AI Podcast Ep. 4: Four Most Commonly Asked Questions About AI with Dr. Jerry Smith

Episode Description

Dr. Jerry Smith welcomes you to another episode of AI Live and Unbiased to explore the breadth and depth of Artificial Intelligence and to encourage you to change the world, not just observe it!

Dr. Jerry is talking today about questions and answers in the world of data science machinery and artificial intelligence.

Key Takeaways

What are Dr. Jerry’s favorite AI design tools? Dr. Jerry shares his four primary tools:

MATLAB is a commercial product. It has a home, academic, and enterprise version. MATLAB has toolkits and applications. The Predictive Maintenance Toolbox at MATLAB, especially the preventive failure model is of great value when we want to know why things fail, also by measuring systems performance and predicting the useful life of a product
Mathematical Modeling with Symbolic Math Toolbox is useful for algorithm-based environments. It is built on solid mathematics
R Programming is Dr. Jerry’s favorite free tool for programming with statistical and math perspectives. R is an open and free source and comes with a lot of applications
Python is a great tool for programming and is as capable as R programming to assist us in problem-solving. Python is very useful when you know your work is directed to an enterprise level

Does Dr. Jerry have any recommended books for causality?

The Book of Why is foundational for both the businessperson and the data scientist. It provides a historical review of what causality is and why it is important
For a deeper understanding of causality, Dr. Jerry recommends Causal Interference in Statistics: A Primer
Counterfactuals and Causal Inferences: Methods and Principles it is a great tool to think through the counterfactual analysis
Behavioral Data Analysis with R and Python is an awesome book for the practitioner who wants to know what behaviors are, how they show up in data, the causal characteristics, and how to abstract behavioral aspects from data
Dr. Jerry recommends Designing for Behavior Change, it talks about the three main strategies that we use to help people change their behaviors
The seven rules of human behavior can be found in Eddie Rafii’s latest book: Behaviology, New Science of Human Behavior

Dr. Jerry shares his favorite tools for causal analysis:

Compellon allows us to do performance analysis, showing the fundamental causal chains in your target of interest. It can be used by analysts. It allows users to do “what -if” analysis. Compellon is a commercial product
Causal Nexus is an open-source package in Python that has a much deeper look at causal models than Compellon
BayesiaLab is a commercial tool that is one of the higher-end tools an organization can have. It allows you to work on causal networks and counterfactual events. It is used in AI research

What skills are needed for data science machinery and AI developers?

Capabilities can be segmented into data-oriented, information-oriented, knowledge, and intelligence. These different capabilities are used in many roles according to several levels of maturity

Transcript [This transcript is auto-generated and may not be completely accurate in its depiction of the English language or rules of grammar.]

Intro: [00:04] You’re listening to AI Live and Unbiased, the podcast where we explore how to apply artificial intelligence and the impact it has on enterprises and society. Now, here’s your host, Dr. Jerry Smith.

Dr. Jerry Smith: [00:19] Welcome to AI Live and Unbiased. I’m your host, Dr. Jerry, I’ll be with you on this journey as we explore the breadth and depth of artificial intelligence. Today, we’re going to talk questions and answers in the world of data science, machine learning, artificial intelligence. If I get crazy enough we may even explore the meaning of why itself and no, I don’t think the answer is 42. Today, I’ve decided to spend some time talking about four primary questions that are often raised. I see these a lot coming my way almost every conversation, whether it’s with a colleague or a client that’s out there. These questions range from tools to skills to that books that I read. So I hope some of these interest you and if they do please comments on the ones that interest you. And if you think I could add to some of these things, please leave comments as well in the comments. Dr. Jerry Smith: [01:14] The first question that we’re going to start off is Dr. Jerry, what are some of your favorite data science tools? I love this question. In the world of data science, especially looking at and studying data versus the machine learning, which is predicting how that data’s going to change in the future versus artificial intelligence, which is about providing intelligence to support human beings in an artificial way, right? Three very different things, right? You got the what’s needed to answer questions and data think about what’s needed to predict how that data’s going to change in the future machine learning. And if you think about, how do we make better decisions, how do we augment the human intelligence in that process AI? Very different things. There are four primary tools that I personally use that we often use on projects and certainly anybody that’s interested in this journey should take a look at. My first is this one right here and it’s MATLAB. MATLAB is a commercial product, It has a cost associated with it as most commercials products do. It ranges from a home version to an academic version to an enterprise version. So, If you are just starting this journey and you want to learn and build your skills up, they have a home version that is very expensive. During academia, during research similar, they have an academic license and of course, if you’re like most of us, and you’re trying to solve enterprise problems, there are other enterprise licenses. My belief is that while this is a pricey product, all those prices, the price of the system is completely offset by the benefits that it gives you. So why is it important to somebody like me? Well, this is a coded approach to solving problem, but they also have tool kits and applications, which has it as a no code approach, which is pretty cool. The tool boxes set the stage for all the things that we need to do, whether it’s from statistical analysis to graphical understanding and display. These tool boxes bring you into a specific environment where we can actually focus on the task at hand. On top of those tool boxes, we then have applications and here is the real power. Take a look at some of these applications that are right over here. Everything from signal processing to control systems, you have work in computational finance and code verification. A couple of my favorite ones are the deep neural network applications. This is a graphical way of designing, testing and deploying your deep neural networks. As you can see right here, we have a set of layers that we can bring into our system. That’s in this center. And then we have its properties on the far right hand side. In most procedural languages program languages. We’re doing this in code. We can do that pretty straightforward designing deep neural systems on a code base, whether you’re Python or any other programming languages. So what I like about this is the visual nature. And as a cognitive psychologist, there’s a difference when you type words and you see things. Textual versus visual, when we’re visually analyzing problem, we tend to be more creatively stimulated than we’re textually analyzing the problem. The second thing is if you’re ever in front of a client, and you’re trying to describe some of the characteristics of what you’re doing in this deep neural era, this is a brilliant way. Anybody can look at a picture like this and look at that stack. And as you go down through the layers explaining what each of those layers bringing in data, reformatting it, doing some functions in the center at the very end presenting a classification characteristics on how we’re classifying or identifying the items that we’re interested in. That’s a very visual way. Clients get that work. So I love it for the deep neural networks and deep neural network application being able to do. Build from scratch, do transfer learning and educate clients.

