Power play: will AI help or harm the climate?

能量博弈:人工智能助力还是危害气候?

Babbage from The Economist

2025-04-10

38 分钟

单集简介 ...

The race to build better AI systems is having an ever-bigger impact on the environment. Data centres, which store and process the information required to train and operate AI models, currently account for around 2% of global electricity consumption. That figure will rise, as new facilities are built to accommodate ever more energy-hungry machine learning. But AI could also help reduce emissions, for example, by optimising electricity grids, detecting methane leaks or developing better battery materials. So, will AI be bad for the environment…or will it turn out to be good? Host: Alok Jha, The Economist's science and technology editor, with AI writer Alex Hern. Contributors: Sasha Luccioni of Hugging Face; Boris Gamazaychikov of Salesforce; and Ronnie Chatterji of OpenAI. Transcripts of our podcasts are available via economist.com/podcasts. Listen to what matters most, from global politics and business to science and technology—subscribe to Economist Podcasts+.
更多

单集文稿 ...

  • The Economist.

  • We're now going to go into the IT room and, as I said, it's going to be unpleasant.

  • If you're trying to think of places least suited to recording podcasts,

  • data centres would probably be right up there.

  • If you want ear protection, you can take it.

  • I recently visited a new one at CERN.

  • the world's biggest particle physics lab.

  • In the enormous building,

  • stacks and stacks of servers were flashing as they stored and processed data,

  • all in service of the thousands of scientists on site.

  • Wayne Salter, a manager at CERN's IT department, showed me around.

  • We have the cooled water which is used by then these large fan units.

  • These are the largest fan walls produced in Europe.

  • We have a potential of 2.5 megawatts of cooling for 2 megawatts of IT equipment.

  • And we blow the heated air through the fan walls,

  • through the heat exchangers to cool down to then go through the servers to cool them down.

  • Data centres similar to this one, but on a much, much larger scale, enable you to do,

  • well, Everything that you do online, from video streaming to banking, or asking ChatGPT a question.

  • Servers need energy to work, and as you heard, fans and running water to keep them cool.

  • More servers mean more energy and more of everything else to keep them running.