Jude Michael Teves [00:00:00] We should not be afraid to leverage these tools, but at the same time, we should be wary of its limitations. We should not be overly dependent. Well, yes, it’s making us very efficient, very productive, but still, learn the fundamentals. Let’s know the bounds of these technologies that we’re using.
Patricia [00:00:35] Hello everyone and welcome to The AI Purity Podcast, your premier destination for exploring cutting edge world of artificial intelligence. I’m your host, Patricia, and today we’re thrilled to dive into the realm of data science and artificial intelligence with an industry leader whose journey embodies innovation and impact. With a master’s in Data Science from the Asian Institute of Management. Our guest has garnered accolades for exemplary work, including being awarded the Top Data Scientist by ASEAN by the Center of Applied Data Science. He is currently serving as Data Science and Analytics Lead, and our guest has a proven track record of leveraging data driven insights to drive innovation and impact. Join us as we delve into his journey, insights and vision for the transformative power of artificial intelligence. Welcome to the podcast, Jude Michael Teves. Hi, Jude.
Jude Michael Teves [00:01:26] Hello, Patricia. Thank you for that introduction. Really appreciate it.
Patricia [00:01:29] Of course. Thank you. We are so glad to have you here today. We would love for you to introduce yourself a little bit to our audience. Please tell us how you got started in this industry and what made you want to become a data scientist and computer engineer.
Jude Michael Teves [00:01:43] Okay, that’s a very long story. How did I get here? Well, I guess even back then, when I was still, in my undergraduate years, I’ve been very much interested in science and technology. So I well, I pursue with a degree in computer science and computer engineering. And, you know, like most computer science and computer engineering people, they would eventually go that route, right? More of like a software engineer. But, even back then, when I was still about to start my software engineer career, I sort of had an existential crisis, I would say. More like me trying to find what it is that I could do that could really benefit the humanity and something, of course, that I’m really passionate about and it should be aligned with what I am studying when I was about to graduate. So I thought to myself if – I think at that time, it could be artificial intelligence, because I think it answers many of the things, those questions that we have like, “What is intelligence? What if we get to a point when the singularity happens?” All of those things. So, I want to be part of that. That was my thinking back then. Before I remember it was not even mandatory to take AI courses – machine learning courses. It’s really a conscious decision of mine to really focus on AI. So, I went that path. I still have my software engineering career, actually. So after graduating, I was a software engineer, but on the side, I was studying lots of AI, and I eventually heard about data science. Because, I mean, it’s not technically the same, but there are lots of overlaps, right? AI and data science case and data science are applying artificial intelligence. And that time, the Asian Institute of Management, where I took my masters, they’re about to offer the first data science masters here in the country, so I was interested in that. I wanted to learn more. And yeah, I got fascinated and I went eventually took that MS degree. I learned about AI in that point as well, but I eventually learned that, you know, yes, it’s something that could really impact a lot of industries like AI. And, at that time, it’s starting to be hyped, not as hyped as it is right now, but back then, you know, it’s still starting. So, I guess I was lucky to be one of the first movers in that space. And eventually, of course, after that, I also got lucky that, you know, it eventually became hyped, and it got many opportunities. So, after taking that MS degree, eventually got this role ADB, Asian Development Bank. I was part of the innovation there. So, we started a lot of AI and DS projects in that organization. So, my typical skill for that would be that. Before ChatGPT became a thing, that’s one of the things that I was focusing on. So back in 2019, my main task at ADB is to create this tech summarization model. And again, this is Pre-ChatGPT. We’re still using GPT before. You know, all of those things took a lot of time. Lots of months for me to really read the research papers and try to implement them and make it work in the context of the bank gathering so many data. We had many people annotate that, basically the typical data collection process and eventually train that a model. So, we’re still getting some gibberish results and that was still normal back then. I mean, it’s not even a long time ago. 2019 was, well, about five years ago now, right? Five years ago, but before I mean, the development in this space in NLP, in large language models, and AI in general, it’s so amazing because before it’s common to see gibberish. But now, we have ChatGPT and lots of models now. Yeah I was focusing on that and lots of NLP chatbot projects. And after that, I went to AIA Insurance. So, still in the financial sector, and on top of that, I was teaching part time at DLSU. So, after my masters, I started to teach part time at DLSU, and I have been teaching ever since, because it’s something that I really wanted to do. I’m really passionate about teaching. Because even in my undergraduate years, I was tutoring. For free, I’ve been tutoring, math, programing, electronics, robotics. And I realized that, you know, it’s something that I really enjoy, and I wanted to continue doing that. But in order for me to teach, I had to have a Masters degree. So, that’s why I had to take my Masters for two reasons. One is I really wanted to go the academic route. I wanted to teach but just part time. And also, that I think that could benefit my career, my corpo career. So, those are the two main reasons why I took my MS. And anyway, so because of that – because of the MS degree, I now can teach. So, I started to teach. I’ve been developing many of the programs DLSU. I’ve been teaching Data Science ever since, part of the co-founders, basically, or the pioneers of the Data Science program at DLSU. That’s, I guess, the general story.