Dr. Jerry Smith: [05:41] And the second package I love in this, for all you folks that are in maintenance or have maintenance related missions, is the preventative maintenance and predictive failure model. You can see here, this is one of the classic visuals that we generate when we’re looking at why things fail, whether it’s a pump or a car or a airplane. What we’re interested in is, can we measure systems performance, and then can we predict remaining useful life for that product? The preventative maintenance application within MATLAB is brilliant, right? I mean, it’s just bringing in data, you’re doing your analysis, you’re cleaning up your poor quality information sources, and then you’re presenting it in a standardized preventative maintenance, algorithms that allow you to produce these results. So MATLAB is one of the first tools that I like.

Dr. Jerry Smith: [06:45] The second tool I like is one that it probably has not been used by a lot of data scientists, but I see it coming out. The first one MATLAB, miracle, miracle data analysis. The second one is MATLAB Wolfram Mathematica. Mathematica is a symbolic modeling environment. A lot of folks don’t play with symbolic modeling when it comes to data science, machine learning and AI. I find it useful anytime that we’re doing algorithm based development work. Now, one of the things that I like about Wolfram Mathematica is this deep computational capability that’s fundamentally built on solid mathematics. For example, I use it a lot when I’m studying quantum computing, quantum computing way over my head, but I’m fascinated by it, right? Reading articles on it, seeing, you know, what it’s capable of and not capable of. The good news, if you’re that sort of quantum computing high end math world, a lot of the quantum computing scientists and work that’s done in that space is done in Mathematica. I create math art. It’s kind of a fun thing to do in that particular space and creating math art which is an algorithmic capability, right? Some of the stuff that you see over here. Kind of fun to do it in Mathematica, it’s ability to generate the art through algorithms and then display, it produces stuff like this. One of the areas that Mathematica has been used is probably been seen by most of you. You ever saw this movie Arrival, which was a 2016 movie, I think it is. This linguistics professor played by Amy Adams as Lousie Banks, right? Leads her team of investigators as they go about exploring this interaction between these gigantic spaceships that touch down across the world. As all these nations sort of tear on the verge of this global war, this linguistics professor leads her sort of ragtag crew against time in order to find a way to communicate with these visitors.

Dr. Jerry Smith: [09:31] And you’re like, yeah okay Dr. Jerry, you know that I don’t have enough to get a Starbucks copy. What does this Arrival movie have to do anything? Well, take a look at this. This was the visual language that the aliens use in their communication. And that happens to be Amy Adams sitting there in the center trying to communicate with them. This visual language this logo ground, if you will, was created in Mathematica. Christopher Wolfram actually spends a couple hours discussing, I’ll post a link to this presentation that you see here, to the work that he did along with the slides that he has. So you can actually see, you can walk through and see how he creates this visual language and mathematics. It’s brilliant work. So it combines both linguistics and logo brand development and visualization in this mathematically based environment. The result of all this is they had a very, very consistent interaction with the aliens in a movie, right? This was just a movie, right? They could have easily just thrown up splatters of paint on the screen. But as you all know in today’s sort of nitpicky world where we analyze every phrasing claim to make sure that’s the correct phrase and usage of that term. You know that people would do the same thing with this. So they decide to sit down, engage Christopher and have him actually create a language. And actually you see this creation done in the movie itself. If nothing else, should you go grab a copy of mathematical, go watch this presentation and sit down and do it yourself. I think it has a set of skills that very few data scientists have. So my second one, my second tool that I really enjoyed is Mathematica for the reasons I have talked about.

Dr. Jerry Smith: [11:38] The last two are standards ones. The next ones are programming, RPR language in our studio. I use our studio extensively. I come at data science from a statistical perspective. I study data. I asked questions more from a statistical perspective. We know that data scientists come in sort of different flavors. We have the math flavor, the statistical flavor, we have the programming flavor, right? I come at it from a statistical and math, right? The R language was generated for people like me who are approaching that area. As you can see here, it’s an open source it’s inexpensive, it’s free, requires your time and energy for it. it comes with a lot of applications, I need to take a look at it, but I think there are 6,000 packages that are now publicly available. It covers just about every realm that every data scientist needs in this particular space. So our program is a great language and a great ecosystem to work in.

Dr. Jerry Smith: [12:46] My last one is this one right here, which is Python. Python is the set of capabilities that a data science needs. Technically coming more from a programming world where people or computer scientists who are used to programming have entered in through Python and begin to discover that they actually develop some interesting viewpoints around data science, machine learning and AI from that space. One of the biggest challenges we have is in this sort of world of R versus Python, right? Even in my organization with the people I work with, we’re always teasing each other on whether R or Python is a greater language. The truth of the matter, this one is that both R and Python are equally capable. You know, a Batman versus a Superman, if you will. They’re both able to provide the kind of superpowers that one would need in solving a lot of problems today. The way I look at it when making a decision on where I am going to start in this world, if my goal is to interrogate data from a math or statistics, I’m looking at it from Python. I mean, I’m looking at it from R, I apologize. doing a Jedi tricks on my mind. I look at it from art perspective, and sometimes I’ll pull MATLAB and I’ll pull Mathematica. If I know my work is going to end up at an enterprise level, where it will actually migrate from some proof of concept, proof of value models and pile it into an enterprise level. I might do some creative discovery just to get a general understanding because I’m comfortable there, but it’ll eventually end up very quickly into the Python. Now, R be the enterprise level. I know there’s going to be a lot of peoples sending me comments and saying, hey Dr. Jerry, R can be run just as well. It’s not about R being run at the enterprise level, it’s about building enterprise applications. R is not an environment rebuild enterprise applications, right? Python isn’t an ecosystem. It comes from that perspective. So if the four tools that I like or MATLAB it’s just a brilliant toolbox and an application, just check out the applications. That’s real engineering approach to data science, machine learning, AI. Mathematica love it. Symbolic modeling try to follow Christopher’s work when it comes to system analysis. And then the last two R and Python. If you don’t mind, let me know what you prefer, are you an R person or you’re Python person. Comments below, I provided a sort of a list of Batman versus Superman characteristics. There are probably an infinite number of lists that are out there to overexaggerate. I would like to hear your opinion, when do you use R when you do you use Python?

