Bart Baesens is a professor of Data Science and AI at KU Leuven and a lecturer at the University of Southampton. He co-authored more than 250 scientific papers and 10 books. Bart received the OR Society’s Goodeve medal for best JORS paper in 2016 and the EURO 2014 and EURO 2017 award. Bart is listed in the top 2% of Stanford University’s new Database of Top Scientists in the World. He was also named one of the World’s top educators in Data Science by CDO magazine in 2021, 2023 and 2024. Bart is also founder of BlueCourses.
Recently, in an exclusive interview with Higher Education Digest, Bart shared his professional trajectory, the mission and vision of BlueCourses, the key insights on the future of big data and analytics, personal role models, significant career milestones, the secret mantra behind his success, future plans, and much more. The following excerpts are taken from the interview.
Prof Baesens, can you tell us about your professional background and areas of interest? How did you get started in analytics?
First of all, thank you very much for your interest in my work and career. I rather coincidentally started working in analytics (or data mining as it was referred to back in the days) in 1998. Having graduated with a Master of Business Informatics at KU Leuven, I started a PhD and got fascinated by credit risk modeling. During my PhD I started studying techniques like logistic regression, decision trees, survival analysis and neural networks to estimate the probability of default of obligors in retail as well as corporate portfolios. The focus of my research was on developing both accurate and interpretable credit risk models, something that would nowadays be referred to as XAI (eXplainable Artificial Intelligence). It was at that time that the Basel Accords, aimed at developing and validating quantitative credit risk models, also gained traction and interest from lenders world-wide. So once graduated, I travelled around the world teaching banks, consulting firms and regulators how to quantify and monitor credit risk using analytics. This allowed me to gain a unique perspective on practical issues and challenges firms were facing when building, deploying and managing credit risk models which in turn led to identifying various new research topics. Meanwhile, I noted that many of the analytical techniques we studied and developed could also be used for fraud detection as well as marketing analytics (e.g., churn prediction, response scoring, customer lifetime value modeling, recommender systems, etc.) and started expanding my research and team in those areas. In fact, my research always centered around creating business impact with AI, either in terms of profitability, interpretability or identifying new AI inspired revenue generating business models. I set up my own research team and DataMiningApps community (see www.dataminingapps.com). Together with my colleague researchers, we managed to create a unique connect between research and industry which resulted into a fruitful multiplicator effect with various companies such as BNP Paribas Fortis, Allianz, Coca Cola, the Belgian government and ING funding our research as such triggering an efficient and productive cross-fertilization between both academia and industry.
Can you please tell us about the programs/courses taught by you at KU Leuven and the University of Southampton?
Yes, sure! At KU Leuven (Belgium), I am the coordinator of the Master of Information Management which is a one-year master taught in English with about 150 international students. The master offers advanced non-technical training in information management. The emphasis is on the efficient and effective application and management of IT in various business contexts such as finance, marketing, HR, production, and logistics. Obviously, analytics/AI now makes up one of the cornerstones of our master. The courses I teach are Business Information Systems, Principles of Database Management and Data Science for Business so very well aligned with my AI research and expertise. I also supervise approximately 20 master thesis students and 6 PhD students yearly.
I have a part-time appointment at Southampton Business School (UK) where I contribute to a Credit Scoring and Data Mining course and supervise various master thesis projects. I also do joint research and PhD supervision with some colleagues over there.
Brief us about the mission and vision of BlueCourses. What sets BlueCourses apart from other e-learning providing platforms?
Our mission with BlueCourses is twofold. First, we disseminate knowledge by offering on-line courses on a variety of AI related topics such as machine learning, fraud analytics, credit risk modeling, web scraping, deep learning, marketing intelligence and text analytics. The courses are delivered by internationally recognized colleague professors in the field. They typically provide a sound mix of theoretical concepts and methods combined with practical insights and case studies. To optimize the learning experience, the courses are interactively designed with multiple choice questions and discussions. Certificates are also provided upon course completion. Seeing people sharing those on their LinkedIn accounts is really very rewarding. Our second goal is to contribute our share in cleaning up our precious oceans from plastic. Hence, we pledge to invest at least 20% of our EBIT (Earnings Before Interest and Taxes) to companies such as Waste Free Oceans and WWF. Right now, we accumulated a very nice and diverse portfolio of international customers from across all continents which we are very proud of (see www.bluecourses.com for the list) and have already conducted various ocean cleanup initiatives.
