Where do we stand in terms of technology adoption?

What AI has been able to do is prove that in today’s day and age, it is indispensable. It is an integral part of business, life event. How we interact with people, how decisions are made are determined hugely by AI – be it from the choice of music we make to the movies we watch online. Every industry has done some small prototype to test whether AI can be used in a field such as chemical composition of drugs, tone and colour of paintings, casting decisions made by film directors, understanding and mobilizing supply chain, agriculture, healthcare. But it leads us to wonder - Is AI a big play here? Not yet. We have proven its utility, but scaling AI’s comes down to data. The adage goes – without IA, there is no AI. Without Informationa Architecture, there is no Artificial Intelligence. Its needs your own data to learn your own processes and how you make decisions. There need to be well defined processes to collect meaningful data that can be fed into the system. This entire architecture sometimes needs a business process revamp, which is no small task. Data cleansing and collection is actually an elaborate and time consuming process. The lack of a structured information architecture in sectors where we need AI, is a major deterrent. Even if this IA isn’t in place, there needs to be data validation and trust factors. The models have to have some element of trust. The third critical aspect is explainability. All three AI techniques have to get incorporated into recommendations like in an HR process. Let’s say an organization is trying to understand how to hire someone – how can this be made explainable? That’s the big question here while wondering about the application of AI. A lot of focus today is about getting the IA in place, establishing trust and explaining data models. We need to do a lot more work in these areas. AI techniques can be used to enhance AI adoption – like AI4AI. Learning from less data and increasing bias detection methods can help push AI adoption. In India, many tech ecosystems are playing with AI, and using it extensively for recommendation systems and consumer facing apps. They do need to go beyond human interaction layer. But one major point of interaction is understanding local dialects, emotions and culture. We need to look at exploring these features for Indo languages – this is where the next big opportunity lies for AI.

Scientists like Gary Marcus think Deep Learning and reinforcement learning is reaching its limit and data heavy models are not viable for the next leg of AI. What do you think is the next frontier in AI research?

Gary is a big proponent of hybrid learning – where the journey is mapped from learning from example to learning from abstraction and patterns. There are two schools of thought – first being, can you learn abstract patterns from within a domain in deep neural net and learn from a related domain that’s data rich? Then, can some of these learnings be transferred in a prioritized manner? We are increasingly dealing with scenarios where there is little available data in a single domain but plenty of data in a related domain. So can learnings can be transferred? Examples are healthcare and IT support, being developed simultaneously with AI. How truly related are these two seemingly diverse fields?

Another piece we’re focusing on is the blend of neuro symbolic research – this is being done extensively by MIT, Boston. Neuro symobolic techniques to vision techniques are being applied to Q&A style problems. In addition, there is an increasing amount of work on integrating background or common sense knowledge, and exploring using neural models to learn logic. While Gary’s approach is more hybrid, the blending of neural and symbolic is another popular way of bringing the data together more effectively, and directly addressing the challenge of learning with lesser data.

Are gender roles in AI skewed? What do you think are the consequences of this?

We need more women participation in STEM, and this is not just about inclusion alone. We need to think of diversity in data too. Else, five yrs from now, we’re going to be reading studies that say AI has a definite gender bias. Just two years ago, when I was participating in a Girls for STEM programme, a study came up that there is gender disparity in education. In the tech community, we don’t want a gender disparity in AI. The bigger problem is to solve it in STEM and encourage women in science, and focus on making STEM a long term career. Technology is so dynamic, you have to constantly learn and unlearn to stay ahead. Affirmative actions need to be taken at all levels – esp in India’s lowest middle class. We need to make technology as inclusive as possible for women, and learn to accommodate/work around gaps a woman may have such as maternity breaks or childcare. Studies show that girls are more driven, especially to affecting change in society so this needs to be channeled for the greater good of all.

How has your journey been as a technologist?

When I graduated from school, it was clear to me the choice I had to make. I wanted to do something that kept me thinking. In college, coding was beginning to become the next big thing, and I developed a passion for it. But what I have also realized is, to build on this passion, it is vital to communicate with people in the field. As young adults, we miss out on getting the right kind of guidance and experience from our elders when it comes to building our career. Very early on, what we know now as mentoring, I used to have casual chats with an elder in the family, who at the time had a PhD in graphic designing. I was able to understand the wider ecosystem pertaining to the life of a PhD aspirant and what it entails, the challenges and opportunities. These interactions shaped my views and led me to develop greater insights on why pursung higher studies and picking a specialized area is vital for career longevity. When I decided to pursue my PhD, systems networking was the “in thing” in tech at that time and I tried my hand at it. But after a couple of years, the trends were more in favour of optimistion modelling. So there are will be many stages in one’s professional career where one feels like he has to take the gamble – of continuing with one’s job as a comfort thing, or push boundaries and learn something new to have a competitive edge in the future. The latter approach really helps you unlearn, and be humble. Every now and then, there will be a point where you have to decide how to invest in the next three to four years of your career, and this is especially the case in technology. Sometimes, you can make a wrong turn, but sometimes the risk will pay off and you would have learnt something new.

The ultimate question – will AI take away jobs?

AI will make some skills more important, and some redundant. At the same time, AI also creates a lot of new jobs like data science and data stewards. Even without AI, there was a risk of certain jobs becoming obsolete. AI isn’t the cause for jobs to go obsolete, we should keep this in mind. A similar hysteria took place when computers came along. But since then, it has created a new generation of technology and adopters. New jobs will come up and AI will help create those jobs.

• Interview done by Jibu Elias, written by Sindhuja Balaji