India’s moment to lead the AI revolution

When I first started following AI decades ago, the dream was clear: machines that could think, reason, and adapt like humans. We imagined intelligent systems that could learn from first principles, understand context, and innovate on their own. Today, that dream feels both tantalisingly close and frustratingly distant.

The hype around AI is louder than ever. Tools like ChatGPT, Claude, and Gemini can write essays, pass exams, and generate stunning images. These are powered by transformers, a breakthrough architecture from 2017 that revolutionised natural language processing.

Yet here’s the truth most headlines skip: AI today is pattern recognition at scale, not true understanding. It is trained on yesterday’s data, not tomorrow’s possibilities. It lacks grounded reasoning and causal understanding – it can tell you what usually happens, but not always why.

Don’t misunderstand me. What we have is extraordinary. AI is now using itself to accelerate its own progress through AI-assisted research, code generation, and automated experimentation. Breakthroughs that once took decades now happen in months. But if neural network architecture like transformers created the wave, India must now create the ocean.

Different Destination: Transformers were never meant to be the endgame. Even their inventors acknowledge they are stepping stones. The gaps are clear. Large language models still fail on multi-step logic without engineered prompts or external tools. They do not “experience” the world; they remix what is in their training data. They cannot learn continuously from small amounts of new data without complete retraining.

Without fine-tuning or integration with specialised systems, they still hallucinate facts. Most critically, AI does not truly understand consequences, ethics, or purpose. These are not minor bugs – they are architectural limits. And they are precisely where India’s opportunity lies.

India’s startup scene is bursting with energy, but too much of it is riding the wave of “wrapping” transformers and building apps on top of OpenAI, Anthropic, or Google APIs. While this can lead to quick wins, it rarely builds core intellectual property. Worse, many survive only until the next big announcement from those platforms, vanishing as fast as they appeared.

Equal Access: Something unprecedented is happening. For the first time in human history, the most powerful tools ever built are available to everyone, almost for free – whether you are in Silicon Valley, Bangalore, or a small town in Kerala. The old world of information asymmetry is gone. The only difference now is how fast and how boldly you use these tools to push the boundaries of research and creation.

This is India’s moment to build startups that create new types of AI from the ground up. Instead of just using existing AI tools, we should focus on building systems that can reason like humans, work efficiently on phones and small devices, understand both images and text together, and solve problems specific to India – such as helping farmers increase crop yields or making healthcare more affordable for rural communities.

New AI technologies are already emerging that work differently from today’s chatbots. These newer systems can remember things better, understand context more deeply, and reason through problems more effectively. At the same time, the infrastructure needed to build and run these advanced AI systems is becoming more accessible, allowing small teams to attempt breakthroughs once possible only for tech giants.

Moonshot Moment: India has everything it needs. We produce some of the world’s best AI researchers and engineers. We have problems worth solving – from climate resilience to affordable healthcare – that demand innovations beyond today’s AI. And the timing is right, as global AI leaders are themselves searching for “what’s next” after the transformer era.

If we want AI that can reason through real-world uncertainty, learn from minimal examples, make discoveries in science and engineering, and align deeply with human values, we must invent new architectures. This is India’s moonshot moment in AI research.

What we need now are research-first startups that invest in long-term breakthroughs, not just short-term features. We need government and private R&D funding directed at fundamental AI science. And we need collaboration between academia, startups, and industry to test new paradigms in real-world environments.

The question for Indian founders is simple: do we want to be consumers of AI invented elsewhere, or do we want to be architects of the AI the world will use tomorrow?

New Horizon: The choice is urgent. If we focus only on fine-tuning models built elsewhere, we will remain consumers in someone else’s revolution. But if we blend AI’s raw power with new approaches to reasoning, learning, and multimodal intelligence, we can be the architects of the next era.

The future is wide open, but the window is closing. It will not be won by following the map. It will be won by drawing it.

Kerala prescribes a MedTech boost

One thing we at TikTalk News often debated was how to create a platform where our medical and engineering students could mingle, so that the MedTech sector would gain firmer roots as they shared ideas. Kerala government agencies have now taken a step in that direction by issuing an order to set up IEDCs in medical colleges across the state and to establish a Health Tech Hub at the Apex Trauma and Emergency Learning Centre (ATELC) at Trivandrum Medical College. A report in The New Indian Express says IMA officials have held talks with IIT Palakkad, NIT-Calicut, Cusat, and the College of Engineering, Trivandrum to take this idea forward.

This move is timely as New Delhi is also pushing the private sector to produce more medical devices and pharma products locally, since the country still relies heavily on imports. The Central Drugs Standard Control Organisation (CDSCO) recently released a revamped Subject Expert Committee (SEC) Guidance Document aimed at enhancing transparency and patient safety. The document provides a framework for the operation, composition, responsibilities, and decision-making processes of SECs. Hopefully this will address long-standing issues such as unclear drug approval norms and inconsistent regulations that have been holding back the industry for some time.

Cureous, Olee Space get funding

It is good to see the deeptech sector attracting investments. Among the recent beneficiaries are health-tech startup Cureous and defence-tech startup Olee Space. Cureous, founded by Asish Mohandas, focuses on automated patient repositioning beds aimed at preventing bedsores and improving caregiver efficiency. It has raised 1.66 crore rupees. Olee Space, founded by Solomon James, has raised funds worth 3 million US dollars in seed funding for the production of its laser-based quantum communications technology and for advancing its directed energy weapon systems. Another notable funding was for Comminent, a Bangalore-based startup that will use the funds to expand its device-agnostic IoT communication solutions for smart cities, starting with large-scale smart metering projects.

Nisar satellite continues to bloom

More updates on the joint Isro–Nasa project, the Nisar satellite. The spacecraft’s 12-metre drum-shaped antenna, the largest ever sent into space, has been deployed successfully. Stowed umbrella-like until its 9-metre supporting boom was extended and locked in place, the reflector was released on August 15 when explosive bolts triggered the “bloom” — its unfurling as stored tension in the flexible frame was released. Motors and cables then pulled the antenna into its final locked position, Nasa reported. Weighing 64 kilograms, the reflector is made of 123 composite struts and a gold-plated wire mesh.

Bot Olympics deliver fun and data

Who would’ve thought people would shell out 25 to 80 US dollars just to watch humanoid robots trip over themselves and chase a football? Yet that’s exactly what happened in Beijing at the first-ever “Humanoid Robot Olympics.” Teams rolled in from the United States, Germany, Brazil and beyond – 192 representing universities and 88 from private companies – all ready to flaunt their metallic prowess. The games weren’t all eye-popping performances, though. There were tumbles, face-plants, and enough clumsy kicks to make any kindergarten team look like pros. But before you dismiss it as slapstick comedy, organisers insist the chaos had a purpose. Each fumble fed valuable data for real-world uses – proving that today’s bumbling bot could be tomorrow’s efficient co-worker handing you parts on the factory floor.