In your opinion, what does the future of big data and analytics look like?
The introduction of Large Language Models (LLMs) has largely revolutionized the AI landscape. First and foremost, rather than being fearful, I can say that I am really very hopeful and excited about the future of AI. Rather than curtailing AI research and its applications, I strongly believe we need more AI. However, there are a few challenges ahead in my opinion.
First and foremost, I think education is really key to the successful usage and further development of AI. Since AI continues to be further permeated in nearly all aspects of our daily lives, it is important that people are properly aware of this and learn to use this to their benefit. As an analogy, think about a car. If you can’t drive it but still use it, it’s a dangerous thing, the same holds for AI. So, we should educate people in terms of the basics of AI usage and structured thinking. Also, our youngsters should be educated to properly use AI. One particular risk I wish to highlight is the antropho-morphization of an AI model or it being perceived as a human rather than a piece of software or pre-programmed chatbot. People (and especially kids) should be aware that they are talking to a machine, an automated chatbot and not to a human and there should never be any emotional bonding whatsoever. After all, what distinguishes us humans from AI are our consciousness, emotions, humor, and free will. I don’t believe AI systems have free will since they are pre-programmed to optimise a pre-fixed objective function (typically using gradient descent and backpropagation) which they cannot go beyond.
The next important thing is the open sourcing of AI models as is currently being done by Meta and their Llama LLM. Given the widespread usage of AI, we can’t afford to have it controlled by only a few (most likely West-coast based US) companies offering proprietary closed-source models. Open sourcing AI models is a key prerequisite to make sure they are used in a transparent, fair and unbiased way and foster a co-creation inspired AI ecosystem. Another expected evolution is the further untapping of new data sources originating from, e.g., IoT devices, smart-watches, self-driving cars (we still don’t have level 5 remember), siloed medical applications, etc. This will allow AI models to learn a lot more about the world we live in. Another challenge will be to have AI models capable of reasoning and planning. Right now, language models like ChatGPT are essentially autoregressive LLMs based on a decoder transformer architecture aimed at predicting the next token (e.g., a word) based upon the previous one. There is some scientific disagreement about their semantical understanding of text, but everyone seems to agree that it is very limited at most.
I expect that in the next few years or maybe decades we will see technological innovations aimed at developing AI agents that can autonomously design experiments and learn. Fundamental to these developments will be an improved paradigm for cause-and-effect reasoning. More concretely, generation of outputs is very different from causal prediction since the latter requires a proper understanding of the world and its state and actions taken. Think about the video generation systems we are seeing nowadays such as OpenAI with Sora and Google with Lumiere (both text to video models). The number of possible videos eligible in a generative setup (i.e., textual prompt to video) is enormous as such facilitating the task since getting one sample video out of the huge solution space is already sufficient. However, predicting the continuation of a real video is much harder since the solution space is a lot smaller and depends upon a proper understanding of the cause-and-effect relations and physical dependencies between objects in the video. That’s why I agree with Yann LeCun that the future of AI is not generative. An interesting approach he co-developed is Meta’s JEPA (Joint Embedding Predictive Architecture) model trained to make video continuation predictions in an abstract representation space only focusing on the essential parts of the video. I strongly believe research and applications of these types of multimodal AI combing text, images, audio and video with hopefully a proper understanding of cause-and-effect relationships will continue to grow.
Next, I also expect a surge in neuro-symbolic AI research and applications aimed at combing neural network architectures (such as transformers, auto-encoders, etc.) with human-readable representations of patterns and knowledge often written down as simple IF-THEN business rules typically elicited from human experts based upon their experience and/or common sense. As to economic impact, in the short term we are likely to witness new non-sustainable thin applications based on an LLM API (similar to the Flashlight iPhone app that was later incorporated in iOS) as well as more sustainable applications (similar to Uber, Airbnb, Tinder, etc.). Finally, regulators will face various challenges as well. It is my belief that one should not regulate technology or R&D as is currently done in the EU (e.g., EU DSA Act, EU AI Act, etc.). Instead, I think it’s better to regulate applications of AI such as when it is used for self-driving cars, credit risk or in medical applications. As a concrete example I, and I think many others with me, would be perfectly fine with a black box complex uninterpretable deep learning neural network if it could cure cancer but would prefer to have a transparent, interpretable model for credit risk as I would like to know the reason why my credit request has been declined. Furthermore, by regulating too much, you risk constantly lagging behind the facts as once the regulation has been designed and is ready to be put in place, the technology is most likely to have further evolved with new accompanying challenges. Finally, creating an international level playing field as to regulation is also a key challenge.
