Water consumption in data centers — no small issue - video Dailymotion

DW (English)

Training and using AI tools like ChatGPT is energy-intensive and heats up processors. Fresh water is often used for cooling. If that warmer water is then recycled to its original source, it impacts the ecosystem.

Transcript

00:00AI tools like ChatGPT are oftentimes based on large language models, which are trained

00:06using lots of data.

00:08That way, they can generate an answer to almost any question.

00:12But that requires a lot of energy.

00:14Every new request triggers computations that are far more complex than what happens with

00:18a traditional search engine.

00:21When you do a search with Google or any other classic search engine, what happens is a simple

00:26data query.

00:28That's it.

00:29With generative AI, there's a huge model behind it, and for each request, the entire

00:33model is activated.

00:36Graphics processors tend to heat up while they're working.

00:39If you search for boiling water on a CPU, you'll find many videos like this.

00:46To prevent chips from getting damaged, they're cooled down by using fans, which also use

00:51electricity, or by using water.

00:53This is also harmful to the environment, because often, fresh water from the local supply is

00:58used.

00:59If the water evaporates during the cooling process, then it's lost.

01:04If it gets partially redirected back into the rivers or water sources, then it can be

01:08used again.

01:10It will just be warmer, which means it will change the ecosystem of the river or water

01:14source.

Water consumption in data centers — no small issue

Category

Transcript

Be the first to comment

Recommended