In spite of having a huge tech talent, India’s AI growth has been slow so far, compared to global leaders like the US and China. This might be because we have little access to advanced hardware or supercomputers. Organisations hesitate to invest in high-end computers or advanced technology unless they are sure of returns. Advances in AI depends on bigger and bigger databases and their storage and processing. Since we have limited high end GPUs and specialised AI processing chips, unfortunately, much of our data generated remains underutilized. Moreover, we are ‘Unity in Diversity’ and there are big data gaps in regional languages, which prevents us from scaling up in AI and making it understandable and affordable to all. Imagine if a farmer gets a message on his phone, from an app, in regional language about the potential of sowing, soil quality, rainfall probability, pest control etc., for a particular land, based on previous data storage, we may be able to create a huge crop !
Before we can think further, we are just amazed to see that China has made AI affordable and accessible and of utility to all. DeepSeek, a free, open-source AI model, whose first version was developed for around $ 6 million has given a shock to the US Industry which took hundreds of millions to develop AI models. The model’s cost effectiveness and remarkable efficiency have shaken investor confidence forcing global market players to think how fragile their investments were. DeepSeek is a large language model that heavily utilizes the “Mixture of Experts” (MoE) architecture, meaning it employs a network of specialized sub-models (experts or neural networks) where only that portion of the model which is required for the specific query is utilised, allowing for faster processing, saving energy and performing diverse tasks. Deepseek has not only challenged the big tech’s economic model but also opened up a new trajectory, that calls for a fundamental re-evaluation of how technology is produced, shared and understood in a neo-liberal economy. Can we join the same trajectory? We think we can change the AI market from oligopoly to perfect competition.
Our AI Start-up and Entrepreneurship system needs more funding to catch up with China and USA. We need to address the gaps in Infrastructure, R& D, Data, and AI Ecosystem development. We have started with “India AI mission”, “ AI for India 2030” and “AI Data Cities” to create a favourable ecosystem for AI. And we now have taken just the needed step. Ministry of Electronics and Information Technology has launched AIKosha, a secured platform that provides a repository of datasets, models and use cases to enable AI innovation. It also features AI sandbox capabilities through an integrated development environment along with tools and tutorials. The AI Compute Portal will initially provide access to 10,000 GPUs, with 8,693 more to be added, offering AI compute services at a highly subsidized rate to support
Start-ups, researchers, and enterprises. It is worth mentioning that India’s DPI framework for AI ensures ethically sourced, consent-based datasets, reducing reliance on synthetic and foreign data. The AIKosha platform hosts over 300 datasets and over 80 models, fostering the development of diverse and unbiased AI solutions.