What types of career opportunities are available in the field of data science? Could you please share some insights?
First of all, there are some people that fear AI will cause massive job loss. Personally, I don’t believe this, quite on the contrary in fact. Just as the industrial revolution automated manual work, the AI revolution automates intellectual work. Since all our interactions with the real world (using, e.g., our smartphone, car assistant, smart watch, AR/VR devices such as Apple Vision Pro or Meta Quest, web browser and social media, IoT based domestic appliances such as fridges, lights, dishwasher, etc.) continue to be further AI mediated, this will undoubtedly create many new types of job profiles. I hope you don’t mind me dreaming together with you or maybe even hallucinating a bit further upon this. At this very moment we see companies posting job positions for prompt engineers, people who are capable of developing and finetune high quality prompts for Large Language Models in order to develop new Natural Language Processing (NLP) applications. Personally, I don’t think this is a job profile that is very sustainable as LLMs will quickly catch up and learn how to properly respond to even very sketchy defined prompts.
Some more sustainable jobs in AI are the following. The first important job is that of an AI translator. These are people who can match AI technology with business needs or speak AI just like they speak English or any other language. Modern day AI technology is not that good yet at reasoning and planning, but we will undoubtedly see more progress in this area with technology aimed at defining AI agents capable of autonomously designing actions (e.g., a marketing campaign, a fraud investigation, a complaint handling, a job posting and hiring process) or experiments (e.g., in a medical, chemical or agricultural setting). The job profile manager of AI agents will then be responsible to manage the 24/7 available AI agent workforce, keep them updated and properly trained with new data, make sure they are well configured and collaborate efficiently (e.g., may be even have them debate and learn from each other), all obviously in line with the continuously evolving ethical and regulatory guidelines. The recent launch of Meta’s Quest and Apple’s Vision Pro both immersive AR/VR headsets providing interactive 3D experiences for, e.g., gaming, entertainment, fitness, travelling, etc. might create job profiles such as AI Augmented/Virtual Reality architect, AI Graphics Engineer or Computer Vision specialist. I also believe new job profiles will be created in, e.g., the cultural sector. I strongly believe in people’s creativity and new technology untaps new sources of creativity. Hence, at this very moment we are already seeing jobs such as AI artist with skill prowess in Stable Diffusion, DALL E, Midjourney, etc. AI ethics specialists are needed to help us decide what data can be used for what purpose and how transparent should the AI models be depending upon the purpose. We will witness AI application-oriented job profiles such as AI Climate Expert, AI Cybersecurity expert, AI Supply Chain expert, AI movie director, AI software engineer, etc. Finally, given that the whole AI domain currently evolves at warp speed, we observe an explosion of AI researchers (especially in industry rather than academia), obviously also a not to be underestimated job profile with growing opportunities.
With extensive years of industry experience, according to you, what skills or characteristics make someone a seasoned data scientist?
I have seen quite a few data scientists and had the pleasure to work with some of them, both academically as well as consulting wise. In my opinion, the key skills for seasoned data scientists are passion, technological background, creativity and team spirit.
The field of data science and AI is evolving faster than ever before such that without a sound passion, it’s almost impossible to keep up with new technological developments. I actually find this quite challenging myself as well. A seasoned data scientist also has a rock-solid technological background, has been trained in Python, Keras, PyTorch, etc. and has actually developed AI models deep down in the data mud filthy trenches end-to-end themselves. S(he) should have experience in the grunt work such as defining the business problem to be solved, the data gathering and preprocessing, the AI model development and evaluation, and the successful deployment and monitoring thereof. Only then they will be able to properly assess the potential of emerging AI technologies and identify business opportunities. Next, creativity is key. A good data scientist is able to think out of the box. S(He) can properly assess new technologies such as Deep Learning, Large Language Models, Stable Diffusion, etc. and identify various ways of leveraging this to either optimize existing business processes or even develop entirely new business models. That’s why we see more and more job openings for AI translators as mentioned above. Finally, no data scientist works in isolation. Given the pervasive and disruptive nature of AI applications both internally across all business departments as well as externally when engaging with customers or other third parties (e.g., suppliers, government, etc.) seasoned data scientists should be excellent team players and be able to successfully collaborate and communicate with, e.g., business people, IT guys, auditors, ethical specialists, regulators, and yes may be even AI scientists like me.