Dr. Jerry Smith: [15:54] Okay. Let’s move on to the next question. The next question is Dr. Jerry, do you have any recommended books for causality. Causality is a really important part of our process. Over here is our circle of life. That circle of life represents the very first step is grabbing our data that’s in there. Data acquisition’s an important part of the real world phenomenon. In that data acquisition after that, what do we do? We go through a set of cognitive services to extract that behavioral characteristics. I talked about that explicitly in another podcast. After we do those cognitive services, grabbing behavioral data out of it, the next thing we want to do in order to change the world and not just observe it, is to look for the causal drivers that drive the concerns that are under investigation, revenue, growth, margin growth, customer satisfaction, making people healthy or happy or selling more. Whatever that set of outcomes that you want to achieve. In order to change them, we have to identify what are the causal drivers. So causal is important, just to finish the process, after we get those causal drivers we build predictors, right? How is this data going to change in the future? Now, once you develop predictors based on causal drivers, we call that special name, we call that a digital surrogate that digital surrogate is designed to represent the entity in life, whether from a simulation or emulation. Simulation is a set of code that achieves for a set of inputs, a set of outputs. Most of your machine learning models are simulations of some sort of function in there versus an emulation which actually breaks down the functionality in a way that mimics the process that we’re doing in this case speaking the mind. Many of the deep neural systems you see today are mimicking a lot of the cognitive processes we have. Again, that the 1.25 still can’t get a Starbucks coffee out there, but it allows us to sort of think through a little bit deeper. What can we do? What are the limits of the approaches that we’re doing when we’re building a digital surrogate? And finally, once we have that digital surrogate, once we have its inputs and then associated outputs, we can optimize it. We can use things like evolutionary computing, which is more or less a functional quantum computer, which is why I’m interested in quantum computing, the ability to search a large search base very effectively, right? Both in terms of what’s called population based intelligence and emergence based intelligence population is things like generic algorithms with Dr. Holland, and that is taking populations of things and combining them together to form a next evolution of generation that is more beneficial to the goal that we have versus emergent behavior, which takes a bunch of simplistic characteristics of like for example, ants in large number. And they come together to formulate a behavior that emerges from the interaction. Both of those are very important but when we’re looking to identify under under what input conditions, under what causal input conditions will we maximize the results that we want to achieve. We use evolutionary computing, we use algorithms to do that.

Dr. Jerry Smith: [19:37] So back to the question, do you have any on causality? Causality Is the third step in our process that is very important, and there are several book that we should talk about. The first one which if you actually look out into the world you’ll see that Judea Pearl is the godfather of causality. You’ll see the very first book I like is the Book of Why. The Book of Why is foundational for both the business person and the data scientist, right? Somebody coming from very little background in math and to someone with a lot of background in math, right? The Book of Why sets the conditions through historical review of what causality is, and why it’s important and then actually begins to bring you into the inference latter that takes us from the observation to the ability to change the world. Judea is a brilliant guy in this case. I encourage you, if you do nothing else read this particular book. If you like that book, and you’re interested in getting a deeper understanding of causal inferencing. This book right here, Causal Inference in Statistics: A Primer by Judea, Glymour, and Jewell is a very deep understanding of this world. It’s a very deep mathematical understanding. This is a kind book that you don’t read and operate equipment on, because it’s mentally intense and it’s going to require a lot of attention and focus, but it’s another great partner as well. Finally, as you get into causality, you’re going to come across counterfactuals. Really quick in your causal inferencing in and causal counterfactuals allows us to test the legitimacy of causal drivers. It allows us to answer the question, what would’ve happened if that world didn’t exist? Think of that as sort of a primer to the multiverse in this particular case. This book right here, Counterfactuals and Causal Inferences, is a great source of information by Winship and Morgan when it comes to thinking about counterfactuals analysis. Now, most of what we need to know can be done by the previous work by Judea on the Book of Why, but if you’re one of those developers of programs, causality programs, like you want to build them from scratch, you really want to get into counterfactuals. While we’re on the subject of books and even though the question around causality, I often ask the question, why are you interested in causality? that leads us to this notion of I want to change the world and not just observe it. That’s really important thing, I want to change the world. You can’t change the world unless you are working with causal. Dr. Jerry Smith: [22:22] When we realize that we want to change the world, we come across the next big thing. Well, what kind of world we want to change? Human world, right? I mean, that’s the premise of the argument is rooted the fact that we want to change human beings, right? When we look at data because we’re in data science, machine learning, and AI. Well, what is data, right? Data is the debris of human activity. It’s because of us, not in spite of us, right? Data does not spontaneously generate in some sort of database log system show up on a network somewhere. It’s the result of some sort of a long causal chain that starts in some human activity. So since we’re talking about human beings and we’re talking about data, it’s natural to begin to think about behavioral data analytics. And in this case, I want to point you to a great book, Behavioral Data Analysis with R and Python. This is a awesome book for the practitioner who wants to learn about what behaviors are, the definition of those are, how they show up in data and also the causal characteristics. And then how to extract behavioral insights from data using your fingers. Coding in R and Python, Florent work is really good in that space. The second book, this one right here, Designing for Behavioral Change by Stephen Wendel talks about the three main strategies that we use to help people change their behavior. So now you’re like, okay, I can almost get that first book. I could almost get the book that what we’re talking about of behavioral analytics and extracting behaviors out of data makes sense to me, but why should I be worried about changing people’s behavior, right? Why should I worry about developing effective designs that are enjoyable for people to use? Or why should I be concerned about how do we combine behavioral science and data science to pinpoint problems and test these potential solutions? Well, if we’re interested in changing the world, changing human behavior, having people partake in healthier outcomes, having people buy your product versus somebody else’s problem, identifying bad behavior, threats to our system, identifying potentially new customers versus old customers, competitive analysis. This all requires a behavioral insights, but also since we’re going to change the world, strategies that allows us to actually change behavior. So designing for behavioral change is typically used by Digital Psychologists or Digital Sociologist. The people on your teams that are responsible for understanding why we do what we do up here, right? Digital psychologists and digital sociologists, those people that say, why do people do things in groups, group dynamics. You don’t necessarily need both in your team, but if you are really serious about changing the world and not just observing it, you should have somebody in your organization capable of talking in this language and understanding it. There’s this somewhere in here, which is on Behaviology. It’s a new book by Eddie Rafii, it identifies the seven rules of human behavior. It goes along the previous book that I talked about. It defines things like why do we want to be comfortable? And we move when we lose our balance? These are all frameworks for us to begin to think about how we can deploy our solutions once we go through the data and cognitive analysis and the causality and the digital surrogate and we optimize. You have to begin to think, well, how we going to take those inputs and deploy into a world where it is a marketing solution, IT, operations solutions, sales, pharmacy and nutraceuticals and that sort of thing. How are we going to deploy these things? These are the rules that we can then shape those optimize inputs into real operational programs. So those are the books I like both in terms of causality and also the books in terms of human behavior that I think should be shaping the world when it comes to data science, machine learning and AI.

