Artificial intelligence (AI) is disrupting life and work, prompting governments and policymakers to take note of its potential and formulate ways to harness its power for greater common good. Last August, the Centre set up a task force to explore the possibilities of using AI for India's economic transformation. The 18-member task force, headed by V. Kamakoti, Professor at the Department of Computer Sciences, IIT-Madras, has been asked to come up with suggestions to drive development in AI research, development and applications. The first report submitted by the panel was released recently. K.T.P. Radhika spoke to V. Kamakoti on the task force, its mission, and AI in India. Edited excerpts:
What was the modus operandi of the AI taskforce?
A: We identified 10 domains ranging from manufacturing, healthcare, agriculture and food processing to environment and national security which are core to the development of the country. Then we found out the challenges in implementing AI in each of them. We also invited public opinion through our website and identified problems that are common to each of these domains. The first report we submitted to the government elaborated specific challenges in adopting AI-based systems, key enablers needed for deployment of these technologies, ethical and social safety issues, and strategies the government can adopt to address these issues.
What is India's AI vision?
A: AI should be seen as a socio-economic problem solver rather than an economic booster. Once we start solving problems through AI intervention in the best possible way, it will lead to good economic growth. AI should solve problems effectively and not create new problems.
Compared to the US and China, Indian work in AI is largely restricted to data analytics and low-end AI. What are our specific challenges?
A: If you look at the entire AI-based system, only 5 to 10 per cent will be core AI. The rest would be system engineering. Today, we have a lot of automation and there is a very good team that can do this core AI. What we lack is ability to bring in the whole system integration. We need a design prototype that has to integrate the whole system into it. When I make a system, I need the whole system engineering. Another India-based challenge is lack of clean data. In manufacturing and agriculture, we need a lot of clean data. In developed nations, they have collected a lot of data, and are applying their convictions on it. In India, the system and convictions are different.
India is a peculiar, conservative and beautiful market. We did not crash when the world crashed. That said, currently, availability of clean data is a major issue since the country is very vast. In automation, there is a lot of data. But if we want to make it intelligent automation based on the data and to make some decisions, I have to take that data, find a methodology to interpret it and make some decisions. And that decision has to be conveyed back to automation for implementation. We have to close this loop.
Once we start having clean data, this will happen. In our report, we have given a lot of samples from India and suggested that with good funding we will be able to catch up with the advanced AI world.
One of the major recommendations you made to the government is forming a National Artificial Intelligence Mission (NAIM) with a funding of `1,200 crore. What will be its function?
A: NAIM will act as a nodal agency for all AI-related activities in India. The finance minister, in the last Union Budget, said that Niti Aayog would be responsible for that. Niti Aayog has already formed an AI committee and will formulate policies. This is a single-window system where every ministry will be involved. Next, we are going to get a list of grand challenges for India in each domain. Then we will map this to the specific ministry. This mapping is important, as enablers and projects of social relevance in each domain need to be owned by a single or group of ministries to ensure correct implementation. Ministries can also fund to bring it to the prototype stage and take it forward. For example, when the agriculture ministry identifies a problem in the farm sector and an AI solution is developed with it, chances of it reaching farmers increase manifold. This will also have industry participation. Each of these domains will expand with the guidance of Niti Aayog now. Also, every ministry will then have an AI wing which will help that ministry and corresponding industry to adopt AI.
The mission will be for five years and we have recommended a funding of Rs 1,200 crore - Rs 240 crore for each year. Of this, `50 crore will go for core activities, `25 crore per six centres of excellence, and `20 crore for establishing generic AI testing beds and a large data integration centre.
Can you elaborate on the major activities of NAIM?
A: The mission will have core activities like forming a network of alliances from services industry, product industry, start-ups, ministries and academia to form knowledge clusters. We should also go out and fund national-level studies to identify concrete projects in each domain to address important social issues. Tools for cataract detection, automation of hazardous jobs like manual scavenging and disaster recovery using AI robotics are some examples. We have also recommended setting up centres of excellence for each industry in line with the RIKEN Center for Advanced Intelligence Project in Japan. These centres will be the biggest enablers for AI adoption. As AI is highly interdisciplinary, these centres will promote interdisciplinary research to facilitate deeper understanding and knowledge of AI and adoption of AI-based technologies. Another function of the mission is to create a generic AI test bed so that developers can bring in their AI technology and test whether it meets current needs.
Today we are getting a lot more data with initiatives such as Digital India. So, we are looking to form a large interdisciplinary and dedicated data integration centre. This can work on multiple data streams in real time to provide information and predictions to public. We also need to have data bank exchanges and an ombudsman to resolve data-based conflicts. We should also set standards to data which would be highly interdisciplinary. These should be strong policies so that people can confidently use data.
How are you going to address the concerns around AI's impact on, say, jobs?
A: We strongly believe people come first, then process and then technology. Technology is not a substitute for humans, it just magnifies human capacity. Humans have to intend to use technology, unless this happens we wont be able to bring AI in a big way. For this, we have to create awareness. We have suggested AI yatras, especially in rural areas along the lines of SpicMacay lecture demonstrations, to create awareness. We have also suggested convening talent conferences with interdisciplinary insights.
There is concern that AI will cut many jobs. But AI will also create a lot many jobs. There is an interesting case study in our report about TT Consultants and XLPAT. It shows that their number of employees increased several fold from 2012 to 2017 after adopting AI. For this to happen, we have to improve skill sets of people and make them job-ready.
We should identify skill sets required for AI-based technology. We need an education strategy to develop human resource with the necessary skills. The National Skill Development Corporation can lead this mission. Another focus point is education.
How will AI impact India's educational scenario?
A: There are two aspects - AI for education and AI education. AI for education is a big deal. AI will help check whether education has properly reached the student or not. Today, scaling up is an issue because the number of students is increasing. The personal attachment of guru-shishya has gone. AI should play a bigger role in bringing students and teachers together.
Institutes like IIT have been doing it for the past 18 years. But why has AI suddenly got this importance? The point is simple. In earlier days, AI was rule-based. Take a washing machine. It was intelligent. It knew it should use 7.5 litres water if you put 10 kilos of clothes and 10 litres if you put 15 kilos of clothes. However, data was not available in full form.
Modern machines can learn. I have some data and machines that can learn using that data and apply it on other data. Earlier, we did not have that large storage spaces. Now, data storage can run into terabytes or petabytes with fast retrieval. In the late 1980s, Daniel Hillis from Thinking Machine Corporation presented the thinking machine. It took more than 24 hours to recognise a key as a key. Today, technology has improved and a lot of data is coming in. Now we are marrying AI with large-scale data. That is machine learning joining hands with data analytics.
While some large companies, e-commerce players and some start-ups are exploring AI, traditional Indian industries and government are not aware of the potential and are not investing. How are we going to tackle this?
A: The traditional industry will take interest once it understands the potential. Once they realise that by adopting AI, they can cut costs and improve quality, they will invest. Cost cutting can also happen by doing the job faster, improving productivity and reducing power consumption. We have asked the government to frame policies to enable traditional industries to adopt AI, like giving tax incentives for income generated by adopting AI. These policies will bring confidence to adopt AI and once that confidence comes in nothing will stop India from becoming an AI-driven economy.