They don't want you to know

preview_player
Показать описание
Artificial Intelligence is here. AI is everywhere, and Chat-GPT is being used often by people across the globe. It may seem like a simple act, but the environmental implications of using Chat-GPT, especially the water consumption is surprising. There is a lot that people don't want you to know about the environmental costs of AI so today we are going to explain them.

Thanks to Professors Ian Clarke and Ayumi Goto at OCAD university for helping to inspire this episode. And thank you to Emmy and Aila for working on this with me.

Written by: Gregory Brown
Edited by: Luka Sarlija
Drawn by: Gregory Brown

References and Further Reading:

Рекомендации по теме
Комментарии
Автор

To clarify for a lot of people, the datacenters don't drink the water. It's used for evaporative cooling (like a swamp cooler). The cool fresh water comes in, runs through systems containing circulating water carrying the heat from the datacenter. The water evaporates when it runs over these pipes/radiators and cools the pipes, while the water vapor is vented to the atmosphere (it appears as clouds on cold days).
The contention is due to the datacenters basically pumping in drinking water and boiling it away. Of course it comes back down as part of the water cycle, but that makes it unavailable for use locally where it was extracted.

This is compounded by their location. For evaporative cooling, starting with dry outdoor air is best because the water can remove more heat for a given volume. Coincidentally, the driest air is usually in water-stressed regions.

jaymzx
Автор

so this is how the water wars will start huh

FlyBoyKhi
Автор

Unfortunately what companies should do and will do are completely different things. Thanks for raising awareness about this!

brken
Автор

Thank you for this video. I feel like most popular science communicators haven't been talking about this but this is what a lot of us have been concerned about for a long time. I'm doing a master's in Data Science and I'm really excited about ML applications for science research, and there are other things it's really useful for, but the way AI is being overused weighed against the cost is really worrying to me. I also just think we don't need to shove AI into everything but that's a slightly different conversation.

Ayshafr
Автор

The small data-center I work for also uses water for cooling. But it has a closed circuit with ultra pure water to run through tiny channels inside the computers. Once the water is hot, it gets pumped to the roof where ambient air gets pushed through huge heat-exchangers (think car radiator), which turns the ultra-pure, hot water into usable, "cool" water again. Cool is relative here. Our computers use 40C "cool" water and exhaust it at 60C. That can of course be regenerated by ambient air.
The water we use is to spray down the radiators on the roof if the ambient air cannot whisk away enough heat. The water evaporates and is basically "gone".

That gives us an ultra high efficiency, because we do not need air-conditioners. Those take enormous amounts of energy and "roof-cooling" is basically free. We only use 4% of additional power to what the computers (100%) use. (PUE of 1.04) But we evaporate water for 1200 people each year.

ChrisBigBad
Автор

Mad Max but the antagonist is just a CEO hoarding water.

Piratewaffle
Автор

I guess less people are thinking about this than I thought. It was the foremost concern I've had since AI started to pick up. Between AI and data farming in general for use in selling people's information, data harvesting has been very concerning to me environmentally. Not just the water cost, but the cost of precious minerals to make the components these technologies need to function.

Lerrae
Автор

My question is: why wasn't there a limit on how much water a company can use to begin with?

annemabrie
Автор

I wonder how much water would a datacenter consume compared to say a meat farm - Would've been good to include a comparison between datacenters and other industries

LGtheCat
Автор

I agree, restarting or building more nuclear reactors is a good idea, actually... creating more AI data centers not so much.

somestuff
Автор

We have a stagnant desalinization plant built forever ago here in Victoria Australia, perhaps we should start using it for this kind of purpose and get the recycled water to be used for farming and agriculture and also possibly reduce the rising sea levels at the same time. Win, win, except for the fact that once governments are agreeable to the idea it will just allow big corporations to get more money and tax breaks without any benefits to the community.

jasonellis
Автор

0:14 just give us the volume. That 1/8th means nothing to most people who don't know that lake

JohnMushitu
Автор

Why not just make a gigantic freezer? Or relocate to a naturally cold environment.

NitroRonin
Автор

money giants ain’t gonna use this AI in moderation, or leave it to a proper ethical use.

itsjestjay
Автор

0:12 Just so you know, that's 25, 500, 000, 000, 000 litres of water... (67363873351323 gallons)

ahanabasu
Автор

I'm okay with AI helping sciences like medication or helping people who need it. For example, a reader who reads text is allowed to read for people who can't see or see well or have other related problems.
But AI that makes art by regurgitating stolen works, be it an image or music or writing texts, we can leave out. They don't deserve our water being used on them.

XxXNightcoreQueenXxXV
Автор

Water doesn't disappear anywhere, it's infinitely reusable.

HungrysitesRu
Автор

Fresh water used for cooling data centers is typically managed through one of the following methods:

1. Evaporation: In many data centers, especially those using evaporative cooling, water evaporates into the atmosphere, removing heat in the process. This reduces water volume and leaves behind concentrated minerals, which are often discharged as wastewater.


2. Recycling: Some advanced facilities recycle water by filtering and reusing it multiple times within closed-loop systems to minimize waste.


3. Discharge: Excess water or water containing concentrated minerals may be discharged into municipal wastewater systems or other approved disposal systems, following environmental regulations.


4. Alternative Use: In rare cases, discharged water can be repurposed for agricultural irrigation or other industrial uses, provided it meets quality standards.



Sustainable data centers aim to reduce freshwater dependency by using alternatives like seawater, treated wastewater, or air-based cooling systems.

Audaxocles
Автор

Of all the ads that could have popped up on this video - a masterclass on how to use AI for anything and everything, from work to your to-do and shopping lists 🤦‍♀️ i have never knowingly used AI, and will never use it if i can avoid it!

TheAueRora
Автор

The Three-Mile Island accident occurred in the Spring of 1979, not 1978... I know because I had planned to drive back from college in Denver to the Philadelphia area (where I grew up) when the accident occurred. Because of that accident, I had to push back my plans for a month... Thank goodness my landlord allowed me to stay for that extra month, with incredibly short notice!

DonnaMSchmid