Dr. Jerry Smith: [26:57] Let’s move to area three. Dr. Jerry, what tools do you use for causal analysis? Well, another wonderful question, right? We, if causality is our third step in the process and it’s super, super important to us, we should have a set of tools that allows us to work in this space effectively. I have three tools that I love. Each of those are a lot like the previous ones but are very different. Probably haven’t heard of a lot them, but certainly they provide us with a set of capabilities that can be deployed from people who are in the analytical space, people who are in the AI space. First one here is one of my favorite longtime tools, the tools called [Unknown name] I’ve been working with this company for a long time. I’ve seen them go from a simple set of causal capabilities, bringing in your spreadsheet that is in the machine learning format, rows of observations with columns of variable data, you know, tens of thousands of rows, tens to hundreds of variables. I’ve seen them in the early stages being, bring that in, figure out what’s causal to the stage they’re in today, we’ll get into in just a second. So COMPAL allows us to do, performance analysis that says, here’s the fundamental causal change in your data world, right? It’ll show you how each individual, leaf node affects the elements above it, and then aggregates into your, your target of interest. You know, for example, you’re interested in student grades, what drives student grades? If you’re interested in breast cancer, what drives breast cancer? If you in sales, what drives sales? It does that, it does that brilliantly, blazingly fast to see it done. It could be used by anyone who can put together a spreadsheet define those columns variables, put in rows of observation, missing data or not, who cares, you upload it, it produces a nice graph, awesome stuff. I encourage you look at it.

Dr. Jerry Smith: [29:17] The other that it does, as it advances overtime, it’s no code. Data, tools, results. The other thing it does is allows us to do this sort of what-if-analysis, this advisory capability, which says, you know, I know my sales are driven by these factors, but what if I wanted my sales to go up another 10%? How would those factors have to change? And it does that. So we have this advisory component, all of this then in the end, the third thing it does that I really like is you can express it in terms of a dockerized container and deploy in your enterprise. Right? You can take this from research to enterprise very, very quickly. So I love [COMPALA] it’s one of easy tools, I actually use it sometimes with my wife who’s in real estate. And she’s trying to figure out what’s causing certain kind of real estate activities that are going on. She brings me a bunch of data. Five minutes later, I got a diagram that I can show her. She goes, oh yeah, that makes sense, pretty cool stuff. So first one is [COMPELA], now [COMPALA] is a commercial product so you’re going to pay for it, right? You get what you pay for in a lot of cases, in this case, you get a lot in this area. The next one is an open source tool it’s called Causelnex. Causelnex is an open source package its in Python, so if you are a Python programmer and you are comfortable PyTorch and things like that, very comfortable in that space, it does things that [COMPALA] doesn’t, it does a much more deeper look at the causal models rather than having causal links. It actually does causal meshes, right? It’ll look at inner relationships between variables, not just in how those variables affect a chain of variables along an area. So it’s really useful in finding confounders in the area and deeper insights that one would want to have, if you read the Book of Why, you’ll see some of those capabilities in terms of different causal networks that comes out of it.

Dr. Jerry Smith: [31:22] It’s also capable of allowing you to augment your knowledge into there. You have an expert that says, you know, we didn’t have the data for this, but I happen to know that A is causing B and, and B is causing this variable over here. You can add that to that system. My last area today is Bayesia Labs, Bayesia Labs is another one of those commercial tools. It’s a brilliant tool, right? this tool is one of the higher end tools that an organization can have. It is, again, a commercial tool, it is no code based tool. It’s a data tool and allows us to do work in causal networks and counterfactual reason. This tool is used a lot in artificial intelligence, research analytics and reasoning area allows us to ingest that data, determine the probabilistic models, the probability models that exist between data understand causal relationships and mutual information in that particular space, and then allows us to do that What if analysis, what if we observe this data to be this, how is the output effective. Not only across the variable of interest, but all the other variables as well. I love this tool. I love it from an enterprise, engineering perspective. And for those organizations who can afford this tool, I encourage you to bring it in and actually design team around it to take advantage of its capabilities. So we have the three tools I like in the causality space are those areas.

Dr. Jerry Smith: [33:47] So we’re into our last question. Dr. Jerry, What skills are needed for data science, machine learning and AI developers? So this is a little different than the tool question, certainly consistent in kind of a way with questions. Its a great question It’s a question that we can probably spend hours talking about in details. But there is a way to talk about this that would give us most of the benefit very, very quickly. And that’s what this diagram is here. This is our capabilities model. Again, I’ll leave you a link the description for a more high end graphic for this, but this graphic, is designed, was designed, to help us address that issue. You’ll see right away to sort of orientation on very far left hand side, you’ll see data sources, our enterprise data sources, our IT data sources, our open data sources or our third party data sources, right? We’re either getting stuff from our enterprise about what’s concerning us about our customers, our IT interactions, on the right hand side, that’s our target, what kind insights can we have? What kind of actions that we want to do again, we want to change the world and not just report on it, right? So that’s the bookends in between our set of capabilities that are in there. Those capabilities are segmented into data oriented capabilities, information oriented capabilities, knowledge and intelligence capabilities, right? Those capabilities are then used by roles, roles like data scientists, data engineers, platform engineers, people who are actually working with the tools and technologies and visualization specialists, right? And those roles are aggregated into different levels of maturity, whether you’re a level one junior person or a level five or level four, very senior person. And then on the very far right hand side, you’ll see some references. There’s some tools which we’ve already talked, most of which are on there, and there are few that we didn’t talk about. So the idea is this, the idea is, can you take your data and then connect it to various capability boxes in order to get the insights.

Dr. Jerry Smith: [36:07] Before I give you an example or two, let tell you how this started. It was about 15, almost 20 years. And we were doing work in this particular area, and I had my data science team around. We had noticed that a lot of our projects were failing, not failing because we didn’t know what we were doing, but failing to meet the goals, you know, they weren’t producing revenue. They weren’t achieving the margin. They weren’t being done on time. There were some cost over went. And I thought to myself, you know, we’re a group of smart people. We have some great programmers, some great data scientists on the project. Why are we having such trouble? So we got together over course of week, we sort of taken a stand down day. And we lied out lots of projects up on the board and we talked about each of those 20 30 projects, and we started to group them, they ones are done really well. And these had produced the results, but didn’t achieve it in the business benefits we wanted. We began to look at that and we began to ask questions, well, how did we do this one? And we, when we talk about it, we realized right away the ones that didn’t work out well started with an understanding and we eventually said, oh, we should have done X. The ones that we did really well were done with the idea that ended up being used successfully on the project. So the missing ingredient here was we were missing a set of capabilities. And when we looked at that, we said we didn’t have the right team members for that one. And once we brought in the right team, they brought in their skills and the behold the project worked out.