You have been a recipient of prestigious awards and recognition such as one of the World’s top educators in Data Science and Stanford University’s Top Scientists in the World to name a few. Our readers would love to know the secret mantra behind your success.
A very tough question which I find hard to answer but will give it a shot. First of all, these awards are obviously nice to receive but I really don’t consider myself any different or more special than other people, I really don’t. Each of us has their talents, merits and career paths and this just happened to be mine. There are so many people out there that deserve awards as well. But four things that I can say drive me are: passion, hard work, team spirit and perseverance.
My work is my passion, I just love doing research, educating and inspiring people. I feel really blessed to do research in AI since it’s evolving so quickly with so many new emerging and exciting applications for, e.g., sustainability, climate change, smart agriculture, history preservation, etc. Also teaching it and seeing how students get fascinated by it further catalyzes my passion. It may sound weird but when students come and see me at the end of a course with a small “Thank You” message (occasionally even a small card), then this really touches me profoundly and has a similar effect as an award.
Obviously, passion creates motivation which in turn results in hard work. I get up every morning at around 6am, do my morning routine (see below) and start working. Since it’s my passion, I really don’t mind working hard on the weekend, during holidays, etc. After all, what can be better than doing what you really like!
In terms of team spirit and perseverance, I got inspired quite a lot by the US Navy Seals. What I learned from them is the values they uphold such as first and foremost respecting everyone, the importance of comradery and team spirit as forged in one of their credos “No Man Behind” and they’re never giving up mentality as echoed in “Never out of the Fight” which really helped me through some really tough and challenging times in my life. If you Google my name, you may find some publications, a few books and indeed some awards but nothing about the many failures and setbacks that I also faced both at a private and professional level. The way the Navy Seals embody team spirit and perseverance inspires me a lot as these are lessons that are very useful in everyday life as well. In terms of team spirit, we all need swim buddies (as the Seals call them) to help us sail through rough seas or tell us everything will be alright at the very moment we need it. In terms of perseverance, I learned a lot from the adventures of Marcus Luttrell, ex-Navy Seal, dropped into Afghanistan with three other Seals set out on a mission to eliminate an enemy target. The horrendous ordeal he went through (lost all his buddies and spent days in hostile territory, all nicely featured in the movie Lone Survivor which he co-directed) is one of the most galvanizing examples of perseverance, incredibly strong will, never giving up and especially believing in yourself. To me, he truly is deeply inspiring that no matter what happens to you in life, no matter how dark it may get at times, you should continue to belief in yourself and move forward. Hence note to me and everyone reading this whilst may be experiencing challenging times in life: You will be fine, You really will! I know this sounds pretty naïve, maybe fluffy to some, but honestly, it’s something I really believe, I really do! Finally, to conclude my answer to your question and come back to my morning routine which I referred to earlier, may be to some a funny secret about myself but Admiral Bill McCraven (a retired Seal who commanded the Bin Laden raid) actually inspired me to start each day by making my bed and taking an ice-cold shower. It gives me a small sense of pride and serves as a perfect energy booster for the rest of my day!
What has been your most career-defining moment that you are proud of?
That’s really hard to say. You make me think hard during this interview. I had to mull this over for a few days. Actually, it would be the founding of our company BlueCourses. The reasons are multiple. First, it was done with one of my best friends, Tim Verdonck, whom I have the pleasure of knowing for several years now. Next, the on-line BlueCourses platform is a culmination of years of research and international consulting all of which resulted in a set of comprehensive online AI courses that became really popular. There is nothing more rewarding than to see and hear people say that our courses helped them shape their future careers, inspired them and/or got them passioned by our course content. Last but least, as mentioned before, our contribution to sustainability, we pledge to invest 20% of our EBIT in companies cleaning up our oceans from plastic. As such we proud ourselves to be industry partners with WWF Belgium and Waste Free Oceans (WFO). In fact, last year we actively sponsored an ocean cleanup of Mbezi beach in Dar es Salaam, one of the most polluted beaches of Tanzania together with WFO and 90 volunteers. That’s a moment that I am really proud of and will never forget!
