The second part of my interview with Prof M Vidyasagar, FRS has appeared in Rediff
( https://www.rediff.com/news/interview/we-arent-security-conscious-as-a-nation/20251106.htm )
'We Aren't Security Conscious As A Nation'
Last updated on: November 06, 2025 14:59 IST
'Government officials use Gmail and ordinary phones without basic security consciousness.'
'Interoperability, especially in joint exercises with countries like the US, worries me.'
'It often means we open our systems to them, but they don't reciprocate.'
'They could have kill switches in their systems and might even be able to affect ours.'
Professor M Vidyasagar, FRS, is a Distinguished Professor at IIT Hyderabad. He earned his BS, MS, and PhD in electrical engineering from the University of Wisconsin.
His distinguished career includes academic positions at universities in the USA and Canada, followed by leadership roles as Director of India's Centre for AI and Robotics and Executive VP at Tata Consultancy Services.
He held a chaired professorship at the University of Texas in Dallas before his current position.
A major feather in the cap of his team at the Centre for AI and Robotics is the creation of world class digital control software for the Light Combat Aircraft Tejas indigenously.
His research interests span stochastic algorithms, convex/nonconvex optimization, reinforcement learning, and machine learning, with recent work focusing on the theoretical foundations of stochastic gradient descent and Large Language Models.
Professor Vidyasagar, a prolific author of 13 books and over 160 peer-reviewed research papers, was appointed a Fellow of IEEE (the largest professional engineering body in the globe) for his advanced contributions to Control Systems at a very young age.
His groundbreaking contributions have been recognized with numerous prestigious honours, including Fellowship of The Royal Society (FRS), the IEEE Control Systems Award, the Rufus Oldenburger Medal (ASME), and the John R Ragazzini Education Award (AACC).
He is also a Fellow of multiple national academies in India and a recipient of a SERB National Science Chair.
"Agreements like BECA (Basic Exchange and Cooperation Agreement) involve sharing geospatial data, which is very dangerous. But too much attention is being paid to getting a 'seat at the high table' than looking after our national interest," Professor M Vidyasagar tells Shivanand Kanavi in the concluding segment of a must-read interview.
How long did the control law development for Tejas LCA, take, and what did it involve?
The first cut was ready in about two years.
The airframe shape was frozen, but the aerodynamics -- how it behaves at various altitudes and velocities -- was still being understood through wind tunnel tests.
The initial development validated our design philosophy. As we got more aerodynamic data, we expanded the flight envelope.
The design philosophy was based on solving complex, coupled non-linear equations--and the most modern control theory which traditional houses like BAE didn't use.
They were using methodologies from the 1940s.
A digital fly-by-wire system means a sequence of bits is generated in real time and sent to the wings or weapons systems.
This is especially critical for a light aircraft like the LCA, where the munitions can weigh almost as much as the plane itself.
The act of firing a missile must also automatically cause the plane to pitch up to avoid being hit by it -- this must be embedded in the control law, not left to the pilot.
The control law itself isn't computationally intensive to design, but the real-time simulation is.
You must solve those non-linear equations in milliseconds to simulate the aircraft's response to pilot inputs and various failure scenarios.
This requires immense computing power, which we lacked in India at the time.
A major hurdle for the LCA was engine development, which wasn't your area. What is your perspective on why developing an engine ourselves has been so challenging?
Mine is a second-hand impression. The challenge is as much materials science as aerodynamics.
The turbine blades in a high-thrust engine operate with tiny clearances at tremendous speeds (6,000 to 7,000 RPM) and high heat.
The blades deform, so the material must have extremely rugged thermal and mechanical properties.
Mastering the manufacturing of these high-quality turbine blades is the biggest hurdle.
Once you have that, the engine control part is manageable.
I don't know if the Kaveri engine project at GTRE was scrutinised with the same intensity as our control law team.
Perhaps if it had been given a true national priority status, a multi-lab, multi-institutional effort involving universities, things might be different.
It's a glaring failure that 20 years later, we don't have an indigenous engine of required thrust.
Technology transfer can work, as seen with the Canadian natural uranium nuclear reactors; the Vikas liquid rocket engine from Ariane/ ESA; and the nuclear submarine.
In each case, Indian teams were smart enough to absorb the design philosophy and even improve upon it. But a jet engine is a more difficult technology.
Even the Chinese haven't fully succeeded yet; they still use Russian engines.
The prudent thing might be to future-proof the LCA program by having engine production facilities in India through partnerships with companies like GE or Rolls-Royce.
If they set up a plant here, it's hard to pull out. But we must also intensify efforts at GTRE.
The DRDO structure isn't well-suited for this. ADA, being a society, had some flexibility.
We once requested to pay a 20,000 rupee per month (non-pensionable) incentive to key engineers for five years.
It was denied because someone in Delhi thought, "How can these engineers make more than I do?"
This mindset -- that patriotism alone should be enough -- is a major barrier.
Lockheed Martin doesn't care what a government employee earns; they pay what's needed to get the talent.
The private sector's entry into defence production is good, but the defence procurement process is unpredictable.
It's highly predictable that it will be delayed! So companies like Kalyani and L&T are looking at export markets first for their guns and tanks.
The risk is that if their market is export, why would they stay in India?
They might move to a cheaper location. While some industrialists are patriots, they also have an obligation to shareholders.
It frustrates me that these systemic issues are well-known but there seems to be no attempt to fix them.
Let's move to your time at TCS. You worked on cryptography, public key infrastructure, and bioinformatics. What was your approach and philosophy there?
I was fortunate that TCS CEO, S Ramadorai, had a very enlightened view.
He told me, "Investment in R&D is an act of faith. Nobody can really track the return on investment."
He hired me in a very direct way.
