The newest addition to IITM’s vast array of departments, the Wadhwani School of Data Science and Artificial Intelligence has been the talk of the town recently. Professor B. Ravindran heads the Department of Data Science and Artificial Intelligence (DSAI), the Wadhwani School of Data Science and Artificial Intelligence (WSAI), the Robert Bosch Centre for Data Science & Artificial Intelligence (RBCDSAI) and the Centre for Responsible AI (CeRAI) at IIT Madras. Currently, his research interests are centred on learning from and through interactions, and span the areas of geometric deep learning and reinforcement learning. In this interview, we learn about the department’s goals, how it affects current IITM students, and the future plans for the department.
T5E: Why was the DSAI department formed, and what are its goals?
Data Science and AI is a fast-growing field and has been in the making for about 50-60 years now. The time has come for us at IITM to stop thinking of AI as an extension of CS and instead start recognizing it as a separate department involving people from diverse backgrounds. AI is being used today in the industry to solve some very hard problems, which is something that we want our department to focus on as well.
How many Indian web-pages do you think there are in India? Just about 6 million!
This is an example of a challenge very specific to India. India is still fairly underrepresented online – the Internet has relatively fewer Indian web-pages, which means that the amount of Indian data we can scrape is quite low. When web data is used for training LLMs, Indian culture might even be ignored! So, there is a dire need for better indigenous tools and specific departments dedicated to this. The name of the department itself is Data Science and Artificial Intelligence – using AI in a data-driven manner. Apart from thinking about AI and its philosophical implications, we’re focusing on using it as a problem-solving tool for data-driven interdisciplinary problems.
T5E: We noticed that the department is also focused on applied AI and integrating AI with other interdisciplinary fields apart from fundamental research. Could you explain the thought process behind the same?
Pioneers in AI come from departments such as psychology, philosophy, mathematics, economics, communication theory, and computer science. Inspiration for AI comes from various fields, and AI models and techniques are inspired by multidisciplinary fields. In recent times, much of the push for progress in AI has been from the industry, where people are trying to solve some very hard real-world problems. We want to focus on this in our department, hence the focus on applied Data Science and AI. However, this doesn’t mean that the department is not doing fundamental research. Several faculty in the department look at fundamental concepts such as explainability, learning in AI, new learning algorithms, and so on.
T5E: Now, let us talk about the new B. Tech in AI and Data Analytics (AIDA). Usually, first-year students in CS and other departments have physics, chemistry, and math courses from different departments. These have been replaced with equivalent courses from the DSAI department like Calculus for Engineers, Applied Linear Algebra, Computational Physics and Chemistry. What is the difference between these courses?
DSAI Department mathematics courses are designed for students to learn all the math needed for further ML courses early on in their academics. They are tailored towards ML, with less emphasis on proofs and more on problem-solving. Likewise, the Computational Physics and Computational Chemistry courses are meant to aid students to learn through practice and coding, i.e. through modelling scientific phenomena in this manner, instead of learning the standalone theory. The aim is to help students learn science concepts through the lens of Data Science and AI.
T5E: ML courses are usually taken by other undergraduates starting from Semester 5, when students have more mathematical maturity. What is the thought process behind introducing these courses much earlier in the B. Tech AIDA (AI and Data Analytics) curriculum?
The B.Tech. AIDA students will be taught math courses like linear algebra, calculus, optimization, and probability right from their first year as core courses. By the third semester (which is when they start ML core courses), the students will have sufficient mathematical maturity and intuition to understand the concepts. The entire curriculum is being designed jointly in a ‘plug-and-play’ manner – we are trying to ensure that the core courses all seamlessly integrate into one another, picking up where the previous course left off. So by the time the students take any ML course, they will already have all the necessary prerequisite knowledge.
T5E: Could you give an idea of the course progression for getting a good grasp on the field of DSAI?
The recipe is simple for any student: start with a deep dive into the mathematical fundamentals – Probability and Linear Algebra (and Optimization, if possible). After this, students should take a good Intro to Machine Learning course to learn basic classification, regression, etc. Once these prerequisites have been satisfied, students can move on to more advanced courses like courses in Artificial Intelligence (AI), Deep learning (DL), and Online & Reinforcement Learning (RL). AI courses will teach students about a different suite of techniques that do not necessarily need data to solve a problem, a complete 180 from the data-driven ML approaches they would have seen so far. Courses like Natural Language Processing (NLP), Computer Vision (CV), and Time-Series Modelling can be done optionally, depending on personal interest.
