Challenges in India's AI Dreams.

During the interim budget today, Piyush Goyal announce that the Government is planning to set up a national centre for Artificial Intelligence (AI). While it sounds like a sweet deal for the growing tech startup ecosystem in India, it has several challenges before it is useful and much longer until it produces impact. 

Let’s start with the fundamentals of AI, it is a way in which a machine is trained to think and act like human would by considering various decision-making parameters. Most people confuse AI with rule-based logic. If you consider a simple traffic, every 30 seconds or so, it alternates between red and green light, this is a simple rule-based logic. But if the same traffic signal were to use AI and replace the need for a traffic cop who often reverts to a manual override, here’s how we go get started. The AI module first needs training data to build a model, what is the average velocity of vehicles? In what direction does it flow? What are the peak hours, when does this junction’s traffic scenario reverses? And so on. Once there’s satisfactory amount of data, a bot can be trained through machine learning models to control the traffic lights ‘intelligently’ based on the real time information about traffic. 

For an effective AI, we need credible data. That’s the challenge number one. We have had our issues with AADHAR’s security and vulnerability. Our statics and census are updated once in a decade, we never have consensus on economic data, we do not accurately know about our citizens (NRC?), infrastructure, economy or pretty much anything. So, a national centre for AI must first act as a centre that provide reliable data sets. 

Accelerating IP protection: Our second challenge is the bureaucracy associated with obtaining patents and protecting intellectual properties. If we have to compete with economies ahead of us, we must match and exceed their speed. In technology industry, having proprietary access, patents, algorithms are key factors to survival and growth of companies. 

Privacy nightmares: Hollywood’s imagination of smart cities may have flying cars and mind reading cops. But in reality, a smart city is where there are ample of sensors that help the authorities measure vital data sets and then take decisions. A vital data may include traffic flow, water and electricity consumption at household level, air and weather at pin code level, and so on. If the state is going to collect all this information and make it available for decision making, then ensuring it is not prone to hacks and leaks becomes paramount. An entire information war can be waged, should these kinds of data fall on the wrong hands. 

While India Stack initiative has been a successful journey, we sure can hope the national AI centre will be at par or better in terms of planning and deployment. Perhaps an emeritus from ISRO should lead the initiative. 

--

Sreeraman Thiagarajan is co-founder of Agrahyah Technologies. He tweets at @sreeraman

This article appeared on Sify on Feb 2nd