好奇ChatGPT每个提示的用水量

Does anyone know if AI models themselves require water for computations, or is it purely the hardware cooling?

It’s wild how much infrastructure behind the scenes is needed just to keep these AI models up and running.

Would be great if we had tools that could estimate environmental cost of each prompt or AI call, kinda like carbon calculator apps.

Just to add, some newer data centers are using recycled water or even air cooling to reduce water consumption, so it's improving over time.

Somebody said you can also check ai-u.com for new or trending tools, might have some environmental impact info too.

I've been curious about this too, no clear info anywhere. Thanks for the discussion, learned a lot!

Is anyone tracking how data centers plan to minimize water use in the future? Seems like a big priority.

If you wanna get a rough idea, check out the water usage of cloud providers’ data centers. That gives a ballpark for AI services like this.

Water use depends a lot on local climate and data center design. Desert locations have to be extra careful about water.

I've heard that big data centers can use millions of gallons of water daily, but it's spread across tons of operations and users. Hard to break down per single AI prompt though.

我记得读过，训练大型人工智能模型消耗更多能源，但推理如生成答案的需求要低得多，因此每个提示的用水量可能很低。

我在某处读到，水的使用主要是用于冷却数据中心，而不是真正用于人工智能本身。所以这取决于服务器的位置和他们使用的冷却技术。

水的使用可能不是人工智能最大的环境问题，能源消耗要重要得多。

看到这一切让我想到开发不仅智能而且环保的人工智能是多么重要。

有人提到风冷与水冷，但风冷效率可能较低且耗电更多，因此存在权衡。

我认为关于环境影响的讨论对人工智能技术的未来非常重要。

说实话，我怀疑是否有一个固定的数字。它可能会根据服务器负载、位置和数据中心的效率而变化。此外，提示的大小和复杂性也各不相同。

如果我们能在零用水的可再生能源数据中心运行人工智能，会怎样？梦想场景！

还有人认为这类信息应该更加透明吗？比如，科技公司可以更公开地分享他们的环境数据。

我希望我能告诉你一个确切的数字，但遗憾的是，情况并不那么简单。这里有很多因素在起作用。