Artificial intelligence systems are creating vast emissions – and it is getting worse, according to a major new study.
The increasing energy required to train and run more complex models, as well as the much broader interest in using them, is bringing serious environmental consequences, a new paper has warned.
As the systems get better, they require more computing power and therefore more energy to run. OpenAI’s current GPT-4, for instance, uses 12 times more energy than its predecessor.
What’s more, training the systems is a tiny part of their work. The energy used to actually run the AI tools is estimated at 960 times those from a single training run.
The researchers suggest that the impact of those emissions could be vast. AI-related emissions could cost the industry more than $10 billion each year, the report suggests, and it calls on governments and regulators to standardise ways of measuring those emissions as well as new rules to ensure they are kept to a limit.
“The exponential growth in AI capabilities mirrors a concerning rise in its environmental impact,” said Meng Zhang, lead researcher from Zhejiang University.
This study underscores the urgent need for the AI industry to adopt greener practices and sustainable standards. Our goal is to equip policymakers with the data needed to address AI’s carbon footprint through proactive regulations.”
The findings are reported in a new paper, ‘Revisit the environmental impact of artificial intelligence: the overlooked carbon emission source?’, published in Frontiers of Environmental Science & Engineering.
