“A sentence can either be positive, negative or neutral – not a mix of all three. Emotionality, on the other hand, has this exciting thing about it that it can have a mixture of emotions within it,” said IIT Bombay professor and computer scientist Pushpak Bhattacharyya, explaining the complexity of understanding emotions in language.
Earlier this month, OpenAI CEO Sam Altman was seen experimenting on X, where he posted, “Is there a word for feeling nostalgic for the time period you’re living through at the time you’re living it?” The next thing you know, everyone was on ChatGPT asking what the word was. And many demonstrated creativity crafting their own versions of the word – like ‘Nowstalgia’. This is truly human. But can it be implemented on LLM chatbots?
Bhattacharyya knows the answer. “Chatbots that are polite, and understand sentiment, emotion, etc give rise to better businesses. Chatbots that are closer to human beings, emotional and sentiment, bring commercial profits along, which is quite motivating,” he added, highlighting a study that reported that organisations that used polite chatbots, benefitted from them, instead of the generic ones.
Bhattacharyya has been working on these emotional and sentimental problems in NLP since his master’s at IIT Kanpur, and has published over 350 papers. “What got me interested in linguistics, emotions, and AI was the similarities between the words of different languages and their respective sounds,” Bhattacharyya said. He narrated how he used to collect proverbs from each country and city he visited to understand how language operates.
Bhattacharyya told AIM that he is also working on Plutchik’s wheel of emotions with eight emotions at Centre for Indian Language Technology (CFILT) lab at IIT Bombay, which is a subsequent work of his recent paper – Zero-shot multitask intent and emotion prediction from multimodal data. This problem deals with combining different types of emotions within one context, a foundational problem that he said no one has taken up before. “At our CFILT lab, we take up problems that no one else has before, which includes not just Indian languages,” he added.
What Indian LLMs need
Bhattacharyya emphasised on building a trinity model for creating Indic language models, which means deciding one language, one task, and one domain for creating models. “For example, creating a model in Konkani for question answers on agriculture or a sentiment analysis system for railway reservation in Manipuri,” he explained, saying that these models are easier to build and then can be connected later into a larger model.
Talking about Indic models built on top of Llama, Bhattacharyya said that though these models are a step in the right direction, it is essential to also build them for specific tasks and domains, and then gradually expand into other tasks.
Bringing Altman’s recent news about raising $7 trillion for making AI chips, Bhattacharyya said that India also needs similar efforts in making indigenous chips. “Being self-sufficient in hardware is very crucial because we cannot wait for the switches and GPUs from outside. We should make in-house capabilities that will facilitate AI research and further development,” he said.
AI Awareness
“In our scriptures, buddhi (wisdom) is above manas (mind) and then comes our body and sensory organs,” Bhattacharyya said, explained that every generation of students is smarter than the last and the volume of information also increases. “It is important for these students to learn the basics and not just get stuck with what is exciting, for instance instant gratification tasks like programming,” he added, talking about the need for students to focus on larger problems, but by starting small.
“Introducing basic mathematics along with programming starting from Class 5 is necessary for education systems, along with teaching them about every other field like social science and others, to give them proper alignment with the world they are building for,” added Bhattacharyya about having holistic education for students.
During his PhD, he spent an extensive amount of time at MIT AI Lab and Stanford University, studying different flavours of AI with pioneers in the field such as Marvin Minsky and the father of modern linguistics, Noam Chomsky. “But my interest in linguistics started way back when I was in Class 4 in school,” said Bhattacharyya.
“Today’s AI takes a lot more computation when compared to when I started doing AI,” Bhattacharyya said. He also highlighted how Bill Gates said in the early 1990s that NLP will drive computation requirements forward. “My advice to young people starting in the field would be to understand the fundamentals of mathematics, build foundations and then stick to a problem for a longer time,” he added.
A Complex Human
Starting his BTech at IIT Kharagpur studying digital electronics, Bhattacharyya came across a circuit board made for adding two numbers. Unlike others who did not think of it much, he was astounded at how a lifeless system made of diodes and resistors had decision making capabilities. This got him into studying intelligence outside bodies of human beings and animals, leading him to AI.
“I’m one of the few NLP researchers who give equal importance to linguistics and computation,” he beamed. He added that his course is inspired by both the fields and how his Master’s thesis was also focused on Sanskrit to Hindi machine translation. Talking about his paper on sarcasm detection in 2017, Bhattacharyya said that the researchers were able to build a computational algorithm that could detect sarcasm in LLMs which would benefit the field of psychology, cognitive science, and philosophy as well.