Dr. Jerry Smith: [37:41] So the idea here was, could we identify all the capabilities across both the projects at work and the project didn’t work and organize them in a way that allowed us to systematically move from data to information to knowledge, to intelligence wisdom. Structurally, not so that would guarantee that we’d be successful just so that we wouldn’t say we fail because we forgot to do something. So that’s what this capability model is about. The idea is this, we take your data on the left hand side, you look at the first column and you draw the dots on there that say, I’m going to do these things, and you draw a line and you keep going progressively to each column asking the question, what in this area do I need? So a lot of our work today is enterprise work, API harvesting, doing some metadata analysis, natural language analysis, cognitive services, doing that cognitive analysis on it, going into causal analysis, out of that building out some sort of anomaly detection or machine learning algorithm, genetically optimized moved into an engine. That’s a traditional path that we see on 75% of our projects. We feel comfortable with that today. One of the benefits that we ended up doing, having with this is, somebody was talking one day, one of our engineers. It’s a story. You give subjects, places, color, stuff like that. You feel it, then you read it. It’s hilarious. Kids love it. Well, we said, well, could we do Madlib, could we do this Madlib thing with our capabilities? If we took a random set of, we took one of the three inputs, the beginning, and one of the inputs and the next across the board, just randomly put ’em together and ask the question. What would that give, would that be a benefit? So we would do that. We’d say, what if we took this enterprise data and we did some text mining on it. And then we ended up doing a design of experiments around that, that then influenced our cognitive analysis that we did feature matching on that we then put into a decision from which we did some, evolutionary program that yourself, that’s we get if with that, that we would not, if we didn’t do that, that doesn’t mean you go do the project. Now it turns out we ended up doing a lot of these projects and we going, wow, look at that. We’re getting some pretty cool results. So the point in this is that this framework allows us to identify a set of capabilities, set of roles in an organized ways that move some data to insights, right?

Dr. Jerry Smith: [40:41] And so, you know, this world is very complex. There’s a lot of other skills you need in there. There’s some, there’s some technology skills. And certainly there’s soft skills in there. When I think about the fundamental, capabilities that projects needs to not these of some of the ones. All right. So with that said today, we covered a lot of stuff. Just sort of scrolling back up to the top of my notes. We covered a set of tools that I think are important to the group. Everything from MATLAB to Mathematica into R and Python know what you think about each of those into several books on causality, from the book of why to causal in to counteracts. We touched a bit on behavioral part, the work done by, foreign in his behavioral data and analysis with arm Python, along with design for behavioral change.

Dr. Jerry Smith: [41:36] We also talked a bit about behavioral, which is kind of a cool new emerging book regard into some of the tools for causality, the compel on the causal, even the Beisya labs work that’s in there and skills capabilities, and the data science AI space. I truly hope that this was, this was useful to you. You found value. And if you did, please leave comments below. If you didn’t leave comments below too, tell me how we can improve this work. I enjoy bringing this kinda information to you, but I, I really enjoy when I hear back. So that’s going to be it for the day. I’m Dr. Jerry, this is AI Live and Unbiased until next time let’s get after some data sciences. We’ll see you later on,

Outro: [42:25] This has been the AI Live and Unbiased podcast brought to you by AgileThought. The views, opinions and information expressed in this podcast are solely those of the host and the guests, and do not necessarily represent those of agile thought. Get the show notes and other helpful tips for this episode and other episodes at agilethought.com/podcast.

Speakers

Dr. Jerry Smith

(Panelist), Managing Director, Global Analytics & Data Insights, AgileThought

Dr. Smith is a practicing AI & Data Scientist, Thought Leader, Innovator, Speaker, Author, and Philanthropist dedicated to advancing and transforming businesses through evolutionary computing, enterprise AI and data sciences, machine learning, and causal AI.

He’s presented at Gartner Conferences, CIO, SalesForce, DreamForce, and the World Pharma Congress, and is often who leaders turn to for help with developing practical methods for unifying insights and data to solve real problems facing today’s businesses.

10 Things a Java Developer Should Learn These Days

We are halfway into the year and, looking back at what the last 2 years of the COVID-19 pandemic left us, we can say that the advancements on technology and all the different languages, frameworks and development tools have not disappointed us and even in many cases have overtaken us. It doesn’t seem that long since Java 8 was released more than 8 years ago and, although it is true that today the latest version is Java 18, we must remember that there are still many companies stuck with Java 8 and even a few with Java 1.4 a 1.6 and a few other legacy frameworks.Because of this great advance in programming languages and frameworks in recent years, many programmers could feel overwhelmed due to the large number of updates that we are exposed to in our industry. However, we have begun to see very marked patterns in different industries, and we believe it is important to mention what are the 10 things that every Java programmer should learn.

6 min read

August 25, 2022

Succeeding at Cost Takeout – Improving Cost Management Strategy

The cost takeout is a strategic initiative that can help your company become more efficient. The goal of the initiative is to determine ways to reduce expenses without sacrificing quality, speed or customer satisfaction. A successful cost takeout requires an accurate understanding of where costs arise in your business process and how those costs can be reduced through new technologies. In this guide, we’ll look at some of the most common drivers of costs in IT departments and explain how you can identify them within your own organization’s processes.Cost drivers are the factors that cause your costs to increase. They’re different for every business, but you can identify them by asking yourself: “What makes my costs go up?”

10 min read

February 7, 2023

AI Podcast Ep. 4: Four Most Commonly Asked Questions About AI with Dr. Jerry Smith

Intro: [00:04]

You’re listening to AI Live and Unbiased, the podcast where we explore how to apply artificial intelligence and the impact it has on enterprises and society. Now, here’s your host, Dr. Jerry Smith.