Patricia [00:06:42] That’s really fascinating. And I’d love to get to know more about, like, how AI is being integrated into finance industries, but I did want to talk about your time at De La Salle University. Tell us how you developed their curriculum, particularly in their Data Science programs and their AI and machine learning courses?
Jude Michael Teves [00:06:59] So, we are offering many programs at DLSU about data science. So, we have this thing we call Minor in Data Science that could be taken by any undergraduate degree. So, I have students coming from, say, liberal arts and law – chemistry, biology, CS, of course, statistics, math, IT… all that stuff. So, it’s a thing now. So, they have this core degree that they have, but there’s Minor in Data Science, and they have to take lots of units for them to get the Minor in Data Science degree. So, that’s one of the things that we’re offering. We’re also offering lots of diploma courses. So, we actually started with that. We started to develop the curriculum for – the materials for the diploma courses. So basically, it’s really catered to industry practitioners. Of course, they want to learn more about Data Science, AI – it’s the new thing. And we have sort of a – you could think like it’s a bootcamp more or less. It’s about two months, real crash course on so many topics, and they’re going to learn the fundamentals of Data Science. So, that’s one of our flagship courses or diploma courses. We also now have the Master of Science in Data Science at DLSU. Yeah, those are the programs that we have, and I’ve been developing the curriculum and the materials ever since.
Patricia [00:08:13] [00:08:13]One of those machine learning courses, Data 103 that you were teaching, you cover a wide range of topics from traditional supervised machine learning techniques to cutting edge tools like ChatGPT and Midjourney. How do you balance introducing foundational concepts with exploring the latest advancements in AI? [19.2s]
Jude Michael Teves [00:08:32] [00:08:32]Right, that’s a very good question. Actually tricky to balance… what’s new, what’s modern, what’s novel, and also the fundamental topics at first. I have a bias towards really understanding the fundamentals. You really have to go through, calculus, linear algebra, all of those things. At least, you [18.0s] know, say you have a [00:08:51]working knowledge that it’s like this. You know how to implement neural networks from scratch, [4.3s] new regressions, and [00:08:57]all of those algorithms. A Few of them. But to be fair, before Data 103, which is machine learning, there’s Data 100, Data 101, 102 – those are the three subjects that contribute to data science, database, and data mining. I teach practically all of them, but generally, I would teach Data 102 and 103 – data mining and machine learning. So, even before machine learning, [19.7s] there’s sort of – I mean, the students are sort of trained through [00:09:20]my approach. So, they really have to learn the fundamentals of the previous courses. By the time that they get to Data 103 – machine learning, they’re more prepared. They have a quite enough foundation for them to really tackle the modern stuff. But still, again, I will tackle, calculus, algebra, make sure that they know at least these – the most basic implementation of linear regression, neural network, all of those things. And after that, now, we’ll explore the modern applications of Data Science and AI as well as the industry nuances, because that’s something that, I guess – I think, most academics tend to forget. Because, I mean, [40.5s] that’s the setting that we’re in. So, [00:10:02]it’s mostly about, you know, the theory, the [2.3s] foundations, but how is it really being used out there. [00:10:08]It’s also one of the things that I add in my courses, because I’m also an industry practitioner. So, I always mention that, “Okay, this is what I’m teaching. This is the ideal state of things in an ideal world.” But in reality, sometimes this happens because this and that. That’s the context. That’s the industry. So, there. [16.8s]
Patricia [00:10:26] [00:10:26]And with the advancements in AI and machine learning, how do you encourage students to stay curious and keep up with the rapidly evolving field beyond the confines of the classroom? [9.2s]
Jude Michael Teves [00:10:36] [00:10:36]I really encourage my students to play around with the tools that are out there. Because there’s just so many things that are happening nowadays that even me, I can’t always be up to date with all of those advanced – all of those technologies. There’s just so much. That’s why I focus on the fundamentals. I show them a few applications, a few important tools, and then I leave it to them. Let them explore, and I also learn from my students. I get to see that “Okay, so there’s this nice tool that could do this and that.” But yeah, I really encourage my students to do that. So, in one of my assessments and one of my homework, I really make them use these technologies, especially like, “Use this to summarize this research paper that I really like.” So, in one of my assessments, basically, I have a list of research papers that I want. I assign them to different groups – use AI to explain it. Use AI to generate a slide. Use AI for everything. And then, check the limitations of the AI approach. Because, eventually they’ll see that it’s not perfect. Sometimes, it’s making up [56.8s] facts. [00:11:34]Sometimes, the lie that was generated is wrong. So, they will learn by using the tool. [6.4s]
