with Ron Schmeltzer and Kathleen Walsh
“Trends in AI”
Ron is also a judge for the South by Southwest Innovation Awards and started and runs the Tech Breakfast. Kathleen is a serial entrepreneur, an expert in AI and machine learning, a savvy marketer, and a tech industry connector.
The birth of Cognilytica is when they both realized that the world is taking a huge leap in technology as AI is becoming more popular. The drive for them to create these podcast episodes came when they realized that people were still surprisingly thirsty for knowledge.
In this episode, they debunk the idea that technology is really the problem; in fact, it’s the challenge of getting people to understand technology. Ron also stated that the world of AI has a much higher overlap than people might have thought. Together, they describe the phases and actions they take whenever they run into a technological issue, outlining the procedure for resolving such issues.
Stream or download Episode 149: “Emerging trends in AI” below
Subscribe to Project Management Paradise via one of the links above and you’ll automatically receive new episodes directly to your device.
Highlights from Episode 149 “Emerging trends in AI”
00:00 – Introducing Ron Schmelzer and Kathleen Walch
01:08 – More information about Kathleen
03:08 – More information about Ron
05:37 – How come they are not running out of episodes?
09:17 – Their biggest episode challenge
12:26 – How similar are the technological problems?
13:43 – Business understanding
20:09 – Data understanding
20:39 – Data preparation
21:18 – Building and evaluating your model
21:46 – Importance of following a Methodology
22:58 – Where can we listen to Ron and Kathleen’s podcast?
…introduce yourselves and tell us more about your background?
Sure, I’ll go first. I’m Kathleen Walch, as you mentioned. I’m a managing partner at Cognilytica. Cognilytica is an AI focused research advisory and education firm. We’ve been around since 2017 and before that Ron and I had worked together at Tech Breakfast, so we’ve been working together for quite some time since about 2014 or so, and Cognilytica really got started and the AI today podcast launched simultaneously with that, because we found back in 2017, conversational systems were really becoming, you know, popular. They started to come out on the market and we said, all right. We started to hear a lot about that and we said, let’s get into it. So it actually started with, conversational systems since then have expanded to the full range of artificial intelligence, including what we call the seven patterns of ai. So conversational falls into that, and the reason we came up with that, we found out when people were talking about ai. It’s really an umbrella term and it means so much to so many different people that we may not be talking about the same thing. So I may be talking about conversational systems. You may be talking about autonomous vehicles. Ron may be talking about sentiment analysis, and so it’s all, you know, falls under that general category of artificial intelligence, but means. Different things. So we came up with the seven patterns as a way to shortcut that. And so at a high level it’s, you know, predictive analytics, conversational patterns, recognition patterns. So making sense of unstructured data. Hyper-personalization treating each individual as an individual goal driven systems. So that’s really around reinforcement learning, trying to find the most optimal path to your solution. And then we have autonomous systems as well. So the goal of autonomous systems is to remove the human from the loop. So this can be both physical systems or software systems as well. And then we have patterns and anomaly detection as well. But I’ll let Ron introduce him.
Yeah, great job. So I’m Ron Schmelzer, also co-host of the AI Today podcast and a managing partner as well at Cognilytica and on the podcast. You know, it’s interesting, you’d think that after, like we’re about 300 episodes into the podcast, in five years of running the podcast, you know, every week without fail, you know, since the very beginning. And our listenership has grown and grown, you know, tens of thousands of downloads. I don’t even know what a current download count is a lot. And you know, one of the things that we have realized is that people are still looking for education. They’re still looking for fundamental knowledge about ai. You think after all these years, AI is not new either. AI’s been around for decades, since 1950s. Right. And that’s, I think, what’s really powered. You know, people still want information. They want education, they want knowledge, they want best practices. They like to hear from others. That’s why podcasts are so popular. I think people like to sit and listen. And hear, maybe on their commute, if they’re still commuting these days, you know, and I think that is what, you know, keeps us going. And most recently, you know, it’s interesting you’d think that the technology issues are the hard ones. The technology issues are the ones that need to be resolved, right. And there are some hard technology issues, but those aren’t the hard problems. It’s the people problems. Getting people to understand things, get to work together, solving process issues, which have to do with the way. And what we’ve realized really over the past couple years and since the evolution of the CPMAI, which we’ll talk a little bit about shortly, is that it’s project management that is highly critical to making AI projects a success. Because the tools are great by some of the biggest and best. But the projects will fail not because of the technology, but because of people in process like almost every single time. So, hey, the world of project management and the world of AI have a much higher overlap than people might have thought.