Who is the one person you look up to and why?
I hope you don’t mind me mentioning two, a professional and private hero. As an AI researcher, I shortlisted Alan Turing (who laid the foundations of modern-day computer science), Pedro Domingos (a professor emeritus in Machine Learning) and Yann LeCun and decided to go for the latter. Yann is a Turing Award winning French American computer scientist who co-authored the first paper on Deep Learning and is currently spearheading Meta’s AI research. Not only his extensive knowledge but also the way he conveys it, often with a funny twist and passion twinkles in his eyes. His classy and scientifically grounded charm really helps people understand how AI works, what it can do for them, how it will likely evolve, how it should be perceived as a value-adding technology rather than a threat, the important challenges and attention points, etc. Yann’s unique comprehensive perspective on the entire AI field, both technically as well as application-wise is something I genuinely look up to.
On a personal level, I had to let this sink in for a few days. Names that crossed my mind were Napoleon, Queen Marie Antoinette (the last Queen of France), Caesar, Queen Elisabeth I, Marcus Luttrell (whom I mentioned earlier) and Monica Lewinski. After some deep thinking, I decided to go for the latter, Monica Lewinski. As you know, she was spit out by almost the entire world after being involved in the Clinton–Lewinsky scandal. Whether she did something wrong or not is not for us to judge upon. Instead of drowning in self-pity, she decided to bounce back up, to fight back and do something useful with her life, and she did! She successfully pursued a master’s in psychology and returned to the public as a social activist against cyberbullying based on her experiences and calling herself patient 0 of online harassment. She is a really nice example of a strong woman full of perseverance despite the many humiliations and insults she had to undergo, a quality I largely admire in people!
What are your passions outside of work?
The first one is undoubtedly my three kids. I really enjoy seeing them grow up, exploring the world, doing exciting things together (e.g. attending rock concerts) and trying to inspire them and help whenever needed. Next, I am a history geek. I binge watch history documentaries and finish each day by listening to podcasts (e.g., currently doing Dan Snow’s series) covering all periods in time but with a particular interest in the Romans, English Kings and Queens, the French Revolution and Napoleontic period, and World War I and II. As part of my history passion, I recently initiated a research partnership with the National Museum of the Royal Navy to study the use of deep learning models (such as Large Language Models and Convolutional Nets) for analyzing and managing their digital collections archives. I really felt like a kid in a ginormous play garden when visiting their home base in Portsmouth last August and seeing all the historical ships (e.g., HMS Warrior and HMS Victory which featured in the recent Napoleon movie) as well as both aircraft carriers that coincidentally were there at the moment. Finally, I also love hanging out with friends laughing with each other’s bad sense of humor (or at least them with mine) all accompanied by a good gastronomic meal paired with an exquisite wine obviously with my favorite ones being Chardonay and Viognier for white and Tempranillo, Carmenere and Cabernet Sauvignon for red.
Where do you see yourself in the next 5 years?
That’s a good question. It may sound weird, but I usually don’t think that far ahead. I really try to live by the day and don’t think too far about the future as I have learnt that it is very unpredictable anyway as evidenced by, amongst other things, the recent progress in AI but also pandemics, geo-political tensions, etc. Anyway, what I really hope for is to stay healthy, spend tons of quality time with my kids and friends, and help anyone who needs a pat on the back. One of my maybe somewhat crazy but big dreams is to fly a (duo-seated) Spitfire to the Normandy beaches to experience the thrills of the WW 2 pilots but not sure this will ever take place. And as to work, I hope to further inspire people about the wonderful things AI can do for them, continue and expand my research, see how it can successfully be applied in emerging and exciting settings (such as history, climate change, smart agriculture, etc.), finish the second edition of our book Managing AI Model Risk, so in other words: simply continue to have fun with my amazing kids, friends and colleagues!