Patricia [00:11:41] [00:11:41]And would you give the same advice? Because your students have you as a mentor to teach them real-life applications of these tools and to see how AI doesn’t necessarily do everything the same way human ability can, you know. And for people who aren’t your students, what would you say to them? About, like, you know, AI is becoming more and more prominent. You said from 2019 to now, 2024, the growth of AI has just been too – a lot. What would you say to these people who aren’t your students? [30.5s]
Jude Michael Teves [00:12:13] [00:12:13]We should not be afraid to leverage these tools, but at the same time, we should be wary of its limitations. We should not be overly dependent on these technologies. While, yes, you know, it’s making us very efficient, very productive. Let’s still learn the fundamentals. Let’s know the bounds, the limitations of these technologies that we’re using. So, in order to do that, we really have to just explore. Test it. Use it. You will learn eventually that, you know, these are things that we can and can’t do. And if you have the time, I mean, not everyone has to take the technical route, but if you have the time, you can study the fundamentals. So, you really get to know that, “Okay, these are the foundations. This is what we use to get the modern applications like ChatGPT.” And you’ll have a better appreciation of what you can do and probably its limitations on a foundational level, because you are studying the fundamentals, but that’s like an extra if you really wanted to go the technical route. [52.0s]
Patricia [00:13:05] [00:13:05]Yes, and I feel like this is a really important discussion. This is why here at AI Purity, we are a platform dedicated to AI text detection. I feel like it’s really important for us to discuss these limitations of AI and talk about, like, the ethical way to use it. And so, I wanted to ask you, as an instructor and mentor, how do you perceive the current landscape of AI use among students? And what role do you think AI Purity platform – AI text detection tool can play in promoting responsible AI use in educational settings?[32.0s]
Jude Michael Teves [00:13:38] [00:13:38]These are uncertain times that we’re in now, I would say. While I’m very excited about all these development, at the same time, you know, this is uncharted territory. So, even I don’t know what’s going to happen the next couple of years generally. And I do encourage to use these technologies but be mindful about the ethical implications of such. Let’s not be very dependent on things. Let’s still apply our critical thinking. And I think these tools could help us like AI Purity in terms of identifying the ones that are really, you know, AI-generated or not. And with that, we’re now applying that extra layer that could be used to at least lessen the noise that we have over the internet. Because right now, there’s just so much AI-generated content, and at least there’s that one more thing that we could use to remove all of these. I would say [55.0s] status [00:14:34]that we have in terms of the content which could really make it hard for people to consume more content, because, you never know what is real – what is not. At least, we’re helping in that aspect of things. [12.2s]
Patricia [00:14:46] [00:14:46]Can you share some specific instances from your teaching experiences where you’ve witnessed the impact of AI on student learning, and how do you navigate the ethical considerations that may arise? [11.1s]
Jude Michael Teves [00:14:58] [00:14:58]Well, generally speaking, I would say that there’s definitely an increase in productivity because it’s still an additional data point that the students could use for them to learn instead of just, you know, relying on, say, books, or me. Maybe there’s another way of saying it that would help the students. Because, I mean, we all have different ways of learning. Maybe this specific explanation – this specific analogy works for you but doesn’t work with another person. But at least with AI – these tools, we could just, keep on asking until you get that certain explanation that you want to hear, so that you can better digest that new information that – and I’ve seen many of that in a lot of my students. But again, because I really encourage them to validate what they’re learning. Check the sources. Make sure that what the AI is spitting out is real. And at least there’s that perspective so that they don’t really just rely on what it is saying. [57.5s]
Patricia [00:15:56] That’s really great, and I totally agree. Students, not even just students, everyone on the internet right now should be practicing discernment with any types of content they consume online, because you’re not really sure if it’s, like, created by a human or it’s AI-generated. But considering your background in machine learning and AI, how do you integrate these technologies into your computer science and data science courses to better prepare your students for the evolving job market?
Jude Michael Teves [00:16:22] Well, aside from the use of generative AI tools, in my case, it’s really at the core of what I would be teaching. Say, for example, the Data Science minor. At the end of it, the finale would be Machine Learning. So, by default I have to teach Machine Learning, and Machine Learning is being used a lot in the Data Science field. So, it’s just one additional piece that we could use to extract insights from the data that we have. So, how do I integrate it? Well, usually, for the whole Data Science minor and even other programs that we have been offering really emphasize the use of modern tools like Machine Learning, so that we can get better insights from data aside from the traditional analytical methods.