That’s awesome. And just to tie it in with the AI Today Podcast, can you tell us how did it originate? You mentioned you’ve over 300 episodes today and that’s just fantastic. Was this an area that you noticed you wanted to have a conversation to spark it among other people, help educate them, and just, you know, really have an area where people can tune in and learn about all these different areas? As you said, it’s not just the one area, it branches off into many different paths.
You know, we do have over 300 episodes and we have never not found things to talk about. So it originated when Cognilytica originated because we wanted to find out what was really happening in AI today. I mean, people have different takes on things, but we said, you know, what’s really going on right now? What are organizations doing? How are people adopting these? What are best practices out there? So we started it as a way so that we could talk to others and we could learn in the industry what was going on. So it was a great way to start. And like I said, five years later, 300 plus episodes later, we still have not had things to talk about, which is great. So, that was originally how it started. And then why it keeps going is because we continue to learn. We have a really strong listenership and they continue to tell us, you know, what they would like to know areas where there’s lacking in the market so that we can help fill those gaps with some of those basic education needs. And then I know Ron had alluded earlier about how there was a lack of best practices, methodologies for actually doing ai, right? So one of our most popular podcast series was our AI failure series. Where a lot of people talk about, you know, oh, the positives, how you can succeed. Here’s some good, you know, use cases, success stories, but you can learn just as much, if not more from project failures as you can from project successes. And people don’t highlight that. So we said this is incredibly important to do. So we have an entire podcast series about common reasons why AI projects fail and how you can avoid that. And so that has been some of our most popular podcasts to date.
That actually really kicked it up a notch. I think what we found was that obviously, you know, you wanna interview people who are similar to your audience because I think what we find is that the audience, you know, identifies with like, oh, that person is like me. They’re struggling like me. They have the same similar problems, maybe another industry or. Maybe the company scaler size or culture is different. There always is, but like there’s some things you could find in commonalities and that works out really well. Then we started mixing in some of the educational stuff. We’re actually now in the midst of recording this glossary series where it’s literally just terms that people should know that maybe they think they know, but may, maybe they don’t. Or maybe there’s some confusion or some disagreement even around the word artificial intelligence. It turns out there is no standard common definition for ai. You might, you would think like, what? That’s crazy, but there really isn’t. Mainly because there’s no definition of the word intelligence. Which, you know, we have a lot of different ideas for it, but like we don’t actually have a common definition. So there’s a lot of characteristics. So you’re like, okay, well what are the characteristics that you expect out of an intelligence system? And that’s what we could talk about. And so it’s kind of funny and that glossary series turns out to be very long and people will listen to that. So I think in terms of the success of AI today, we found this like, you know, introducing some people that people may identify with the education component, as Kathleen mentions, not always highlighting the good stuff, sometimes highlighting the bad stuff. And of course our focus at AI today has always been, what’s happening with AI today and what can you do with AI today? Because there’s lots of podcasts on the topic and some of ’em focus on say, the research or what’s potentially happening in the future, or maybe some of ’em focus on sort of like the past and what’s happened then, or maybe some tangential issues in terms of like, you know, ethics or other. They’re important, but we’re like, well, what do you, what can you do about it today? So I think that’s always been our little angle and I think why our podcast is called AI Today, there you go.
Has there ever been a podcast guest or a topic that you’ve had on the show that has really left you thinking like, wow, I wasn’t expecting it to go this way, or, okay, that was insightful? Has there been anyone or any topics such as that just comes to mind?
