It is a strange coincidence(巧合) that as humanity attempts to greatly reduce its carbon emissions(排放), it is also rushing to develop a technology that could, in theory, consume an unlimited amount of energy.
Doing things against facts is a dangerous game, but you can picture a world in which, having chosen to start dealing with climate change properly in the 1990s, we would be just wrapping up the gentle path to net-zero emissions in time for a rapid AI increase fuelled by green power. Instead, we find ourselves at risk of running a 21st-century technology on a 20th-century energy supply.
If you live in the US, every time you use an AI model, around 20 per cent of the electricity required will be produced by burning coal. The explosive growth of AI makes this an even more pressing concern. According to one analysis, if Google chose to shift to an entirely AI-powered search business, its electricity consumption could match that of some countries.
Such estimates may be magnified, but evidence of rising power consumption from AI is all around. Microsoft is placing bets on nuclear plants to power its data centres, while the English government has promised to boost its national computing capacity by 50 per cent by 2025, which would help it keep pace with rapid AI developments in the world.
Thankfully, we may soon reach a turning point. As is reported, this year the global power department came close to reaching peak greenhouse gas emissions, as the switch from fossil(化石) fuels to clean and renewable sources is well underway. The question now is whether we can quicken the pace of decarbonisation (脱碳) to match the coming growth in energy consumption as AI becomes increasingly rooted in our society. Silicon Valley's intelligent machines may grab the headlines and the imagination, but the people really inventing the future are working in the energy department.