Patricia [00:17:04] I mean, with the prominence of, AI tools right now, a lot of people are really scared of, you know, maybe losing their jobs. There’s been a lot of talk about that. [00:17:14]In your opinion, what are the key skills and knowledge areas that students can focus on to excel in the rapidly evolving field of AI? [8.0s]
Jude Michael Teves [00:17:23] [00:17:23]Okay. So, definitely the first thing would be critical thinking. In a stage where there’s just so much information out there, and you never know which one is real or not, you need to have that critical thinking to determine whether it is fake or not. And I guess on top of that, I encourage my students to – well, in the context of this course that I have, right? We then really develop their foundations. Well, yeah, I mean, we could just make use of the tools and just apply it to whatever. At some point, I would say you’ll get tired. We’re just catching up with the newest trends, all of these technologies, you need to have a stable foundation. You need to have a core. And if you know the fundamentals, like, really knowing, the math, not I mean, you know, the hardcore level math, but at least enough for you to be able to understand how these models work. At least, you know, you’ve implemented some – you’ve seen it in action. Maybe try to implement a few models, even the basic ones. Something that, you know, you could use whatever happens. There’s this new technology, you know that you understood it well enough, at least the foundations. So, I’ve been applying the same thing for many years now. I’ve seen many developments in the field. I’ve seen all these new architectures and whatnot in AI. But at the end of the day, it’s still the same core algorithm. I know I’m trying to reduce it. This is the reductionist approach, but yes, there’s still that core, and it’s still the same. For example, the back propagation algorithm. It’s still the core of a neural network of all these models that we’re using. If you understand that, then at least you can withstand all these changes. More or less.[113.8s]
Patricia [00:19:19] [00:19:19]And how can students and educational institutions better prepare students for the integration of AI in their future careers? [6.2s]
Jude Michael Teves [00:19:27] [00:19:27]I would suggest for schools to really encourage the use of these tools. For them not to be overly wary of such things. Yes, of course, we should still be cautious, but at the same time let them know about its limitations by actually using it. Even the faculty. I really encourage my co-faculty to make use of these tools, so that we all know together what it can and cannot do. Again, it’s still a relatively new tool. It just released at the end of 2022. Yeah, it was 2022. Number 2022, ChatGPT. And I would say up until now, we still haven’t fully understood it. So, by making use of these tools, we will know better as a species – as humanity. We’ll know the extent of what it can and cannot do.[51.5s]
Patricia [00:20:21] [00:20:21]And given the newness of the industry, how do you approach teaching students about the ethical considerations surrounding AI? [7.0s]
Jude Michael Teves [00:20:29] [00:20:29]It’s a very hard topic, how to discuss the ethical considerations of such. I would say the – what initially happened when these tools were used is that it was met with some resistance. They generated lots of resistance. My approach was different. I mean, I’m not saying that this is the best thing, but I guess it worked for me. I embraced it initially, okay. Well, my thinking is sort of like, “If you can’t beat them, join them.” Something along those lines, actually. So, I don’t really understand the tools back then. I’m not sure about what it can and cannot do as I’ve been saying. So, okay, let’s explore this together. Let’s see. Encourage my students to use this while ensuring that I keep on mentioning that these are the limitations whatnot. But, you know, maybe I’m wrong. Let’s just see. Let’s just use. Maybe it’s now performing better than expected. So, by doing such, and I’ve been doing this for the past couple of terms. More than a year now. I’ve seen different ways for my students to leverage these tools. I’ve seen different approaches on how they could use the AI to explain all these topics that I’ve been teaching. So, that has been my approach. Because even me, I don’t know the full extent of these technologies. [82.0s]
Patricia [00:21:52] That’s a great way to see it. If you can’t beat them, join them. Especially when using these types of AI tools. But [00:22:00]how do you envision the future of education? How do you think AI will be impacting education in terms of the curriculum development and teaching methodologies? [8.0s]
Jude Michael Teves [00:22:09] [00:22:09]What I foresee is that we’ll focus more on the applications that showcases the nuances of these things that we’re teaching instead of the typical, I mean, the typical skill set, I would say. It’s just the foundations. I mean, that’s still important, but we could now focus on other things, because it has been done for so many times. There are these tools that could help us learn this new topic. So as educators, we can now focus on other things, and that’s what I am trying to do right now. So, I – generally, I still have my typical lecture sessions, but I augmented with – you know, there’s this existing course out there, you know. If you’re gonna learn programing, might as well just use this [44.7s] DataCamp [00:22:55]for the typical syntax. That’s something that has been taught a lot already. There’s just so many resources out there. Thousands even, and it’s for free. Why do I have to keep on repeating the same thing? But what I’m teaching are the things that are not on the internet. The context. The nuance. Okay, yes, this is the foundation. Yes, we use this model, this algorithm, this technique. But how is it really being used out there? Those things. So, I think, we could now focus on those things, the higher level things instead of the fundamental ones. [33.9s]
Patricia [00:23:30] [00:23:30]And as you teach these courses on Computer Science – Data Science courses, what challenges have you seen or observed amongst your students in facing, and understanding, and applying AI concepts? [10.0s]
