Talks

Key Note Address by Prof V Kamakoti, Director, IIT Madras

Read Time:8 Minute

MMA 21st All India Management Student Convention 2022

Key Note Address by Prof V Kamakoti, Director, IIT Madras

When I started teaching computer hardware architecture to my undergraduate students, some of them were not really interested in it and they just wanted to clear the course, as they thought it would never be used in management. But understanding of engineering discipline, especially computer science in the context of AI and ML, is now becoming extremely important from a management perspective. That’s perhaps the reason why engineers become very successful management professionals.

Understanding AI through Chess

Today, as we look at the world, there are many disciplines which, once upon a time we thought we could handle using basic understanding of processing. For example, chess is a highly rule-based game. How do computers play chess with human beings? For every possible move the human makes, the computer has some K possible moves, let’s say K1 to K6. There can be other set of possible moves from the human, for which there can be another set of possible moves by the computer. So if you look at six such possible moves between each, then the total number of the combinations will be K1 x K2 x K3 x K4 x K5 x K6. Just by sheer computing power, the computer will calculate the possible move by the opponent and for that every possible move, it will calculate what its own move would be. Based on that, it can look 6 to 7 moves ahead, find out which move will be the best for it and then it will make that move. This is how the computer chess game evolved. According to older versions of chess, a level 1 player will look at three moves ahead; a level 2 player six moves ahead; and level 3 player 10 moves ahead. The amount of time the computer takes to respond to a move that a human being makes increases as you increase the level. This was the first way by which intelligence was built into a game like chess but it is highly process-oriented. There are rules and we just have to follow those rules. The computer decides the best choice just by following the rules in all possible ways.When it came to things like rapid fire round, the computer also had a limitation in terms of the time it took to make the move. The computer started sensing the pattern. ‘Rather than looking at ten different moves away and making a choice, can I look at the pattern and make the best move?’ This essentially led to a transition from a rule-based system to an intelligent pattern analysis based system. And that is when real machine intelligence started building up. Today, we can have programs that will learn from themselves. The evolution of chess is indeed a very interesting field to understand AI in its proper perspective.

Coming back to management, organizations are indeed going to become more and more digital in nature. For every activity that happens in the organization, we’ll have a sort of digital signature or a digital trace. When you recruit somebody with a certain degree from a certain institution, after the person joins the organisation, his performance appraisal is captured in a database. From this, if HR can predict the performance and progress of a person who is recruited from a certain institution, it can help them to decide to which institution they should go next and what sort of reskilling the people need to undergo at a regular interval. You can learn many things by understanding the data, processing the data and identifying patterns in the data. Then you can quickly come to some conclusion. That is what we call business intelligence.

Use of AI to address NPAs

A much more interesting example would be the application of AI in the banking profession. Today, one of the important problems the banking sector is facing is NPA – Non Performing Assets. Is there a way to predict if I can get back the loan given, by doing a reliable credit appraisal while giving out the loan? This is a million dollar question. If this question can be easily answered, then our economy would have been much better but it depends on many factors. The banks today have enormous amount of data that includes both positive and negative data. Positive data point out that the money loaned out had been promptly returned and negative data point out that the money that was financed did not come back. If a person cannot repay the loan, banks go for settlement to get whatever they could make out of that. We can start using AI for this most important problem of addressing the notion of NPAs. A lot of things have evolved. There are risk management systems that are put in place which look at the institution as a whole and not individuals, like Basel’s Level 1, 2 and 3. If all these theories had worked properly, our state of economy should have been much better but they don’t really work. That is why AI is certainly going to play a very important role for the managers in the BFSI sector.

If you are always taught to think negatively, you will only think negatively. That is true for a computer program. If I feed the computing system only with negative examples or patterns, for instance, of people who received the loans but did not repay, then the entire machine learning AI engine will look at these patterns and will always predict that every loan will not be returned, because it has only seen negative examples. So when we try to build these types of AI engines to manage NPAs, we need to feed positive examples also. We call this ‘balance of data.’ We need to teach the computing engine that there are both positive and negative examples. Based on that, the computer will aid the credit committee whether a loan could be given or not given; or if they should go for a settlement or not.

Will AI replace human beings?

AI will definitely replace human beings in mundane and stupid tasks which the human being is not supposed to do. For example, in earlier days, a clerk in a bank could be promoted to the next level, if that person could count accurately 500 numbers of rupee notes accurately in a certain time. Today, we have a counting machine, which automates this process. Counting notes is a mundane task and one will normally get frustrated if one keeps on counting notes in their job. When we evolve any manual process, that process will need a lot of innovation but over a period of time that process can become so well-oiled that we don’t need a human intervention. Then we can automate it. This is precisely why AI is going to be here.

Deep Thinking

Garry Kasparov, chess grandmaster and the former world chess champion, came out with a book called Deep Thinking.  One example that he gives there is that he goes into a room where there are 100 computers and he plays a rapid-fire. He goes to the first computer and plays, then he moves to the next computer and plays; he goes on playing with 100 computers. Now, the question is, ‘can we have one robot playing against hundred Kasparovs?’ If computers can replace human beings, can we make the opposite also true? If we have 100 Kasparovs with chess boards and one robot to move around, playing with those 100 Kasparovs, what are the types of problems that we will hit? The robot has to move around from one table to another. It should first recognize the table, then the chessboard and the coins. It should properly hold the coin, lift it and keep it in another point. This involves a complex problem to solve which, perhaps every discipline of every IIT must come together. Therefore, we can safely say that AI can only be an aid to human beings and it can never replace human beings, at least in the near future. This is an important input that we get by reading books like the Deep Thinking.

Learn Data Science

Lastly, the most important foundation that management students need to understand data and then subsequently AI, which will help them build a career around AI, is data science. A very good understanding of data science and programming is extremely important to build a worthwhile career where you can appreciate and make sensible use of artificial intelligence. Data science is not just Excel sheets where you start plotting some pie charts. Data science involves deep mathematics; good understanding of statistics and probability theory; a lot more understanding of Behavioural Sciences and many more things. The UGC has come out with a very good order which says that you can do two degrees at the same time provided your time tables don’t clash. Management graduates must surely do a serious course on data science and try to get a degree in it. There are now many online degrees. IIT Madras offers a BS in Data Science program, which is available for all. You just have to pass a qualifying examination. There is no entrance examination. Importantly there is no age bar. Presently, students from the age of 17 to 81 have attempted at joining this course. We have seen father-son; mother-daughter; father-daughter and such combinations of people doing this course.

Exit Options

It has multiple exit options. If you exit at the end of the first year, you get a certificate with an understanding of foundational aspects of data science. If you complete the second year, you get a diploma, which is much more advanced than what you have learned in the foundation course. At the end of three years, you get a BSc degree and at the end of four years, you get a BS degree which is equivalent to any university degree across the world. More than 15,000 students are now enrolled in this program at IIT Madras. You are also welcome to join and be a student of IIT Madras.