The regulation of AI should take its climate impact into account

AI is so costly for the planet to deploy that a cost-benefit analysis ought to guide its rules of usage

is an independent expert on climate change and clean energy.

India is undoubtedly a leading country in the deployment of artificial intelligence (AI) systems. A recent survey on cloud complexity by NetApp, a data infrastructure company, revealed that 70% of firms in India have AI projects up and running, as against 53% in the US, and that 91% of the companies in India plan to use their data to train AI models, as against a global average of 62%.

Globally, the AI industry is expected to grow by 37% annually till 2030, with great excitement over the opportunities it creates as also trepidation on its misuse. Regions/countries like the EU, US and China are expected to implement sweeping regulatory measures to bring about greater transparency, establish new standards and address activities that pose a systemic risk to citizens, with the EU already working on a bill to institute an “AI liability directive” that would provide financial compensation to adversely effected people. India too is cognizant of the need to develop a regulatory regime for AI, although it is likely to be light-handed, at least initially, with potential job losses its special focus.

It is important, however, to recognize the two-way relationship between AI and the energy sector. The electricity grids of today are only going to grow in complexity, with multi-dimensional flows of electricity now integrating all kinds of distributed and decentralized systems, including rooftop solar panels, electric vehicle power-packs, and so on. Combine this with the complexity that accompanies the deployment of hybrid systems such as wind-solar or offshore wind-solar-tidal power, or other combinations thereof. Planning and operating such systems, in particular the vexed challenge of demand and supply forecasting, will greatly benefit from the use of AI systems. In fact, AI will likely be indispensable for sustaining the green energy grid of the future. While AI is already helping with the predictive maintenance of current energy systems, it is easy to imagine it playing a vital role in the maintenance of emerging systems that will be vastly more complex.

On the flip side, the energy guzzling nature of the AI industry is assuming alarming proportions. A search driven by AI could consume four to five times the energy needed for a conventional web search. It is estimated that the computational power required to sustain the rise of AI is doubling roughly every 100 days. To achieve a 10-fold improvement in AI model efficiency, demand for computational power could surge by up to 10,000 times. In Davos, Switzerland, OpenAI’s CEO Sam Altman sounded a warning on the vastly higher power demand of the next wave of generative AI systems that energy systems will have to cope with. The water demands of the AI industry to cool data processors could be enormous too.

Research on other environmental consequences of AI is still in a nascent stage. However, staying with energy for the moment, there is no doubt that the world is off-track in meeting its sustainability goals. India’s renewable energy programme, while impressive, has thus far only helped make a dent in incremental electricity demand and therefore incremental emissions; it has not yet started taking emissions out of the country’s base of energy demand. Unchecked deployment of AI could increase incremental electricity demand enormously, making the task of reducing the growth of greenhouse gas emissions and eventually reversing them that much more challenging.

In other words, India and the world could be locking themselves further into fossil energy use for much longer. And, while we have little evidence on other environmental consequences of AI, the adverse effects of fossil energy use on the planet is well established and may be proportionally attributable to AI in the future.

To remind ourselves, scientists have identified nine planetary boundaries, staying within which would provide a safe operating space for humanity. We have crossed six of these boundaries, capturing climate change, fresh water use, air pollution, biosphere integrity and plastics pollution—mostly attributable to our use of fossil fuels.

The relationship between AI and energy becomes even more consequential in the context of climate concerns driving net-zero ambitions around the world, the emerging jurisprudence around recognizing our right to escape the adverse impacts of climate change as a human right, and the rise of liability directives in the context of AI regulation. The Supreme Court of India said in a recent judgement that climate change impacts the constitutional guarantee of the Right to Life, and that citizens have a right to freedom from the adverse effects of our climate emergency.

Not only does the AI industry need to take responsibility for its climate impact, every user industry too, by virtue of its Scope 3 emission responsibilities, must weigh the costs and benefits of AI deployment carefully. We cannot afford to allow its use for frivolous ends. Similarly, the regulatory regime needs to recognize the high cost of AI application—in terms of investments, climate effects and other environmental factors—and provide clear AI-usage guidance and incentives to contain costs. India and Indian industry should work towards strong, precautionary and enabling regulatory regime for AI. The sooner, the better.