After I left DRDO, I sent him a letter outlining technologies I worked on that could be useful to TCS.
He called me a few days later and simply said, "When are you joining?" When I asked what I would do, he said, "We'll sort it out later."
At TCS, we were a showpiece, like Bell Labs or Motorola's research wing, demonstrating technical depth to clients.
The bioinformatics group, for instance, helped win deals by showing cutting-edge capability, even if not used directly on the project.
The public key infrastructure (PKI) group became the largest issuer of digital certificates.
After the US passed the Millennium Digital Signature Act in 2000 and India passed its IT Act in 2001, I saw a big market.
Another company simply imported a foreign vendor's 'sealed jar' software. Whereas we wrote our own PKI software at TCS.
A funny story illustrates the value of self-reliance.
The NSDL wanted to use PKI for their depository record access.
They installed our software, and soon called me angrily because it was rejecting all clients.
The problem was names with apostrophes, like D'Costa or D'Mello -- common among Goan Christians.
Our software flagged the apostrophe as an illegal character. We fixed it in a week.
The company that imported their software had the same problem but couldn't fix it because they didn't have the source code.
They had to go back to the foreign vendor, who essentially asked, "How big is the Indian market? Not worth it."
That's why self-reliance is crucial.
We also worked on Indian language technology. We improved the rendering of Indian language characters in early browsers.
The Swaminathan Foundation approached us to help farmers access information in local languages.
We created a bootable version of Red Hat Linux on a CD (since power outages were common) that had a fully Indian language-friendly interface.
You could have the desktop in Gujarati but type a document in Tamil.
We gave it to them -- it was a form of CSR before the term was popular.
You also advocated for open-source software. Where do we stand on that, especially in government and defence?
I don't think we've come far at all.
I once advised Dr Kalam to mention open source in a speech.
The next day, Microsoft was very worried and called me to ask how serious he was and if this could influence government policy.
The security aspect is key. It's not that open source can't have viruses, but the good guys are more likely to find weaknesses before the bad guys.
Yet, government officials use Gmail and ordinary phones without basic security consciousness.
We are not security conscious as a nation.
Interoperability, especially in joint exercises with countries like the US, worries me.
It often means we open our systems to them, but they don't reciprocate.
They could have kill switches in their systems and might even be able to affect ours.
Agreements like BECA (Basic Exchange and Cooperation Agreement) involve sharing geospatial data, which is very dangerous.
But too much attention is being paid to getting a 'seat at the high table' than looking after our national interest.
Finally, you were recently working on applying AI to cancer research at UT Dallas. Can you explain that work?
It's based on new Machine Learning algorithms.
In cancer, you have a vast amount of data, hundreds of thousands of parameters for a single tumor sample.
But you only have a few hundred, maybe a thousand, samples for a specific cancer site.
This is a 'too many measurements, too few samples' problem. We developed a new algorithm to identify the most predictive measurements.
We applied it to predict which ovarian cancer patients would respond to platinum chemotherapy.
It usually takes 4 to 6 weeks to see if the therapy works, wasting precious time for patients who may only have 6 months.
We wanted to identify non-responders quickly.
IMAGE: Professor
Mathukumalli Vidyasagar. Photograph: Kind courtesy Professor Mathukumalli Vidyasagar
We validated the algorithm on NIH data (TCGA) and then tested it on an independent dataset from Australia with 275 samples.
I tell my students to stare at the data, not just run algorithms.
We looked and found that for 101 out of the 275 samples were wrongly entered.
We removed those 101 samples and ran the algorithm on the remaining 174. It worked perfectly.
This is a lesson: Data quality in biology is often poor compared to engineering, and you must check it meticulously.
The current AI community isn't always careful. The machine only does what you tell it to do.
If you train it on biased data you get biased outputs.
The original goal of 'machine learning' and 'neural networks' in the 80s/90s was different.
It was more about 'operational AI' -- trying to replicate the internal working of the human brain.
This failed for two reasons: 1. Our electronic switches are not reversible like the brain's chemical switches, and 2. It wasn't necessary.
The shift happened to 'functional AI' -- just getting the job done, regardless of how. The creation of large, standard datasets like ImageNet allowed people to compete purely on accuracy of image recognition, making the 'how' irrelevant.
This functional focus, combined with market hype, has diluted the original mission of AI.
A final point on the much talked about Quantum Computing and Quantum Communication.
Quantum computing is a fascinating case study of the importance of theory.
It existed only on paper for about 15 years after Feynman first suggested a quantum simulator and David Deutsch formalised it.
Theorists like Peter Shor analysed what it could; do if built, leading to the entire field of post-quantum cryptography -- devising new cryptographic schemes that would be secure even if quantum computers could break current ones like RSA.
Quantum communication, which the Chinese are advancing, is different.
It's about Quantum Key Distribution (QKD), using quantum entanglement to distribute encryption keys securely over an insecure channel.
It's not about building a quantum computer but solving the problem of secure key exchange, which is a very real and practical application.
That's a fascinating distinction. So, quantum communication is about secure key distribution today, while the full quantum computer is still a future prospect. This has been an incredibly insightful conversation, Professor.
We've covered a vast landscape, from the challenges in India's defence R&D and the story of CAIR to AI in healthcare and the philosophical shifts in computing.
Thank you for sharing your experiences and perspectives so candidly.
Shivanand Kanavi, a frequent contributor to Rediff, is a theoretical physicist, business journalist and former VP at TCS.
He is the author of the award winning book Sand To Silicon: The Amazing Story Of Digital Technology and has edited Research By Design: Innovation and TCS; a chronicle of over 30 significant outcomes and case studies of TCS R&D from 1981-2006.
Feature Presentation: Ashish Narsale/Rediff
