Global AI race should not create a risk to humanity

The global competition in artificial intelligence should not lead to a ‘race to the bottom’ by overlooking AI governance or risks to humanity, said Amandeep Singh Gill, UN secretary-general’s envoy on technology. He said India, with its technological prowess, has the potential to emerge as the leader of the Global South in digital transformation. Gill, who is at the forefront of international cooperation on AI governance, spoke to ET’s Annapurna Roy & Himanshi Lohchab during his ongoing India visit. Edited excerpts:

India’s role in AI governance

India is at the cusp of the next phase in its digital transformation, leveraging data and AI, leveraging the foundations of DPI (digital public infrastructure) for accelerating growth in the digital economy. India is in the middle of these interesting debates on what direction to take. The Digital Personal Data Protection Act is an interesting and important development..


How can India contribute to the digital transformation efforts of the rest of the world, out of a sense of solidarity as a leader within the non-aligned movement, as a leader within the Global South? How can India help others, not in the sense of providing charity or assistance, but in the sense of coming together for pooling of resources, for sharing of applications and the benefits that flow from these exciting technologies and also sharing the governance experience? India has a lot to share with the rest of the world.

On equitable distribution of AI benefits

That’s a major concern that AI may end up reinforcing inequalities and AI’s fruits, AI’s benefits, particularly for sustainable development, would not be shared equitably. There is an existing digital divide. You don’t want an even bigger divide on the AI side.
The UN provides a space where all 193 member states can engage in shaping the AI future. On capacity-building on AI, how can benefits be shared, access to critical AI enablers, compute—GPUs are in short supply—how can data and talent be facilitated through a network of collaborating centres through, potentially, a global fund on AI where public and private resources can be pooled. I think this is a significant issue for developing countries, because we need to widen the innovation base.

On managing risky systems

In areas such as lethal autonomous weapons systems, it’s important that the competition is managed in the light of existing obligations under international humanitarian law and other agreements, because there are risks for humanity. Even this kind of exciting opportunity around AI could be undermined if we start to have a destructive competition that goes into spaces which are risky and hard to predict—systems in which it’s difficult to establish human control, human accountability, in particular for life and death decisions, systems that can have significant societal implications in terms of how information is presented—you can call it the misinformation or disinformation problem, but it’s essentially about distortion and manipulation of reality. There are advances in generative AI that have heightened the risk systems that can be misused by non-state actors to develop weapons for terror, whether bioweapons or cyber weapons. There has to be a modicum of collaboration and cooperation to manage those risks.

On AI’s impact on jobs

This requires urgent attention from policymakers. GenAI is a labour-saving, productivity-boosting technology. Fundamental relationship between labour and capital is going to get impacted. In a diverse, populous, democratic country like India, this needs careful nurturing and we can’t just sit back and say, ‘look, you know, the market will take care of this’. For instance, leading BPO companies are preparing for work shifts from low-end work on the call centres to how can we help the development of AI model training, fine-tuning?

On need for an AI law

India will decide for itself. I see the laws are already there, like the DPDP Act, the IT Act and the Right to Privacy judgment. You can’t separate AI governance from data governance. But I think at some stage India may need to look at some specific aspects of AI regulation. It’s not enough to say that only domain-specific regulation is sufficient.