Dr. Jerry Smith: [00:19]

Welcome to AI Live and Unbiased. I’m your host, Dr. Jerry, I’ll be with you on this journey as we explore the breadth and depth of artificial intelligence. Today, we’re going to talk questions and answers in the world of data science, machine learning, artificial intelligence. If I get crazy enough we may even explore the meaning of why itself and no, I don’t think the answer is 42. Today, I’ve decided to spend some time talking about four primary questions that are often raised. I see these a lot coming my way almost every conversation, whether it’s with a colleague or a client that’s out there. These questions range from tools to skills to that books that I read. So I hope some of these interest you and if they do please comments on the ones that interest you. And if you think I could add to some of these things, please leave comments as well in the comments. Dr. Jerry Smith: [01:14

Dr. Jerry Smith: [05:41]

And the second package I love in this, for all you folks that are in maintenance or have maintenance related missions, is the preventative maintenance and predictive failure model. You can see here, this is one of the classic visuals that we generate when we’re looking at why things fail, whether it’s a pump or a car or a airplane. What we’re interested in is, can we measure systems performance, and then can we predict remaining useful life for that product? The preventative maintenance application within MATLAB is brilliant, right? I mean, it’s just bringing in data, you’re doing your analysis, you’re cleaning up your poor quality information sources, and then you’re presenting it in a standardized preventative maintenance, algorithms that allow you to produce these results. So MATLAB is one of the first tools that I like.

Dr. Jerry Smith: [06:45]

The second tool I like is one that it probably has not been used by a lot of data scientists, but I see it coming out. The first one MATLAB, miracle, miracle data analysis. The second one is MATLAB Wolfram Mathematica. Mathematica is a symbolic modeling environment. A lot of folks don’t play with symbolic modeling when it comes to data science, machine learning and AI. I find it useful anytime that we’re doing algorithm based development work. Now, one of the things that I like about Wolfram Mathematica is this deep computational capability that’s fundamentally built on solid mathematics. For example, I use it a lot when I’m studying quantum computing, quantum computing way over my head, but I’m fascinated by it, right? Reading articles on it, seeing, you know, what it’s capable of and not capable of. The good news, if you’re that sort of quantum computing high end math world, a lot of the quantum computing scientists and work that’s done in that space is done in Mathematica. I create math art. It’s kind of a fun thing to do in that particular space and creating math art which is an algorithmic capability, right? Some of the stuff that you see over here. Kind of fun to do it in Mathematica, it’s ability to generate the art through algorithms and then display, it produces stuff like this. One of the areas that Mathematica has been used is probably been seen by most of you. You ever saw this movie Arrival, which was a 2016 movie, I think it is. This linguistics professor played by Amy Adams as Lousie Banks, right? Leads her team of investigators as they go about exploring this interaction between these gigantic spaceships that touch down across the world. As all these nations sort of tear on the verge of this global war, this linguistics professor leads her sort of ragtag crew against time in order to find a way to communicate with these visitors.

Dr. Jerry Smith: [09:31]

And you’re like, yeah okay Dr. Jerry, you know that I don’t have enough to get a Starbucks copy. What does this Arrival movie have to do anything? Well, take a look at this. This was the visual language that the aliens use in their communication. And that happens to be Amy Adams sitting there in the center trying to communicate with them. This visual language this logo ground, if you will, was created in Mathematica. Christopher Wolfram actually spends a couple hours discussing, I’ll post a link to this presentation that you see here, to the work that he did along with the slides that he has. So you can actually see, you can walk through and see how he creates this visual language and mathematics. It’s brilliant work. So it combines both linguistics and logo brand development and visualization in this mathematically based environment. The result of all this is they had a very, very consistent interaction with the aliens in a movie, right? This was just a movie, right? They could have easily just thrown up splatters of paint on the screen. But as you all know in today’s sort of nitpicky world where we analyze every phrasing claim to make sure that’s the correct phrase and usage of that term. You know that people would do the same thing with this. So they decide to sit down, engage Christopher and have him actually create a language. And actually you see this creation done in the movie itself. If nothing else, should you go grab a copy of mathematical, go watch this presentation and sit down and do it yourself. I think it has a set of skills that very few data scientists have. So my second one, my second tool that I really enjoyed is Mathematica for the reasons I have talked about.

Dr. Jerry Smith: [11:38]

The last two are standards ones. The next ones are programming, RPR language in our studio. I use our studio extensively. I come at data science from a statistical perspective. I study data. I asked questions more from a statistical perspective. We know that data scientists come in sort of different flavors. We have the math flavor, the statistical flavor, we have the programming flavor, right? I come at it from a statistical and math, right? The R language was generated for people like me who are approaching that area. As you can see here, it’s an open source it’s inexpensive, it’s free, requires your time and energy for it. it comes with a lot of applications, I need to take a look at it, but I think there are 6,000 packages that are now publicly available. It covers just about every realm that every data scientist needs in this particular space. So our program is a great language and a great ecosystem to work in.

Dr. Jerry Smith: [12:46]

My last one is this one right here, which is Python. Python is the set of capabilities that a data science needs. Technically coming more from a programming world where people or computer scientists who are used to programming have entered in through Python and begin to discover that they actually develop some interesting viewpoints around data science, machine learning and AI from that space. One of the biggest challenges we have is in this sort of world of R versus Python, right? Even in my organization with the people I work with, we’re always teasing each other on whether R or Python is a greater language. The truth of the matter, this one is that both R and Python are equally capable. You know, a Batman versus a Superman, if you will. They’re both able to provide the kind of superpowers that one would need in solving a lot of problems today. The way I look at it when making a decision on where I am going to start in this world, if my goal is to interrogate data from a math or statistics, I’m looking at it from Python. I mean, I’m looking at it from R, I apologize. doing a Jedi tricks on my mind. I look at it from art perspective, and sometimes I’ll pull MATLAB and I’ll pull Mathematica. If I know my work is going to end up at an enterprise level, where it will actually migrate from some proof of concept, proof of value models and pile it into an enterprise level. I might do some creative discovery just to get a general understanding because I’m comfortable there, but it’ll eventually end up very quickly into the Python. Now, R be the enterprise level. I know there’s going to be a lot of peoples sending me comments and saying, hey Dr. Jerry, R can be run just as well. It’s not about R being run at the enterprise level, it’s about building enterprise applications. R is not an environment rebuild enterprise applications, right? Python isn’t an ecosystem. It comes from that perspective. So if the four tools that I like or MATLAB it’s just a brilliant toolbox and an application, just check out the applications. That’s real engineering approach to data science, machine learning, AI. Mathematica love it. Symbolic modeling try to follow Christopher’s work when it comes to system analysis. And then the last two R and Python. If you don’t mind, let me know what you prefer, are you an R person or you’re Python person. Comments below, I provided a sort of a list of Batman versus Superman characteristics. There are probably an infinite number of lists that are out there to overexaggerate. I would like to hear your opinion, when do you use R when you do you use Python?