Labs are also important, as students need to have practical coding experience. The department offers plenty of labs for AIDA undergrads: Optimization Lab, ML Lab, AI Lab, DL Lab, and so on are all part of the curriculum.
T5E: How does one go about getting a DSAI minor? When will the minor courses list be released, and when can we start taking the courses? (Exciting news, read on!)
We are planning to offer three minors. They are:
- Machine Learning: A more fundamental minor focused on theoretical machine learning.
- Applied AI: A minor focused on the applications of AI including NLP, CV, Time-Series Modelling, etc.
- AI for Healthcare: A minor offered in collaboration with the Medical Sciences and Technology (MST) department which focusses on AI in healthcare.
However, the details have not been fleshed out yet. We are still working on the list of courses, so the minor is not out for registration this semester.
T5E: Are electives in the DA department open to other department students?
There are already some electives like Privacy in AI that are open to all students. Some courses like FML (Foundations of Machine Learning) are not open for other department students this semester since we already have a very large number of students taking it as part of their core curriculum. More electives will be opened up next semester and can be taken through SEAT allocation or the add-drop procedure. For people outside the DSAI department, we will probably be opening up ML 1, AI and maybe DL as electives in the future, but we are yet to finalize these.
T5E: Can we credit DSAI courses in existing minors (i.e. the AI and ML minor from the CS dept.) and vice versa?
We have yet to take a call on this. At any rate, this is currently not possible.
T5E: Are there any changes in the IDDD (Inter-Disciplinary Dual Degree) in Data Science curriculum to include DSAI department courses?
Yes, the curriculum has been changed already. Starting from this sem, many of the core IDDD Data Science courses will be replaced by equivalent DA (course code for the DSAI department) courses.
T5E: Can you elaborate on research opportunities for B. Tech. and M. Tech students? Is there a UGRC (mini-project) equivalent for B. Tech. students?
There are already plenty of research opportunities in all the labs and centres in the DSAI department. And yes, there will be a UGRC for B. Tech. AIDA students, similar to what other departments have. AIDA students can do a maximum of two UGRC overall. It hasn’t been decided yet if we should open up crediting DSAI UGRCs to other departments or not.
T5E: Are there plans for new centres in the department?
There are 6 centres of research already:
- AI4Bhārat (Artificial Intelligence for Bhārat): Developing Natural Language Processing (NLP) tools and large language models for Indian Languages (essentially indigenous ChatGPT)
- CeRAI (Centre for Responsible AI): Fundamental and applied research on the safe use of AI and responsible AI
- IBSE (Integrative Biology and Systems Medicine): Using AI/ML in the analysis of biological and clinical data
- RBCDSAI (Robert Bosch Centre for Data Science and AI): Interdisciplinary AI research in manufacturing analytics, smart cities and healthcare
- Walmart Centre for Tech Excellence: Using AI to tackle problems in manufacturing
- MInT (Mobility and Intelligent Transportation): Using AI to solve transportation problems
More centres are in the making as we speak – there is one planned centre integrating AI and Healthcare (in collaboration with the Medical Sciences and Technology department), another centre for Agricultural technology, and so on.
T5E: AI involves a lot of computation, are there plans to expand computing resources to be used by the DSAI department?
Yes, of course, the department needs more computing power! We plan on using cloud credits in addition to existing computing resources. Buying our own hardware is also an option, but that comes with several maintenance issues like cooling and having to upgrade hardware every couple of years.
T5E: Tips for developing a career in AI: what should a student aspiring to develop a career in DSAI focus on?
Kaggle! It’s very important to be able to practically code up AI models. Handling a 100-gigabyte file is not an easy task, and knowing theory alone won’t help you get better at this. Websites like Kaggle have large amounts of data and fun competitions where students can hone their coding skills in developing ML models. There are people who make a career out of being Kaggle Grandmasters, so keep practising! Internships are also a great way to get better at AI – working on a practical project is always helpful and also likely to improve your career prospects. One last thing, make sure you are always up to speed with at least some AI tools. ML has become such an applied field in recent times, so staying up to date with recent tools is imperative.
T5E thanks Prof. Ravindran for giving us the opportunity to interview him.
Edited by Shreya .S. Ramanujam