Jude Michael Teves [00:23:42] [00:23:42]There’s so many facets. In learning AI, the challenges, generally, I would say, in the courses that I teach, it would be the diversity of the background. Let’s start with that first. Because in the courses that I teach, other students come from all sorts of background, and some of them are coming, say, from liberal arts, law – courses that are very far from STEM, right? So, with the courses that I teach, it is the first hurdle that I encounter. How do I teach this very practical course to a very diverse audience? Of course, there’s going to be some compromises, right? But again, well, if there have been multiple courses leading up to the Machine Learning course that I have. So, that’s the first thing. Of course, I really have to be, I guess, creative as an educator. [47.6s] At least pull that off, [00:24:32]given the diversity of the audience. But assuming they now have enough fundamentals, okay, they’ve learned the calculus, and algebra, statistics, programming side of things, actually, that’s their next problem. Learning all these concepts and integrating them together. Say, for example, you might be an IT person, you’re very good at programing, but you might have some math [26.1s] and stat, [00:25:00]but you’re not proficient in it. Same goes for the others – those people – they’re proficient in this kind of topic, but you now need this. So, in the field of VSM, it’s the intersection of all these topics, and that makes it tricky for a lot of my students to grasp. Like, they could probably understand each of the topics alone. In silo. But how do you now synergize? How do you integrate all of those things – combine all those learnings that we’ve had and apply them? That’s another thing, and you really just have to keep on practicing. See the different use cases for those topics and how they’re being utilized together. That’s really it. So, those are the typical hurdles that my students are facing. Let’s just say that, okay, now they can build the model. They can now, like, use machine learning. They’ve been doing that for a couple of months or a year or two in my DS minor. The next problem would be the application – the industry perspective. Well, like, yes, they can now create a pipeline. They can, like, do – train and predict models, but is it really necessary to, like, train a model? Maybe for this problem. We don’t even need to train our model to begin with. It is simple, I don’t know, say, correlation. Maybe that could already answer the business problem. That’s something that, again, they could learn by experiencing many things. This is not something they could, like, easily get by reading books. So yeah, those are the typical hurdles, and yeah, they just need a lot of time for that. That’s normal. That’s why I’m there as their mentor – as their prof to really guide them. So yeah, I’ve been teaching the fundamentals, but we will also discuss those nuances. Like, would you sacrifice accuracy versus runtime? It depends on the context. Maybe in this context, accuracy is very important. It doesn’t matter if it, like, takes days or even weeks. I mean that’s absurd, right? It doesn’t matter if it takes that long if it will save a life in the context of health. I don’t care if it takes another day or two, but that 1% increase in accuracy means a lot, because that’s a life being saved, right? So, it really depends on the context. Like, what if we think of the context of self-driving cars? Maybe this model is very accurate to detect if this is a person 100%, but it takes like, I don’t know, a day for it to process. Of course, they’re going to hit that person already. It has to be fast while not, you know, sacrificing the accuracy that much. So again, context is important. You have to know. Context. Don’t just blindly apply some model, because you know, this has the best accuracy. You have to know what it’s for. It really depends on the industry – depends on the application. [163.8s]
Patricia [00:27:45] And in your experience, [00:27:48]what are some common misconceptions or myths about AI that students often have and how do you address them? [6.8s]
Jude Michael Teves [00:27:55] [00:27:55]Like, it’s magic. I mean, of course, if you haven’t taken any of my courses, most people would just think that it’s like magic. It’s something that we can’t do. But yeah, it can be advanced, but if you break it down properly, you could demystify a lot of things. But of course, with the more recent models, it’s just too complicated for us to understand, because again, these models have billions of parameters. But if we start with the more explainable models, then yeah. [30.3s] It’s not black box. It could really explain why this is really the output. [00:28:32]I believe that everything is deterministic. If again, we have enough computational power to process that information, we could really determine why this is output. But there’s just so many combinations, that’s why it’s hard for us to identify why this is the final output. Again, because of the billions of interactions of the nodes in the models that we have right now. But if you think of a simple model, very clear. This is output because of this rule and this mathematical equation and whatnot. Yeah. [27.1s]
Patricia [00:29:00] Now, I want to pivot a little bit away from the educational setting, and I wanted to ask you a little bit more about other industries, especially since you work in the finance banking industry. In your role as a Data Science and Analytics lead, [00:29:13]how do you foresee AI changing the professional landscape in finance, for example, especially in areas like credit scoring and what are the ethical considerations that may come to play in that domain? [10.3s]
Jude Michael Teves [00:29:24] [00:29:24]Yeah, very good question. Generally, in this sector we have a preference towards explainability. While, yes, these models are something that we could use. We generally don’t recommend the use of such just because it is really important for us to explain every single thing, like in the context of credit scoring. Would you want me to say to you that you have been rejected, because the AI model said so? That’s something that most people can’t accept, right? It has to be clear, because this feature, this feature, feature, blah blah blah… This is the effect on your score. Very clear. But in certain applications, yes, we could definitely make use of these advanced models, but for the most part, we are very cautious, and also because that’s mandated by law. We really just have to explain why we’re giving you this score, for example. But, say, for cases that are more on the cognitive AI side. Take, for example, face recognition – those kinds of things, definitely you have to make use of advanced models, [77.6s] advanced neural network models [00:30:44]for that. But others, it has to be very explainable. [3.3s]