Dr. Jerry Smith: [15:54]

Okay. Let’s move on to the next question. The next question is Dr. Jerry, do you have any recommended books for causality. Causality is a really important part of our process. Over here is our circle of life. That circle of life represents the very first step is grabbing our data that’s in there. Data acquisition’s an important part of the real world phenomenon. In that data acquisition after that, what do we do? We go through a set of cognitive services to extract that behavioral characteristics. I talked about that explicitly in another podcast. After we do those cognitive services, grabbing behavioral data out of it, the next thing we want to do in order to change the world and not just observe it, is to look for the causal drivers that drive the concerns that are under investigation, revenue, growth, margin growth, customer satisfaction, making people healthy or happy or selling more. Whatever that set of outcomes that you want to achieve. In order to change them, we have to identify what are the causal drivers. So causal is important, just to finish the process, after we get those causal drivers we build predictors, right? How is this data going to change in the future? Now, once you develop predictors based on causal drivers, we call that special name, we call that a digital surrogate that digital surrogate is designed to represent the entity in life, whether from a simulation or emulation. Simulation is a set of code that achieves for a set of inputs, a set of outputs. Most of your machine learning models are simulations of some sort of function in there versus an emulation which actually breaks down the functionality in a way that mimics the process that we’re doing in this case speaking the mind. Many of the deep neural systems you see today are mimicking a lot of the cognitive processes we have. Again, that the 1.25 still can’t get a Starbucks coffee out there, but it allows us to sort of think through a little bit deeper. What can we do? What are the limits of the approaches that we’re doing when we’re building a digital surrogate? And finally, once we have that digital surrogate, once we have its inputs and then associated outputs, we can optimize it. We can use things like evolutionary computing, which is more or less a functional quantum computer, which is why I’m interested in quantum computing, the ability to search a large search base very effectively, right? Both in terms of what’s called population based intelligence and emergence based intelligence population is things like generic algorithms with Dr. Holland, and that is taking populations of things and combining them together to form a next evolution of generation that is more beneficial to the goal that we have versus emergent behavior, which takes a bunch of simplistic characteristics of like for example, ants in large number. And they come together to formulate a behavior that emerges from the interaction. Both of those are very important but when we’re looking to identify under under what input conditions, under what causal input conditions will we maximize the results that we want to achieve. We use evolutionary computing, we use algorithms to do that.

Dr. Jerry Smith: [19:37]

So back to the question, do you have any on causality? Causality Is the third step in our process that is very important, and there are several book that we should talk about. The first one which if you actually look out into the world you’ll see that Judea Pearl is the godfather of causality. You’ll see the very first book I like is the Book of Why. The Book of Why is foundational for both the business person and the data scientist, right? Somebody coming from very little background in math and to someone with a lot of background in math, right? The Book of Why sets the conditions through historical review of what causality is, and why it’s important and then actually begins to bring you into the inference latter that takes us from the observation to the ability to change the world. Judea is a brilliant guy in this case. I encourage you, if you do nothing else read this particular book. If you like that book, and you’re interested in getting a deeper understanding of causal inferencing. This book right here, Causal Inference in Statistics: A Primer by Judea, Glymour, and Jewell is a very deep understanding of this world. It’s a very deep mathematical understanding. This is a kind book that you don’t read and operate equipment on, because it’s mentally intense and it’s going to require a lot of attention and focus, but it’s another great partner as well. Finally, as you get into causality, you’re going to come across counterfactuals. Really quick in your causal inferencing in and causal counterfactuals allows us to test the legitimacy of causal drivers. It allows us to answer the question, what would’ve happened if that world didn’t exist? Think of that as sort of a primer to the multiverse in this particular case. This book right here, Counterfactuals and Causal Inferences, is a great source of information by Winship and Morgan when it comes to thinking about counterfactuals analysis. Now, most of what we need to know can be done by the previous work by Judea on the Book of Why, but if you’re one of those developers of programs, causality programs, like you want to build them from scratch, you really want to get into counterfactuals. While we’re on the subject of books and even though the question around causality, I often ask the question, why are you interested in causality? that leads us to this notion of I want to change the world and not just observe it. That’s really important thing, I want to change the world. You can’t change the world unless you are working with causal. Dr. Jerry Smith: [22:22

Dr. Jerry Smith: [26:57]

Let’s move to area three. Dr. Jerry, what tools do you use for causal analysis? Well, another wonderful question, right? We, if causality is our third step in the process and it’s super, super important to us, we should have a set of tools that allows us to work in this space effectively. I have three tools that I love. Each of those are a lot like the previous ones but are very different. Probably haven’t heard of a lot them, but certainly they provide us with a set of capabilities that can be deployed from people who are in the analytical space, people who are in the AI space. First one here is one of my favorite longtime tools, the tools called [Unknown name

Dr. Jerry Smith: [29:17]

The other that it does, as it advances overtime, it’s no code. Data, tools, results. The other thing it does is allows us to do this sort of what-if-analysis, this advisory capability, which says, you know, I know my sales are driven by these factors, but what if I wanted my sales to go up another 10%? How would those factors have to change? And it does that. So we have this advisory component, all of this then in the end, the third thing it does that I really like is you can express it in terms of a dockerized container and deploy in your enterprise. Right? You can take this from research to enterprise very, very quickly. So I love [COMPALA

Dr. Jerry Smith: [31:22]

It’s also capable of allowing you to augment your knowledge into there. You have an expert that says, you know, we didn’t have the data for this, but I happen to know that A is causing B and, and B is causing this variable over here. You can add that to that system. My last area today is Bayesia Labs, Bayesia Labs is another one of those commercial tools. It’s a brilliant tool, right? this tool is one of the higher end tools that an organization can have. It is, again, a commercial tool, it is no code based tool. It’s a data tool and allows us to do work in causal networks and counterfactual reason. This tool is used a lot in artificial intelligence, research analytics and reasoning area allows us to ingest that data, determine the probabilistic models, the probability models that exist between data understand causal relationships and mutual information in that particular space, and then allows us to do that What if analysis, what if we observe this data to be this, how is the output effective. Not only across the variable of interest, but all the other variables as well. I love this tool. I love it from an enterprise, engineering perspective. And for those organizations who can afford this tool, I encourage you to bring it in and actually design team around it to take advantage of its capabilities. So we have the three tools I like in the causality space are those areas.

