Shri Santosh K Misra, IAS & the CEO of Tamil Nadu e-Governance Agency spoke to Nasscom (Jibu Elias) and shared the government’s vision of massively improving on governance through responsible AI.

On being prodded about the ethical considerations of AI, he was candid enough to say that these concerns were real and justified. He gave the example of a welder using protective gear while carrying out the work. Without the gear, he risks blindness. Drawing a parallel, “protective gear” from an AI standpoint is the Responsible AI Framework which his government has crafted and wants other states to use it too while customizing for their needs. Please read on to find out how governance can be greatly improved through cutting-edge technologies such as AI & Blockchain. For instance, an impoverished man doesn’t have to drag his sick and bed-ridden mother on a cot for 10 kilometres (a harrowing incident that recently came to light) just to prove that she is alive and eligible for the government pension. AI-led technology can verify the uploaded image within a matter of seconds and impact the lives of millions of pensioners in India. 

On some of the remarkable applications that TNeGA is leading 

The TN Government is amongst the early pioneers using AI, Blockchain, and allied technologies for a better future. As CEO TNeGA says, “the vision is inspired by what’s happening in these fields since 2018 when I joined this organization. And, with support from the government, we have started to create a set of solutions using these technologies which will impact governance in a significant way. AI is a key technology and along with Blockchain, we are looking at creating digital identity”. He was of course referring to the State Family Database, a project which is envisioned on creating a digital identity for all TN residents.       

The biggest challenge to more effective governance, he says, “is being able to break departmental silos.” He adds, “silos have existed for good reasons too, due to concerns about the privacy of data, and not wanting them to go out of control”. Beyond data, these concerns extend to transactions as well and the critical need to preserve them. Today there are 40-odd departments and every one of them has their version of “truth” and supporting databases including record-keeping. Moreover, these databases have been built over a period, while some may be of more recent creations, many others were created 15 years back. Their tech stack, architecture, and granularity aren’t uniform which poses additional challenge while integrating. So every time one wants to build an AI or Blockchain solution, it has to be on the foundation of federated and harmonized databases. 

Tamil Nadu – marching towards predictive governance 

The second pillar is a Blockchain Highway which will be on top of the Digital Identity project. And, the third pillar is about delivering predictive services. The philosophy behind this is inclusivity & convenience – why should people knock on the doors of government offices to get the benefit of government services? Instead, why can’t services reach them directly? Mr Misra gave the example of the pre-natal MIS. In Tamil Nadu, this database has been able to capture 99% of the cases. Once the database has captured the details of women (pregnant mothers), eligible government services under the programme (e.g., the national version is Janani Suraksha) can be rolled out in a much smoother manner. Every trimester, the expectant mother is entitled to receive 2000 rupees; she should be able to get this amount without filing additional forms/applications/or visiting government offices. This kind of granular data capture should set off a chain of alerts related to immunization at timely intervals. In under-privileged sections, the information is lacking and often, young children aren’t administered the vaccines on time. This data can be captured as a part of the National Health Mission and it can reside in the digi-locker. Immunization as a service should reach the families without having them run pillar to post. This is a powerful example and there can be many others too – alerts sent to the family when the child turns five and the details of Aadhaar enrolment centres are provided. If the data is captured effectively, cutting-edge tools enable these services to reach citizens within specific timelines; hence they can be termed as predictive services where the government is always a step ahead in delivering citizen-centric services. This will ensure smoother delivery and provide great convenience to people.

It’s a grand vision, and the TN Government is looking towards partnering with the private sector to create the desired ecosystem and unlock great value in the process. He also mentioned the facial recognition system and the pest/disease identification in agri-crops. Mr Misra also spoke about the CoEs in cutting-edge technologies and wants the private sector to contribute towards solving governance issues. He prided over the fact that most of the tech developmental work is being done in-house.                      

Customizing for Indian conditions 

Mr Misra made a very important point – “Facial recognition systems that are available globally aren’t geared or optimised for Indian conditions. Our solutions have to work for low-resolution cameras, and in inadequate lighting conditions as well, where a part of the face may be in the shadow.” 