So many. It depends on sort of like what you’re interested. So there’s one, we’ve had some like influential thought leaders who have come on, like, actually one of our first podcasts was , with James Barrett who had this,, the big book on super intelligence. And it was like one of our very, very first podcasts. And it was kind of interesting because a lot of the folks who are like really involved in the space, AI is our final invention. Thank you that that’s the name of the book by James Barrett. And you know, he’s a very famous guy. He’s a producer of some movie as well. And the truth of the matter is, is like, is that there’s this cycle, there’s this component of AI that’s not really based on sort of like the realities of where we are because a lot of systems really are not that smart. And we tell people, it’s like, have you talked to Siri lately? Or Alexas? These are not the smartest machines, but they’re, and they’re by the companies that are really working on the edge of innovation. So you, you know, when things are good as, as was told to us, it’s like, you know when you need to be scared of an AI system, when it could tell a good joke, . Then you should be scared cuz you know, it requires so much understanding of the world and context to really be able to tell a good joke. And so we realized that there’s a whole lot, there’s a whole aspect to AI that has to do with the way people feel about smart machines. Whether they’re scared, feelings of loss of privacy, feelings of the like lack of control of algorithms that are making decisions that impact their lives. Even algorithms that are not ai like, You know, possibly getting an account band and an account. People are fearing these things right now. Right? And there’s recent, there’s actually reason to be concerned because computers aren’t that smart. And we may put a little too much trust into these systems then as truly warranted. So there’s that. There’s have a lot of other guests. I’m, you know, I’m thinking about some of the ones who told us is like, you know, there’s too much math in ai, but yeah. , I know Kathleen, you wanna talk more?
Yeah, I mean, I think cuz we’ve had so many interviews, right? And podcasts in general. So some of the themes fall under, you know, as Ron mentioned, kind of these AI luminaries in the space. So we had James Baron, we also had Colin Engel, who’s the founder of iRobot. We had Ben Gerel who is with Singularity Net, and he also helped create Sophia Bot. So it’s interesting to get their perspectives on things and how they see things in the industry. Then we’ve also had alot of implementers on. So these are folks who are putting AI into practice in both the government and private sector. So that’s been interesting to see. And they’ve been, you know, from governments all over the world. We had, the chief Data officer of the Scottish government. We’ve had many people from the United States on. We had a lady from Oslo, Norway, so it’s really nice to get, you know, that kind of global perspective. We’ve had folks from Australia as well with how it’s being implemented across the world, and there’s been some common themes there. Everybody has the same struggles. You know, Ron had mentioned earlier, it’s nice to talk about use cases because sometimes you talk to people and they’re like, well, my use case is so unique and this is so specific exactly to my industry and my problem. And we’re like, why don’t you just step back a little? And look at this from a different lens because it’s probably not unique and you can learn from others. And so those have been really nice to highlight and showcase when we’ve had listeners come to us and say, you know, thank you so much for this podcast because it’s really opened my eyes. This is the exact same problem that I have. It just happened to be in a different industry. And what we found is that there’s not a lot of crosstalk and people are not collaborating. Types of ways. So that’s been something wonderful about the podcast. And then the, one of our most favorite series in particular, I know it was at least my favorite, I don’t wanna speak for Ron, was our AI failures. Because we’re able to say, this is where common reasons we’ve seen AI projects fail. Let’s explain it to you and then don’t make the same mistake. You shouldn’t have to. These are around data quality issues, data quantity issues. What is your ROI on a project? At Cognilytica, we’re advocates of best practices methodologies, so in particular the cognitive project management for ai, CPMAI methodology. Phase one is business understanding. That means make sure you are actually solving a real business problem. You would be surprised at how many people jump into projects , and they’re like, well, we don’t really know what our ROIs gonna be. We don’t really know what our problem is that we’re solving, but I was told that we should do AI or AI’s cool. So let’s move ahead. And we’re like, okay. And then, you know, 5 million dollars later, you wonder why your project fails. So that, yeah, .
