IBM CEO and managing director Arvind Krishna made a very pertinent point this week when he said that the government should look at developing sovereign capability in artificial intelligence (AI) and consider setting up a national AI computing centre. While there have been a lot of talk on the capability of AI and the need to regulate it, the topic of the government developing capabilities in terms of a computing centre has been highlighted perhaps for the first time. The need for government’s participation in AI is mandatory if data-sets and large language models are to be developed for larger, macro purposes. Left on its own, industry will only develop language models for AI for its business needs. This is fine, but their applications will be limited.
AI language models are basically predictive in nature and to develop such models, large sets of data are required. Using the right language models, the government will have the tools to predict, for instance, the nature of monsoons, the health requirement of the citizens over the years, and even the manner in which the economy and the national income would grow. These are not the areas where a private company developing language models would invest. Simply put, the government’s participation in creating AI capability centre will give it the right kind of AI tools for policy making.
For developing AI language models and computing capability, server farms are the basic hardware that are needed. These require substantial investments. To quote Krishna, such an initiative may require a few hundreds of millions of dollars, which is within the government’s capacity. The server farms need to be housed in big buildings for which only the government will be able to acquire land in a proper location. Having developed such infrastructure, the government can easily involve private companies to develop large language models for its needs. The benefit for industry will be that it will be able to leverage the data available with the government for its own purposes through such participation.
This was clearly indicated by Tata Sons chairman, N Chandrasekharan at the recent B20 summit, where he pointed out that the new data protection law has created an architecture for making private applications on the internet using public data, while maintaining complete privacy. Since the government has created the infrastructure for securing personal data, the timing is perfect for a public-private partnership to use such data for developing AI language models. The funding part can be handled through various innovative methods the government has recently devised.
The production-linked incentive kind of schemes or the subsidy model offered to companies for design and manufacturing of semiconductors are some of the ways which can be explored and improvised to create such sovereign capabilities. Since the government is in the process of finalising the draft Digital India Act, which will be shortly put up for public consultation, one hopes the issue of developing AI capabilities and the role of the government and industry will be enunciated in it. The Modi government is seen to be quite technology savvy and has been using such tools for better delivery of schemes aimed at the socially disadvantaged. Hopefully, the right initiatives will be taken on AI as well to ensure a constructive government-industry participation in an area which is still nascent but has the potential to grow at a fast rate.