The contents of this essay are based on conversations during the 7th India-US Forum
AI is a nascent industry which must be leveraged without running into overwhelming legal challenges.
There is a North-South divide in approaching AI. Countries in the global North, such as the US, have taken a risk-centric approach to creating rules around AI. On the other hand, the AI community in India, and much of the Global South, has approached this technology as an opportunity to extend empowerment.
There is a need to reconcile these two approaches, to not discard important risk considerations while allowing space to leverage AI for public good.
China is currently behind the West in developing AI models. However, it continues to be a serious competitor. Chinese mobilization of the economic and political system, along with their Civil-Military fusion, will allow them to catch up fast, even if the Chinese models look different to the current Western ones.
Chinese investments across the board in critical technologies make them a serious challenge in the coming 5-10-year period. From 2025 to 2030, most of the world’s data can go through Chinese-made infrastructure.
The India-US cooperation lacks a sense of urgency. There is a pressing need for a US-India strategic technology partnership. Over the next five years, the India-US partnership will have to focus on exploring the space realm for opportunities to test and deploy AI models, bringing investments in hardware to relieve current hardware supply chain dependencies in chips, drones, and robotics, and ensuring cooperation in the Lethal Autonomous Weapon Systems (LAWS) space.
India and the US converge as partners in democratic principles that neither can risk while weighing the opportunities of AI. The AI systems that this partnership builds can benefit the needs of the Global South through its democratic processes. Democratic processes, however, are slow which gives non-democratic AI models an edge in mobilizing resources and catching up.
The two countries can also leverage the China challenge to gain in talent, software, and educational opportunities. They can also provide alternatives to Chinese models and set a framework for AI in the coming decade.
In the last 10 years, China has built clear strategies, amassed resources, and identified national AI champions that they can deploy globally. The release of their AI models in 2023 indicated their transformative potential in the economic, societal, and national security space.
The US established the National Security Commission on Artificial Intelligence in 2018 to combat Chinese competition in the AI space. In recent years, the US has taken further steps in this regard. The White House passed an executive order to move departments and agencies to scale up the adoption of generative models and build the regulatory framework around the use of these operations. The latter was supported by a bipartisan Congressional push on providing a regulatory framework.
There are complaints in the United States that AI models are becoming too expensive, which will impact access of small- and medium-sized companies. The US is testing a National AI Research Cloud, which is a consortium of universities with tools access to Cloud. Legislation such as the Create AI Act of 2023 has been brought to fund this initiative.
But the current moment is not just a geopolitical contest between the US and China, but a transformative moment in the way democracies are organizing themselves around platforms that are emerging from AI.
The current generative models may have come from US-based companies, but the release of open-source models can be leveraged by emerging nations to invest resources and get ahead in the geopolitical space.
There are three elements to AI: data, algorithms, and computational power. India has AI-skilled manpower and large datasets but lacks compute capabilities, which are concentrated in select companies and countries.
80 per cent of the cost of AI innovation is in compute capabilities. The compute required for the newer models is doubling every 6 to 10 months and the space is witnessing a massive concentration of capabilities. Through the iCET program, India and the US have been working together to ensure that the Indian government and private actors have access to chips that are being developed across the world, which are making AI compute easier and faster.
The focus of the Indian government will now be to democratize access to compute to leverage the Indian data mine, ensure skilling and reskilling, and build a proper data-governance framework.
Generative AI will be uniquely leveraged in India, and the country needs to play with the full stack of AI to develop local capabilities. There are multiple private sector and start-up efforts that are building models which offer unique Indian solutions to challenges that are unique to India. This speaks to the potential of Indian large language models and attracts equity investors. The challenge is to bring these models to the mass of the people and to build solutions in Indian languages.
India will need to develop its stack with investments that enable frugal innovation and business models that will allow newer startups to invest in developing compute infrastructure.
India’s Semiconductor Mission has provided incentives for chip manufacturing. Competition in compute capacity in India should drive down prices over time. It will require significantly increased investment from the government and private actors to compete with the best chip-manufacturing companies in the world. India’s aim currently is to produce 28 nm chips.
The current data piles for training AI models have been exhausted, and there is a need for better data to train these models. The data generated by automated vehicles, for instance, can be very useful for AI model training. China stays ahead by deploying new vehicle models, since they control the entire spectrum of the supply chain.
India is not operating at the scale of China and requires building alternatives for full stacks to the current dominant players in the market. While India has a large data exhaust, the data to train large language models in Indian languages is very small. India must get creative in generating this data and create the right regulatory framework to support this. These research clusters can also be made available to friendly countries in a spirit of cooperation similar to that between the US and India.
Investments in Indian language models are already being made. The advantages of these investments can be distributed by creating interoperable frameworks and open data sets such as the US AI Research Cloud.
Trusted partners, such as India and the US, can also benefit from extending these models across borders and plugging into each other’s AI research data sets.
AI will take jobs with the 4D Rule, that is, Duplicate, Delicate, Dirty, and Dangerous jobs can be replaced by AI. Both in India and the US, there will be a co-piloting phase with AI models as early adopters take up AI, followed by a massive wave in AI adoption. This will create a need for mass skilling of populations in order to keep them competitive in the market. Thus, the jobs that AI takes will be replaced with jobs that require skilled AI talent.
Personalized education based on individual needs will become possible and cost-effective. The education and healthcare sectors will be provided with a range of new toolsets with AI.
