The need to build sovereign AI capabilities

During his recent visit to India, Arvind Krishna, chairman and chief executive officer (CEO) of IBM, said India needed to develop sovereign capability in artificial intelligence (AI), including Generative AI. Indian policymakers need to take his advice seriously. AI — especially machine learning (ML) and deep learning — has increasingly been used in the past few years for everything from e-commerce recommendations to self-learning cars. But the entry of Generative AI — ChatGPT and DALL-E from OpenAI, Bard from Google, Claude from Anthropic, Llama 2 from Meta, and others – has kicked off a whole new ball game.
 

Generative AI models with billions of parameters can train on unstructured data to create new content, give coherent answers to queries in natural language, find new patterns from masses of jumbled-up data, and come up with options and answers that can revolutionise everything from drug discovery to video and audio creation. It is likely to have a deeper impact on the global economy and society than other technologies of the recent past.
 

While many Indian companies have enthusiastically jumped onto the Generative AI bandwagon — from information technology (IT) services giants like Tata Consultancy Services and Infosys to a dozen-plus startups — no Indian company so far is talking about developing its own foundational models. Indian technology firms, and even the Indian government, have focused far more on the applications space. Various arms of the Indian government have also started building specialised data sets and Cloud computing infrastructure and initiated programmes to build innovative new applications. Sticking primarily to applications would be a mistake. India must also develop capabilities in building new, indigenous foundational AI models. It also needs to start research on the next frontier of AI — General AI. Building innovative applications is immensely important, but it is not enough.
 

As this columnist had written earlier, AI is a general-purpose technology (GPT) and GPTs tend to dramatically alter the power dynamics in society and business. Countries that dominate a GPT become techno-colonialists, while those that depend on others for access to these technologies end up as techno-colonies. (Haves and Have-Nots in AI, Business Standard, July 8, 2020).
 

While the US has typically held the pole position in technology research — digital and others — China has been working hard for over a decade now to build its own sovereign capabilities in a range of technology areas, from AI to genomics. India has focused on the applications and services space in general.
 

Building AI capabilities — especially at the cutting edge such as Generative AI models — will not be easy for India because it will start out late. In the US (and Canada), University research institutes, as well as technology companies such as OpenAI, Google, Meta, Microsoft, Anthropic and others, have devoted significant resources to cutting-edge AI research. In China, Baidu and SenseTime as well as several startups have built their own Generative AI models and the Chinese government has prioritised AI research for a long time. European countries have lagged behind the US and China, but dozens of firms and research institutions in France, Germany and other European nations are building their own transformer and other AI models that can be viable rivals to those built in the US. In South Korea, internet giant Never is getting into the game.
 

To play catch-up, India will need to have a clear plan and also execute it properly. One option, as Krishna suggested, is for the Indian government to take the lead and set up a national AI computing centre. But there is another model as well — as the original charter of OpenAI showed. OpenAI came about because some of the most influential billionaires (Elon Musk, Peter Theil) and other tech luminaries (Sam Altman, Greg Brockman et al) banded together to set up an AI research organisation that would democratise AI programs. Infosys, then under CEO Vishal Sikka, was one of the founding investors. Even though OpenAI now has a for-profit arm that reserves the latest AI models and tools for its clients like Microsoft, there is a healthy open-source movement that has developed in Generative AI. The big IT firms in India could collectively fund an AI research institute the way OpenAI was built. It would mean putting aside market rivalries in order to focus on a long-term common good.
 

Even if the latter model is followed, the government needs to create policies that can bring to India the best Indian brains currently doing research in the US or Europe in the area of AI research. China’s formidable AI research programme has been built by encouraging the brightest Chinese engineers and AI researchers studying in US or European Universities or working for Silicon Valley giants to come back to the mainland. The Indian government needs to figure out what is necessary to woo the best Indian AI talent back from Silicon Valley and US universities. The government will also need to think about new laws in copyright and data usage that will give the first rights of non-personal data generated within the country to its own players. India is after all the second-biggest generator of digital data in the world, next only to China.
 

It will not be an easy task — but building national AI capabilities is imperative if India does not want to remain a techno-colony.

The writer is former editor of Business Today and Businessworld, and founder of Prosaic View, an editorial consultancy