Dr. Jerry Smith: [33:47]

So we’re into our last question. Dr. Jerry, What skills are needed for data science, machine learning and AI developers? So this is a little different than the tool question, certainly consistent in kind of a way with questions. Its a great question It’s a question that we can probably spend hours talking about in details. But there is a way to talk about this that would give us most of the benefit very, very quickly. And that’s what this diagram is here. This is our capabilities model. Again, I’ll leave you a link the description for a more high end graphic for this, but this graphic, is designed, was designed, to help us address that issue. You’ll see right away to sort of orientation on very far left hand side, you’ll see data sources, our enterprise data sources, our IT data sources, our open data sources or our third party data sources, right? We’re either getting stuff from our enterprise about what’s concerning us about our customers, our IT interactions, on the right hand side, that’s our target, what kind insights can we have? What kind of actions that we want to do again, we want to change the world and not just report on it, right? So that’s the bookends in between our set of capabilities that are in there. Those capabilities are segmented into data oriented capabilities, information oriented capabilities, knowledge and intelligence capabilities, right? Those capabilities are then used by roles, roles like data scientists, data engineers, platform engineers, people who are actually working with the tools and technologies and visualization specialists, right? And those roles are aggregated into different levels of maturity, whether you’re a level one junior person or a level five or level four, very senior person. And then on the very far right hand side, you’ll see some references. There’s some tools which we’ve already talked, most of which are on there, and there are few that we didn’t talk about. So the idea is this, the idea is, can you take your data and then connect it to various capability boxes in order to get the insights.

Dr. Jerry Smith: [36:07]

Before I give you an example or two, let tell you how this started. It was about 15, almost 20 years. And we were doing work in this particular area, and I had my data science team around. We had noticed that a lot of our projects were failing, not failing because we didn’t know what we were doing, but failing to meet the goals, you know, they weren’t producing revenue. They weren’t achieving the margin. They weren’t being done on time. There were some cost over went. And I thought to myself, you know, we’re a group of smart people. We have some great programmers, some great data scientists on the project. Why are we having such trouble? So we got together over course of week, we sort of taken a stand down day. And we lied out lots of projects up on the board and we talked about each of those 20 30 projects, and we started to group them, they ones are done really well. And these had produced the results, but didn’t achieve it in the business benefits we wanted. We began to look at that and we began to ask questions, well, how did we do this one? And we, when we talk about it, we realized right away the ones that didn’t work out well started with an understanding and we eventually said, oh, we should have done X. The ones that we did really well were done with the idea that ended up being used successfully on the project. So the missing ingredient here was we were missing a set of capabilities. And when we looked at that, we said we didn’t have the right team members for that one. And once we brought in the right team, they brought in their skills and the behold the project worked out.

Dr. Jerry Smith: [37:41]

So the idea here was, could we identify all the capabilities across both the projects at work and the project didn’t work and organize them in a way that allowed us to systematically move from data to information to knowledge, to intelligence wisdom. Structurally, not so that would guarantee that we’d be successful just so that we wouldn’t say we fail because we forgot to do something. So that’s what this capability model is about. The idea is this, we take your data on the left hand side, you look at the first column and you draw the dots on there that say, I’m going to do these things, and you draw a line and you keep going progressively to each column asking the question, what in this area do I need? So a lot of our work today is enterprise work, API harvesting, doing some metadata analysis, natural language analysis, cognitive services, doing that cognitive analysis on it, going into causal analysis, out of that building out some sort of anomaly detection or machine learning algorithm, genetically optimized moved into an engine. That’s a traditional path that we see on 75% of our projects. We feel comfortable with that today. One of the benefits that we ended up doing, having with this is, somebody was talking one day, one of our engineers. It’s a story. You give subjects, places, color, stuff like that. You feel it, then you read it. It’s hilarious. Kids love it. Well, we said, well, could we do Madlib, could we do this Madlib thing with our capabilities? If we took a random set of, we took one of the three inputs, the beginning, and one of the inputs and the next across the board, just randomly put ’em together and ask the question. What would that give, would that be a benefit? So we would do that. We’d say, what if we took this enterprise data and we did some text mining on it. And then we ended up doing a design of experiments around that, that then influenced our cognitive analysis that we did feature matching on that we then put into a decision from which we did some, evolutionary program that yourself, that’s we get if with that, that we would not, if we didn’t do that, that doesn’t mean you go do the project. Now it turns out we ended up doing a lot of these projects and we going, wow, look at that. We’re getting some pretty cool results. So the point in this is that this framework allows us to identify a set of capabilities, set of roles in an organized ways that move some data to insights, right?

Dr. Jerry Smith: [40:41]

And so, you know, this world is very complex. There’s a lot of other skills you need in there. There’s some, there’s some technology skills. And certainly there’s soft skills in there. When I think about the fundamental, capabilities that projects needs to not these of some of the ones. All right. So with that said today, we covered a lot of stuff. Just sort of scrolling back up to the top of my notes. We covered a set of tools that I think are important to the group. Everything from MATLAB to Mathematica into R and Python know what you think about each of those into several books on causality, from the book of why to causal in to counteracts. We touched a bit on behavioral part, the work done by, foreign in his behavioral data and analysis with arm Python, along with design for behavioral change.

Dr. Jerry Smith: [41:36]

We also talked a bit about behavioral, which is kind of a cool new emerging book regard into some of the tools for causality, the compel on the causal, even the Beisya labs work that’s in there and skills capabilities, and the data science AI space. I truly hope that this was, this was useful to you. You found value. And if you did, please leave comments below. If you didn’t leave comments below too, tell me how we can improve this work. I enjoy bringing this kinda information to you, but I, I really enjoy when I hear back. So that’s going to be it for the day. I’m Dr. Jerry, this is AI Live and Unbiased until next time let’s get after some data sciences. We’ll see you later on,

Outro: [42:25]

This has been the AI Live and Unbiased podcast brought to you by AgileThought. The views, opinions and information expressed in this podcast are solely those of the host and the guests, and do not necessarily represent those of agile thought. Get the show notes and other helpful tips for this episode and other episodes at agilethought.com/podcast.

Agile Podcast: AI Live Unbiased