Patricia [00:30:49] And [00:30:49]as a consultant in Machine Learning and AI, what challenges do you often encounter in developing end-to-end applications, and how do you address them? [7.8s]
Jude Michael Teves [00:30:59] [00:30:59]Well, it depends on the organization that I’m working with. But, like, the general problems in deploying end-to-end applications would be, of course, where to get the data from. That’s the first problem, yes. Well, we have been hearing about all these cases and whatnot, but in reality, it’s hard to pull it off. It’s hard to even, like, start a proper project, because the data is not there. Or maybe there are data restriction issues. For various reasons. Maybe, data privacy related matters, the data collection in general, the different data formats out there. Like, yes, there’s data. You would expect that there’s still – the data is still on some paper out there. It’s not digitized. So, that’s typically one of the problems. Yeah, the data exists but not digital. Typical problems. [52.4s]
Patricia [00:31:52] And [00:31:52]how do you foresee AI impacting the future of finance and banking, and what potential challenges should industry professionals be prepared to address this? [8.2s]
Jude Michael Teves [00:32:01] [00:32:01]Well, I’m not speaking on behalf of any financial institution. It’s gonna be safe. But I think my opinion – one thing that could really impact would be the customer experience. That’s a given, I think. Because of all these tools that are available to us, that are very modern, that sometimes feel very human-like, we would have a better customer experience. Say for example, we have chatbots we’re talking to. And overall, you’ll have a very [28.4s] human [00:32:31]experience, instead of, like, a typical chat response, which are very limited. So, you know, the onboarding process could be more seamless, could be more smooth. Generally, I think that’s one big impact of the technology in the financial sector. Also with, say, something like voicebots, something like that. Yeah, because they actually feel very human-like now. Like, when you listen to them, right? When you listen to those AI voices, I sometimes think that they are like real human in some cases. So basically, better customer experience overall. [31.5s] On a rating [00:33:04]from a more, I guess, more technical side of things, the fraud. With better AI tools, we could have better predictions on the risks and the fraud. But again, the explainability factors on everything. But anyway, if you could catch those outliers that means a lot. So, I think that’s one of the things that AI could really happen in this particular industry. [25.4s]
Patricia [00:33:30] And can you discuss a specific project or initiative where you implemented machine learning or AI applications, and what impact did it have on that industry or organization you were working with?
Jude Michael Teves [00:33:42] One is in ADB, and I’m part of the innovation division many years ago. So, we started many of these tool projects – tool-based cases, and one of which is the one that I’ve mentioned earlier, wherein we’re creating this model that could do tech summarization. And generally, I mean pre-ChatGPT, that’s very rare. That is one of the hardest if not the hardest tasks in AI and natural language processing in general. Basically, creating text – summarizing text. Now, it’s very common with ChatGPT. But again, it wasn’t that good before, especially in Asian Development Bank. That really means a lot, because we’re really developing something, and we’re really proud of it, because we took lots of months, but we’re able to create something that could summarize a text. So, what what is it for? Basically, in ADB, we are – we have so many projects across Asia funding lots of large scale infra project. And at the end of those projects, we have this thing called the Independent Evaluation Report. So basically, it’s like an assessment of what went right, what went wrong, what could be improved, those things, right? So, it’s a long report, lots of pages, about hundreds of pages in each of the projects. And we basically got many projects from ADB, but those PDFs extracted the data, and then we had help from many of the people in ADB to annotate that with now for the specific, say, PDF or paragraph or text data. This is the output. So, lots of manual work involved. But anyway, we’re able to have a good corpus of data that we could be train our model on. And with that, we’re able to create a NLP model, a large language model back then that could summarize this ADB themes, instead of you spending hours reading those PDF papers. So, basically helps everyone save a lot of time, right? Yeah. Than if you spending hours – That would be very impactful – Just read the summary. Yeah.
Patricia [00:35:44] I wanted to pivot back to ethical considerations and AI integration, especially in education. What do you think are the guidelines or best practices that should be in place for the responsible student engagement with AI technologies? Particularly with tools like AI Purity, especially since we were talking about ChatGPT earlier. You know, I feel like that’s been one of the biggest things, like students using tools like that to, you know, create their work. So…
Jude Michael Teves [00:36:13] Guidelines… In my classes, what I do is, well, yes, I encourage them, but make sure you add references. You add notes. Because of the help of ChatGPT, there is certain prompt I got is all. It’s okay for you to use these technologies but just be honest. I think that’s a really important thing. I just want honesty from my students. At least they know – I know that they use this tool for this thing, and they’re able to answer that because of that.
Patricia [00:36:40] And can you share instances where ethical considerations in AI projects presented challenges during your consultancy work? And how did you navigate these challenges?
Jude Michael Teves [00:36:50] I guess the typical ethical consideration in general when I do any AI or ML-related projects would be the use of personally identifiable information. That’s very important. So, what is that? Basically, an information that you could use to identify a person. And in general, we avoid such, because it might, well, induce some bias. Of course, we cannot include, for example, race or gender or other similar features in your models. So, that’s something that I’m really emphasizing when we do any of the projects. But we have – in the context of my class, if there’s something like that, I tell them that, you know, we can’t use that. Remove that. Try to find another feature that could explain whatever it is that we’re trying to predict.
