Modern AI models like ChatGPT operate based on large language models and are trained using huge amounts of data. This process requires high-performance processors and is very energy-intensive.
Category
🗞
NewsTranscript
00:00Using AI tools like ChatGPT can be a time saver, but they're a huge drain on resources.
00:08Apart from electricity, these systems require water and metals.
00:14Training AI models requires powerful processors, mostly GPUs or Graphic Processing Units.
00:20These are installed in servers and consume a lot of electricity.
00:25The amount of compute power we need for things like ChatGPT 4.0 is beyond good and evil,
00:33and beyond anything that can be consciously considered an environmentally sustainable
00:39part.
00:40AI systems take weeks to months to train on large computer clusters with thousands of
00:48GPUs, and these compute systems consume electricity resources that are far larger than any small
00:59city or even large cities.
01:01The International Energy Agency predicts that, by 2026, the amount of electricity needed
01:07to power data centers, artificial intelligence, and cryptocurrencies could be twice as much
01:12as in 2022.
01:14Already today, the amount of electricity used by data centers in some countries is huge.
01:23In Ireland, for example, right now, 18% of the power of the entire country goes to powering
01:28data centers, and the International Energy Agency estimates that could be as much of
01:32a third of the country's entire power by 2026.
01:36So these are really significant numbers, and these issues are really growing, not just
01:40in Europe, but around the world as well.