll give you a special plan or we've got something for you. And that was extremely, extremely effective. Combining data that we had some access to that was built over time that was just sitting there ready to be used.

Yeah, I think that's exactly right. That's a really good example. And now you're getting to the point where you can sort of collect massive amounts of data much more cheaply. And then you have tools that look through confined patterns that humans can't perceive. And so you're able to get to the point where you're able to even target even more.

Yeah, my example is a crude, simple one, a simple matching thing. But now you can combine even bigger data sets of things that you're collecting via your web server or your other tools.

Yeah, and combining different types of unstructured data. I mean, if you could see if you could come up with a model to understand the quality of the call the outcome and then combine that with structured data about rating systems and things like that. And then bringing that all together and creating a sophisticated model. That's the kind of stuff that's going to be terrific.

There's so much more to talk about and I think there's a lot of then the tools, the methods, the approaches, the cases, things like that. I know you'll get into more of that when we have more time in your talk at the summit. And I know you've also got a ton of this stuff in the book and so I want to let folks know that you've made a chapter available for free online of the book so people can get a taste of it. And then of course, you know, you can buy it from your regular book source. But if you go to Doug AI Free, you can download a free chapter of the book there. Artificial Intelligence for Business. Doug Rose is the author. He'll be speaking at our Scaling Up Summit in Denver this October. So register for that or go to Doug AI Free to get a free chapter of the book. Read some more examples of this, learn about the tools, how to use them like that, and begin to look at where to do this. I think you've really just begun to get our listeners minds sort of going about where and how this applies, which is I think a great place to start.

Terrific. Thank you very much again for having me.

Yeah, thanks so much for coming on the show and I want to remind everyone that if you'd like tools and topics like this for scaling up every week we've got a new podcast for you pretty much every week. You'll find that on our website@scalingup business.com where we also have free tools and downloads for getting your business scaling up, getting your team started. So how to run team strategy and planning sessions, worksheets and tools for getting that stuff done, how to apply all the basic ideas of scaling up to your business right now for free@scalingupbusiness.com thanks again, Doug.

Thank you, Bill.

I want to give special thanks again to our original growth guru, Vern Harnish. Our show is produced by niob, audio production by Podfly Productions, audio editing is done by Albert Burge and our show notes are compiled by Ayn Kodina and proofread by Tim McGowan. If you got some value out of the show today, please share it with somebody else and let us know the parts that you loved. Thanks again for listening and keep Keep scaling up. Are you ready to turn your goals into daily wins? You should check out our Scoreboard software Powered by alignment, it's the official software for scaling up. With Scoreboard, you'll get an AI driven coach for priorities and KPIs and it'll help keep your team focused on what matters most. With Align, you can track your goals in real time and you'll even see results on the go. With their mobile app, you can rest easy knowing that Align has world class onboarding and support to set your team up for success. This is really more than just software, it's what I actually use to manage growth in our scaling up client teams. Unlock clarity and results with Scoreboard by Aligned. To try it now, go to aligntoday.com Bill G One more time. Aligntoday.com Bill G that's aligntoday.com Bill G Check it out and try it now.