Yeah. And I wanna piggyback on that because what made the failure series work for us is like, we didn’t talk about it, sort of like in theory or like as like a general learnings. But look, we actually did the rip from the headlines. And so we talk about, you know, Walmart canceled this major million dollar shelf scanning robot that they had invested all this time. As we talk about that, Amazon had to pull out this HR system, the AI system you’re using for hr, and they got to all sorts of trouble. We talk about the fact that, you know, Uber had these autonomous vehicles and actually killed someone. So that’s actually kind of problem, right? So we talk about that and we talk about, and so we go like, yeah, this company did this, and these are not small companies. These are big companies making big mistakes. Their failures are costing millions of dollars. Personal lives, you know, and there’s lots of, there’s like the Dutch government had used this algorithm for, you know, benefits that was inherent, had some inherent, you know, bias issues in terms of the data was, was bad. So database bias issues and, you know, they had to pull that back. So you don’t want, you don’t wanna, you don’t wanna be in the headlines for the wrong reasons. So the other really helps that, you know, that really helps. So we can go on and on and on about this, but, uh, you know, I think it’s really great. There’s lots of insight to be learned from people that hopefully are experiencing the problems before you have to experience them yourself.
Absolutely. And as you said before, like lot, not a lot of people actually talk about the failures in AI with their projects and that, but I find you would learn more from failing than always succeeding cuz there’s always that element of, you know, for the next project that when you did X not to do Y instead it, you know, it’s very beneficial. And there is that overlap with the AI and project management and it’s coming more evident in the coming years of how it’s used, sorry, in the applications of software and that, but can you tell us some lessons maybe you’ve learned with the overlap of AI and project management?
Yeah. I think a good place to start is that, as we mentioned, that ai, the one thing that people may or may not realize about AI projects, they’re not really about devel application development. They’re really about data. Because AI derives all of its ability from the ability to learn from data and to create generalizations from data and to create predictions and do all the things that we talk about every single pattern Kathleen mentioned about having a conversation or doing recognition requires that the system be trained to do that sort of thing. And that’s all because of data. And it turns out that data management issues are some of the harder problems, and not in terms of the technology, but like a lot of the the data’s not in good quality. We don’t have enough of it. You know, it needs to be augmented. Someone needs to go in there and add some, all this sort of stuff. Data ownership, data privacy, data security, data governance. If you start bringing these things up, you’re like, oh yeah, there’s a lot of problems, and if the success and failure of your AI system depends entirely on that data, it’s the old garbage in is garbage out, which is 100% the rule for ai. Then your AI systems will fail. So, what we realized that there was a methodology in, in this case, not a generic project management methodology, a methodology specifically for doing data projects called CRISP DM, the cross industry standard process for data mining, which has been around since the late 1990s, but hadn’t been iterated develop for years. And what we did is that we brought in new iterative and agile styles of, of project management combined that with, of course, the new requirements for AI and machine learning. And that’s what brought about CPMAI, the cognitive project management for ai, which was put into place about when Cognilytica started in 2017. First with some banks, some really big. And then it evolved to some large government agencies, and now there are thousands of people and, and tons of organizations. We have like this whole logo pile that you can take a look at, you know, from Coca-Cola to Rio Tinto to, you know, all the, the, these big banks and whatever. It’s a methodology that they’re using for AI project success, but it’s mostly mostly about doing things in the right order. So maybe Kathleen, you can chime in on that pro process and project management methodology.
Yeah, so what we found, you know, I said CPMAI starts with business understanding. So project managers understand methodology and what we also have found is that in order to successfully run AI projects and make sure that you’re running them the same predictable way every time, doing it in the right order, you need a methodology to follow. So that’s, you know, kind of where this overlap of AI and project managers comes into play. And how CPMAI fit so nicely into it, because there’s methodologies that you need to follow, but they’re data centric. And so when you can do that, then you’re doing it in a repeatable way. And what we’ve found also as, as Ron mentioned, so it’s not like software application, so you’re not gonna wanna apply, you know, just agile methodologies for this. You need to enhance that as well. But starting with, you know, using some of those terms and also starting with that kind of base methodology we found really helps with success and helps project managers and different folks on the team understand things better.
Are there any AI best practices that you find will be best implemented in this area? Or have you any experience (of AI) which would be, in your opinion, the best practices?
