Dr. Arpan Pal was invited to deliver his talk as a part of Chamber's effort toward providing a mutual knowledge sharing platform to its members to organize series of presentations by Managing Committee members and their organizations on new and emerging business trends and practices as part of the Managing Committee Meetings.

Dr. Pal has more than 26 years of experience in the area of Signal Processing, Communication, Embedded Systems and Robotics. Currently he is with Tata Consultancy Services (TCS), where, as Chief Scientist, he is heading the Embedded Systems and Robotics Research Area in TCS Research. His research interests included Unobtrusive Sensing for loT, Al using Sensor Signal lnformatics and Edge Processing, Al and loT driven Connected Health, Smart Cities, Smart enterprises and lndustry 4.0, Cognitive Robotics and Collaborative Drones. Dr. Pal has more than 140 publications and book chapters till date in reputed journals and conferences. He has also authored a complete book on loT. He received both his B.Tech and M. Tech from lndian llT, Kharagpur in Electronics and Telecommunications and PhD from Aalborg University Denmark. He is also a Senior Member of lEEE and is engaged in the innovations space in different industry bodies and start- up accelerators.

With the above introduction, Dr. Pal was invited to speak. He thanked the Chamber for having invited him to speak at its Managing Committee Meeting.

Having said that while loT is becoming a buzz word, loT, Al and Robotics were all evolving as the next drivers of business and our lives. ln the 1990s when internet first came in, people were involved in generation, compilation and analysis of data, which was the first part called traditional lT. Now the transition has been to generation of data by physical objects and infrastructures. Sensors came to us which were used in different ways and on different objects as per need. That is how use of sensors came into primary manufacturing processes, into the produce and in the healthcare process. The use of sensors extended from machines and objects to human beings - Apple watch.

Then came objects which can track people, example mobile phones and today it can track one's health and wellbeing, which signifies the people context. The physical context and people context together are now known as lnternet of Things (loT). The next phase of 'Context Discovery' which searched answers to two questions - "who is doing what and when?" and "what is happening where and when?". Finding answers to these two questions through data is what 'analytics' is all about. Often it is easy to do this, but many times it is not easy to this and that is where Al comes in to do that analysis. loT and Al are becoming a buzz word suddenly, and the reason for this was that the cost of sensors is coming down, Cloud is becoming mature and internet was become efficient. Effectively, there are lot of enabling technologies now available and affordable.

Robots and drones have come in next as pure sensor carrying devices, which also respond. Sense from loT, brain from Al and actuation from Robotics, this is how the system would perform. Dr. Pal explained that through their work over years they realized that the working of the system was extremely multi-disciplinary and complex. While working on projects, he said that one has to have domain experts, computer science technologists, hardware experts together as ultimately the data which would come in and get analysed would produce results. Al, Dr. Pal said, was a system's ability to correctly interpret data, to learn from such data and use those learnings to achieve specific growths. The process works through a codified computer system.

For human being prescriptive solutions are easier then predictive as data processing capabilities of human being have limitations. For computers predictive solutions are easier as it can analyse large volumes of data quickly. This is the reason why stock markets, weather predictions etc. are done faster by computer systems. Prescriptive output by computer systems are very difficult mainly because of being prescriptive calls for wisdom.

Dr. Pal went on to explain in detail the Al driven and non-Al driven systems and enablers. ln a normal computer system once data is fed an output is received. This was changing in Al. ln Al the input is provided along with an output specifying that given a particular input, the output should be as provided. There is no need of coding, the code (or called 'Label' in Al terminology) is generated by Al. For example, an ECG data can be fed specifying the areas which is/are normal and those which are not and the Labeling takes place. He explained the concept of Learning by Al, which was through (i) Unsupervised Learning - Training without Tutor (ii) Supervised Learning - Training with a Tutor and (iii) Reinforced Learning - Training with Rewards / Punishment. With the example of Google's "Did you mean..?" question in the search option, he clarified the Al training concept. Al is capable of self-learning through huge number of models.

Dr. Pal spoke about the concept of Al being there for several years buy the acceleration in Al's advent has been enabled by the devices which have become available now. Mobile phones have started coming up with intelligent technology enablers. The enablers hitherto rare and expensive have now started moving from military and critical use domain to the consumer domain.

Having explained the concept of loT, Al and Robotics, Dr. Pal went on to explain the various projects in Health Care and Manufacturing that they were working on. They had done several projects on non-communicable diseases like cardio- vascular diseases, mental disorders and so-called lifestyle diseases. Devices enable early detection, treatment and cure of these deceased. They are also working on age related diseases like dementia, collecting data from various hospitals and also through sensors on patients for monitoring, collection of data, training Al through the data for diagnosis and getting the trends verified by the doctors and going for medication again through Doctors for treatment. They are also working on Parkinson disease and drugs through Al. There are many challenges as data has to be personalized and accuracies verified. lt was also necessary to put the domain expert's knowledge with Al.

Dr. Pal spoke about projects being done in Manufacturing. They have a project for automobile engine monitoring systems, precision engineering equipment efficiency and analysis on their predictive life cycles, process monitoring in steel plant, crane operative safety analysis, welding technologies, where data is collected and Al is trained to arrive at solutions. He spoke about projects being handled by them on Robotics in manufacturing and the experiences and challenges they faced.

Having spoken about the hugely facilitating role that Al, lot and Robotics is going to play in the future, especially Al, Dr. Pal state that it was is not free of pitfalls. Failures are imminent. The important question was whether human being would learn to trust machines. Errors may prove costly and there are questions to be answered concerned assignment of responsibility of damages caused to life or property by use of Al. To address this issue, the final decision, has to be ratified through qualified human intervention. lf it is a medical decision, it has to go to a doctor and the doctor takes a final call.

Dr. Pal concluded his talk mentioning that there are concerns also with increasing use of Al but the problems and challenges which this would throw up will have to be addressed. On the general concern of job loss with Al coming in, Dr. Pal's view was that more jobs would indeed be created with use of Al - jobs of data trainers, trainers for robots etc.