Their solution has worked well for schools and government offices. Especially of great benefit - during the time of contact-less attendance. Pan-India, this solution has evinced great interest, and the goal is to scale. In schools, 45 minutes a day is consumed due to the traditional system of attendance-taking. This time can now be saved. It’s an edge-computing device and to make light of the situation, Mr Misra calls it “a self-contained dabba” which one deploys and the rest falls in place.” All it needs is a source of power. Multiple applications are possible through this technology.

The agricultural app for farmers is yet another big bet for the state – it has reached 5 lakh farmers’ families. Users can access market-related information through this app. They’ve added a new module – pest & disease control. Farmers can click images of the crops which are afflicted by pests or diseases and get feedback. The challenge right now is to get an adequate number of digitized good-quality images (pest-inflicted crops) that can be used to fetch more accurate results. At present, the technology works well for two crops (paddy & maize) and 3 crop-related diseases. As he explained, “we need to do another round of data collection and expand this technology to include the top 5 crops in Tamil Nadu.”

Mr Misra spoke at length on the cataract solution – “Blindness Control Society of Tamil Nadu had reached out to us and said that they faced staff shortage. Our AI solution is very effective for routine diagnosis especially in environments that are stretched for resources,” 800 – 850 images have been used to train the system and the accuracy is between 70 – 80%. Simple 10-minute training can enable volunteers to take pictures of the elderly (their eyes) on their phones, and within a matter of seconds, the results come back. India houses 5 million blind people and 2 out of 3 cases may well have been prevented if diagnoses were timely. In low-income groups, general apathy and lack of knowledge are major hindrances to the early diagnosis of cataract. This technology, currently at the field trial stage is showing promising results. The impact is much better public services delivery that helps in early diagnosis, prioritizing cases that eventually leads to the prevention of blindness. As new data gets captured, the level of accuracy will go up substantially. 

Using AI Responsibly         

Mr Misra acknowledges deeply the pitfalls that AI can have. Towards mitigating related risks, the government has come out with the Ethical AI Policy. They are very clear that any AI solution rolled out in the public domain must adhere to the 7 Ethical Principles – the acronym is DEEP MAX (Diversity, Equity, Ethics, Privacy, Misuse Protection, Accountability & Explainability). India is a huge country and extremely diverse. This framework needs to factor in the diversity angle and a customized framework may be required for different states. But overall, the TN Government will play by these stated rules.

India must develop low-cost, scalable solutions that aid in better governance. And states should be open to sharing their learnings – TN for one, is most keen to help out other states. They are even ready to share the source codes if required – that is the level of collaborative thinking we need to embrace. 

Responsible AI is a deeply researched area and studies continue at a frenetic pace. The Deep Max Toolkit is very powerful and easy to use, and much thought has gone into it. Policy administrators can gain significantly by using it. At times it’s difficult to meet all parameters but even a simple binary answer – yes/no – to these 7 principles can be a good start. Though India is a massive data generating country, the data is held by a few entities. These entities are benefitting massively from this and not necessarily, everyone else has access to it. 

That’s why the Data Protection Bill must address these issues. For instance, who is the “ultimate owner” of the data? And, the owner of the data should be given an equitable share of the profit that the enterprise is making through its use. Mr Misra was quick to add that this is his personal view. He added, “data owners (individuals or companies) should be fairly compensated”. 

Sure, technology is a double-edged sword but we have to be open about the benefits and not worry about the pitfalls only. There have been ethical concerns around facial recognition technology but one should also look at the huge benefits. A person can shoot a video / click a static image and relay it to the system which tells immediately if the image clicked is of a person who is alive or not. This can be of immense value to pensioners across the country. Imagine the kind of outreach it can have. A postman can physically verify 10 individuals in a day (as an example) but this technology can be used for millions of people and do the job much more efficiently than humans.

Finally, he laid great emphasis yet again on the importance of having a responsible AI framework.          

Watch the video of the conversation here.