Yeah, there’s, and that’s another great thing. You want to borrow best practices that have proven to work from other approaches that other industries, you don’t want to create something from scratch. And I think that’s one of the things we learned is that, you know, as you’re putting together CPMAI, bringing in more of the agile methodologies and other methodologies that have already addressed some of the aspects of just generic project management. But, making them thinking about them from a data-centric perspective really works. So what we found was that you have to address the data issues early. So, as Kathleen mentioned, phase one and this methodology as business understanding. Phase two is data understanding, which means you need to understand. What data’s needed? What data do you have? Where are the sources of data? What’s the quality of that data? Issues of privacy, security, governance, all that sort of stuff. You can’t move forward without that. There’s this thing called the AI Go, no-Go. It’s like these nine traffic lights, they all need to be green before you can go, you know, anywhere. Otherwise you are going into dangerous territory. Right? And then after data understanding, the third phase is data preparation. So you actually start building, what do you call these pipelines where you start dealing with trying to get the data to where it needs to be in the right quality, with the right additions, enhancements, and all the things you need, transformations before you even start building your model, because there’s no point to to doing that. Especially since we’re talking about big data. And big data, as you know, is not just about a lot of it, which is the big part, but it’s also about the big data that’s changing a lot. That’s in different levels of quality, that’s in different levels of, of issues of speed, in terms of how, how much is changing and all that sort of stuff. Different variety as well. All these so-called VS of big data. So the methodology deals with that. And then fourth phase, then you can start building your model to the business requirements. And then after that you can evaluate the model, which is the phase five to the business needs as well. And then you, go ahead and push that out. And obviously in very short, iterative sprints. So that’s, that’s what we learned was that, you know, really starting making data, the start of the and core of the whole approach.
Exactly. Yeah. I was gonna say, make data core have a methodology that’s data centric and also, follow a methodology because we’ve seen far too many organizations, I mean, you know, Ron had talked about RIP from the headlines. We wanted to say, this isn’t just small organizations. This isn’t just organizations that maybe are doing the first AI project, that are failing. It’s major organizations as well. And that can come from this fundamental lack of following a methodology. And if you’re not doing it across the board, then different groups are going to be doing different things. Unfortunately, far too often when we’ve asked, you know, companies, what are you doing? Either they kind of look at us with like, what are you talking about? Or they say things that are kind of crazy, like they’re doing the scientific methodology and we’re like, I don’t think you’re doing the scientific methods for your AI projects. Like if you really think about it, you’re not doing that, so you shouldn’t be sharing that. So that’s why I think those are some of the major lessons that we’ve learned.
This has been a really interesting podcast and I’ve really actually learned a lot myself now. If any of the listeners would like to listen to the AI Today Podcast, which would be the best platforms to stream it on?
Yeah, listen to all of them on any of the platforms you like. We’re on iTunes or or Apple Podcast. We’re on Spotify, we’re on Google. We’re syndicated everywhere. We’re also on our website. You can go to AItoday.live and see all the episodes there. The other thing we have is, That we offer a lot of training and certification and courses, especially in the CPMAI methodology. But we have a lot of free stuff. So if you, if you check out, we have an intro to CPMAI. So if you’re a project management person, as you all are listening to this podcast and you’ve been like, Hey, maybe, maybe I should think about AI and data. Maybe I’m involved. Maybe I want to expand my career or something like that. We have an intro to CPMAI. It’s like a mini-course for free. You can check it out. AI today.live/cpm a i We even have its own landing page there. Check it out. But you know, as far as like, you know, listening to the podcast, if we’re on a platform that if we’re not on a platform that we should be, you should let us know, but we we’re hopefully on all of them that we could possibly be on.
Exactly. Yeah. We try and make it super accessible so you can listen wherever you like to listen to your podcast.
Thank you so much for being on the podcast today, and we’d love to have you again on the podcast in the future.
Yeah, and I have to mention, we also will be having, the folks from Cora on the AI Today Podcast, so stay tuned for that. We’ll have some news about when that’s out, and we’ll promote that to all of our channels. We’d love to hear your perspectives on project management as it, overlaps with AI.
Connect with Ron Schmelzer on LinkedIn here
Connect with Kathleen Walsh on LinkedIn here
Check out Cognilytica: cognilytica.com
Learn about the magic of digital twins by accessing a complimentary guidebook at corasystems.com/digitaltwins.