Patricia [00:37:39] And what what would you say are the paramount ethical considerations that professionals in the AI and data science field should prioritize? And how can they actively address these considerations in their work?
Jude Michael Teves [00:37:53] I would say it’s the same thing. Still, the use of personal identifiable information. It’s really a big no no for me to use those things. So, the same applies in the context of corporate industries.
Patricia [00:38:07] I wanted to talk about the recent award that you were given, being a top data scientist in ASEAN. What insights can you share about the current state of AI adoption and innovation in the region?
Jude Michael Teves [00:38:17] In the region? Well, definitely there’s a lot of adoption since ChatGPT was released. I’m pretty sure we’ve been hearing about AI in the past couple of years, but a lot of people out there are not very interested in that. But with the advent of ChatGPT and, you know, start of 2023, they call that the year of generative AI. Anyone’s interested. Like even my grandparents. They’re asking, like, “What does it do?” They’ve been using the tools and whatnot. It’s really a lot of adoption because, you know, everyone’s talking about it. Not one of the most. It is the most downloaded application. I was able to open the top social media platforms. Like, in the first couple of months, there have been so many users of ChatGPT. So, definitely we can clearly see that there’s a lot of adoption in the field for AI, not just in the Western world, but also in Asia.
Patricia [00:39:08] And as a Data Science ambassador for the Asean Foundation, what initiatives are you involved in to promote awareness and education about AI in the region?
Jude Michael Teves [00:39:19] As part of me being an ambassador for the ASEAN Foundation, as a data science ambassador, I’ve been conducting many Data Science sessions. So, it’s specifically for the Data Science Explorers competition while I do some upskilling about Data Science analytics and Cloud. We could also utilize that knowledge to compete in that specific competition where they could win thousands of dollars. So, I have been doing that last year. I’ve had many sessions both onsite and offsite, mostly held at DLSU, De La Salle University. That’s my alma mater. Had many of my sessions there, but I also had lots of participants from other ASEAN countries such as, you know, Vietnam, Thailand, Singapore, Indonesia the had hundreds of participants. And even if, like, I’m not doing it for the ASEAN Foundation, again, I’m an educator to begin with. So, I still do teach in my classroom and many other engagements that I have here in the country. I’ve been active in the community. So, it’s not just for ASEAN Foundation – it’s not just for the DLSU. I had lots of speaking engagements like, for Google, for one of the regions here, had like thousands of people, thousands of students attending my talks in person. That’s what I do, I guess, as an ambassador. I really promote the use of Data Science and AI – responsible use of those.
Patricia [00:40:39] And what role do you see AI playing in skill development and continuous learning, both for students and professionals?
Jude Michael Teves [00:40:46] Just keep on asking the questions until you get what you want. Try to be more productive with the use of these technologies. Yeah, that’s really what I see.
Patricia [00:40:56] And I wanted to ask you some advice since you are a mentor and just general future directions for AI in education in other industries. [00:41:05]In your role as a mentor, what advice do you give to your students and professionals entering the AI and data science field to navigate ethical challenges effectively? [9.9s]
Jude Michael Teves [00:41:16] [00:41:16]Good question. General advice. Well, I have given so many advices, but what’s the most general one that I would give? It really depends on the context for the most part, but one of the things that I often say is, you know, you have to learn the fundamentals. All those things as I’ve mentioned earlier. But I guess at the end of the day, what I would like to impart to most people is that don’t forget to be human. Sounds cliche, right? But at the end of the day, what are we using these tools for? Why are we doing all of these things? It’s for us. It’s for the betterment of our race. Of us. For us humans, right? So, just have that in mind in whatever it is that you do, whatever it is that you develop where the end users are us. That’s going to be your Northstar. It will guide us in whatever it is. Don’t forget to be human. [37.4s]
Patricia [00:41:55] That’s beautifully put. That’s one of our, you know, core beliefs here at AI Purity, actually. Purifying, you know, the AI. Keeping things human. That’s really great. Thank you for sharing that. And how do you see the field of Machine Learning education evolving in the coming years? What trends or advancements are you anticipating?
Jude Michael Teves [00:42:14] I’m looking forward to more developments in the reinforcement learning space. There are three areas in Machine Learning. For the most part, we have been focusing too much on supervised learning things. That is the easiest next to unsupervised learning, but there’s reinforcement learning. And I think a lot of the developments that we could still do, we’ll definitely be in that space. So, I am looking forward to that. So, what are some typical examples of reinforcement learning, like the AI models that they use for Dota 2, for example. I was able to be, like, the best Dota player. Same, for the game of CSGO. Basically, those games, the AI used for that, that’s reinforcement learning related. So, I’m really looking forward to that. Haven’t really done much work on that personally. I did make a few examples, but that’s not something that I am actively working on. So yeah, really looking forward to that, and I also want to do more research on that space eventually. Maybe if I do, maybe once I start that, maybe I’ll do reinforcement learning, but yeah.
