Transforming governance: The AI pathway
This article is authored by Gaurav Goel, founder & CEO and Rahul Kulkarni, chief technologist, Samagra.
When tackling the needs of 150 crore citizens, technology acts as a powerful tool with unprecedented impact. Take education, a key driver of progress, which often faces complex challenges. For instance, in Himachal Pradesh a few years ago, the problem seemed simple: Too many schools and not enough teachers. But the real issue was more intricate—schools were clustered too closely, and some had more teachers than needed, while others faced shortage. Enter optimisation algorithms, which streamlined the number of schools and teachers, merging 400 schools and greatly enhancing teaching quality.
Consider the challenge of managing the transfer of about one lakh teachers in Haryana's government schools. How do you coordinate such a vast movement effectively? Similarly, in agriculture, how do you allocate 3000 village agriculture workers across 30 districts in Odisha to ensure balanced workloads and efficient last mile delivery for farmers?
Machine learning algorithms have tackled these issues with precision. Today, Haryana handles teacher transfers through a fully automated system, while Odisha uses advanced optimisation algorithms to distribute the workload of village agriculture workers, ensuring fair and effective service delivery.
Building on these successes, the recent advancements in Artificial Intelligence (AI) have shown remarkable potential for enhancing individual productivity. But ever wondered how the potential of AI can be unleashed at a population scale in the context of governance to transform the experience of citizens?
We can break this down into three parts, based on the type of AI model being deployed: First, there is interpretative AI--this involves AI's ability to understand and process human input. For example, it includes technologies like Convolutional Neural Networks (CNNs) for reading handwritten homework, or analysing and interpreting video content. This can provide personalised, real-time feedback to students and adjust their learning trajectory based on their individual needs. Second, is cognitive AI--these are machine learning algorithms being utilised for enhanced problem-solving and decision-making by mimicking the human cognitive processes. For instance, by analysing trends from historical data, tools like TimeGPT or TemporAI can forecast future demand for healthcare services, plan resource allocation, and manage staff schedules more effectively in a given district hospital. Lastly, there is generative AI- this involves generating new content or data based on patterns and information from existing data. For example, in the state of Odisha, through the Ama KrushAI chatbot, village agriculture workers and farmers today, have access to personalised advisory services in Odia, Hindi, and English on best agronomic practices, government schemes, loan products among others at their fingertips.
This represents only a glimpse of AI's transformative potential when applied to real, complex, on ground governance challenges that impact millions of lives. As India sets out on its ambitious goal of becoming the AI use case capital of the world, five key pillars will shape its future trajectory:
- Approach: AI should ‘enhance’, not ‘replace’, human capabilities in governance, complementing rather than substituting human judgement, especially in critical tasks affecting lives.
- Building capability and capacity: Just as foundational literacy and numeracy forms the bedrock for education, India requires a comprehensive approach to AI literacy which cuts across grades, trades and professions. Establishing state of art (ISRO-like) institutes in every domain working on strategic use cases of AI will foster collaboration between academia and industry, maximising value for all stakeholders.
- Cost: Cost efficiency is key in scaling AI technologies. The inference cost, the cost associated with running a trained AI model, on Graphics Processing Units (GPUs) can be both high and inefficient. Thus, maintaining a lens of frugal innovation is important while building population scale solutions. One way to achieve this is by using smaller models that can operate directly on mobile devices, thus eliminating the need for costly cloud-based GPUs.
- GPU availability: AI scalability, specifically at the time of training AI models, is hindered by limited GPU availability. There is a need to thus democratise GPU production. Embracing specialised chips for specific tasks, such as Language Processing Units (LPUs) or Tensor Processing Units (TPUs), is important. By designing and manufacturing these chips in India, ideally developing them in the open-source way like RISC-V, we can strengthen our domestic technology capabilities.
- Responsible use of data: To ensure ethical AI development and safeguard sensitive data, consent is paramount. We need to effectively utilise the Digital Empowerment Architecture (DEPA) framework established by MeitY, which underpins consent mechanisms in platforms like Digilocker and Ayushman Bharat, to build and maintain public trust.
Once we secure these foundational aspects, we open the door to the full spectrum of AI’s transformative potential. Imagine an AI-powered society where healthcare is revolutionised: Portable diagnostic tools in apps instantly capture body images, monitor activities, and perform preliminary health screenings. An AI co-pilot assesses the results, offers advice, or escalates urgent cases to nearby hospitals. At the hospital, you receive prompt, accurate care without long waits or repeated explanations. Meanwhile, the public health organisations analyse patterns to identify at-risk populations and detect endemics early, enabling proactive service delivery. Similarly, in an AI-powered society where agriculture is revolutionised: Farmers use satellite imagery and drones to monitor crop growth and soil health in real time at the farm plot level, receiving accurate and timely advice on irrigation and pest management, among others that enhance their farm’s productivity. Similarly, in the domain of education, teachers are equipped with tools to measure their students' learning levels real-time, made possible through spot assessments in their vernacular language. The possibilities are endless and unfathomable, because AI has the potential to change the world more than anything so far ever has.
This article is authored by Gaurav Goel, founder & CEO and Rahul Kulkarni, chief technologist, Samagra.