Patricia [00:43:18] And with the continuous evolution of AI technologies, what are your hopes and concerns regarding their impact on education and society as a whole?
Jude Michael Teves [00:43:28] I am genuinely and I’m hoping that we use it as a means to ascend. Basically, be more productive. But of course, we have to be mindful of the bad players out there. We have to put some policies – some rules in place so that it cannot be abused. So looking forward to see it being used in a positive manner. But yeah, there has to be some rules. There has to be some checks in place. Just in case there will be some bad actors. You never know.
Patricia [00:44:02] I agree, and this is why we, you know, have these discussions with a lot of professionals across different industries. Because I truly believe that with enough discussions, we can, you know, really promote the ethical use of these technologies, because I don’t believe they’re going away. They’re becoming more and more prominent in society. So this is really important. And I wanted to ask you just quickly before you go, has there been any regulation so far in De La Salle University, for example, about the use of AI among students?
Jude Michael Teves [00:44:29] Yes, well, say for the most part. For example, in what I teach, it is explicit there in DLSU that they cannot use AI for these projects or these assessments unless we explicitly say so. So, at least that’s there for a lot of my assessments. Well, it depends on what it is., but generally, I would encourage the use of that. If I am confident enough that, you know, it’s something that they cannot research on the internet, then I tend to give very hard assessments that I am generally confident that, you know, you can ask AI. You can ask ChatGPT, and a lot of the times it cannot answer. I actually like that because at least, they know that they can’t fully rely on AI.
Patricia [00:45:11] Oh, that’s great.
Jude Michael Teves [00:45:12] That’s a good lesson already. Yeah. Try to use that. Up to you, but don’t expect that it will be the right answer.
Patricia [00:45:18] And I don’t want to take too much of your time. I just wanted to ask you what does the future hold for you? Do you have any other plans? What’s the future of you and AI machine learning in your career, at least?
Jude Michael Teves [00:45:32] Well, maybe eventually in the future, I’ll continue with my PhD. So, if that’s the case, then definitely there’s going to be more AI research work for me. But if that’s not going to be the route that I’ll be taking, then that’s practically what I do every day. Like, I do a lot of ML, AI, and the different engagements that I have with the government, with the industry and with the academia. So yeah, still lots of AI, but I guess I wanted to explore more. Maybe, you know, as I mentioned earlier, studying more parallel reinforcement learning. Try to see the limits of that particular sub-area in AI. I still don’t know a lot, and that’s okay. That makes it fun. I’ll learn a lot of things.
Patricia [00:46:23] And you are an expert in Machine Learning, AI, Data science – Is there any advice, anything you’d like to say to our audience today just to impart on them?
Jude Michael Teves [00:46:34] Find your passion. Find your call in life. Can be hard, but it’s going to be worth it. But at the same time, know what’s relevant out there. So, I always think about “ikigai” if you’re familiar with that concept. Ikigai?
Patricia [00:46:49] It’s a Japanese concept, yes.
Jude Michael Teves [00:46:51] Yeah, it’s a Japanese concept. I forgot that verbatim, but basically, find that thing that is in the intersection of what you’re good at and what the world needs, what you can get paid for… I think there’s one more thing I forgot, but generally, good ideas. So, those things and hopefully it aligns. I mean, you’re listening to this podcast, so I’m hoping that, you know, it’s something AI or ML related. But if it’s not your thing, then that’s okay. Find whatever it is, and hopefully you’ll find that soon. If not, just just give it time. I guess in my case, I got lucky that, you know, I got into this field, and luckily, this is what the world needs. So yeah, I’m passionate about doing AI – ML stuff. Eventually, it became hyped.
Patricia [00:47:33] Yeah, you were a very visionary person. Because who would have, like, envisioned, you know, AI being such a huge part in our lives today. So, that’s a really great foresight on your part there. So, we want to thank you so much for taking the time to join us in an episode of The AI Purity Podcast. We really appreciate you, and anything else you’d like to say before we end the episode?
Jude Michael Teves [00:48:00] None. I enjoyed this. Hopefully, we can have collaborations in the future, and I’m looking forward to this tool that you have created, so really excited for you guys.
Patricia [00:48:09] Yes, absolutely. Thank you so much, and of course, thank you to everyone who’s listening today and for joining us on another enlightening episode of The AI Purity Podcast. We hope you enjoyed uncovering the mysteries of AI-generated text and the cutting edge solutions offered by AI Purity. Stay tuned for more in-depth discussions and inclusive insights into the world of artificial intelligence, text analysis, and beyond. Don’t forget to visit our website, www.ai-purity.com, and share this podcast to spread the word about the remarkable possibilities that AI Purity offers. Until next time, keep exploring, keep innovating, and keep unmasking the AI.
