apparently, when water is used for purposes that annoying people dislike for ideological reasons, it disappears into another dimension, lost forever.
'Thirsty' ChatGPT uses four times more water than previously thought
61–70 of 105 posts
Re: 'Thirsty' ChatGPT uses four times more water than previously thought
#62From the article: > Using the chatbot for between ten to 50 queries consumes about two litres of water, according to experts from the University of California, Riverside. From ChatGPT: > A single BTC transaction consumes approximately 1,000–1,500 kWh of electricity, with cooling processes for fossil fuel and nuclear power plants using about 2–3 liters of water per kWh. This translates to an indirect water consumption…
> If accurate, I still prefer ChatGPT.
It looks like you are using ChatGPT too much and feeding yourself untrustworthy information for no reason.
Internet search still exists, there is no need to use ChatGPT for this. Simply typing "how much water per btc transaction" into Google will give you this top response much faster than ChatGPT would:
> Buying or selling bitcoin uses 16,000 litres of clean water for every single transaction.
This is reported by New Scientist, who's reporting is based on actual calculations (using real data) from a PHD candidate at the Amsterdam School of Business and Economics.
Whereas the number given by ChatGTP is based on nothing but the statistical probability of what would sound correct in this context, and has no guarantee of accuracy at all. In this case it was off by an order of magnitude compared to the number based on real data.
The internet search probably used a lot less water as well (beside being both faster and more trustworthy).
Re: 'Thirsty' ChatGPT uses four times more water than previously thought
#63> Many data centres use water-based systems to cool the plant with towers evaporating the heat, like a huge perspiration system, which means that the water is lost. Someone doesn't know what "evaporation" or "lost" means.
Someone doesn't believe in global heating.
Total datacenter energy use is on the order of 1PWh/yr. Total global solar surface irradiance is on the order of 400,000PWh/yr.
The direct heat contribution is negligible to global temperature.
There is an argument that water vapour in the atmosphere is a greenhouse gas, but also an argument that clouds reflect solar energy, and that this water vapour is emitted at ground level. I don't think it's so obvious that water evaporation from datacenter cooling systems is directly either good or bad for global heating.
There certainly are negative environmental and social effects of this water usage in some places and implementations, and certainly it could include severe heating potential, but this is an over-reduction without further context.
Re: 'Thirsty' ChatGPT uses four times more water than previously thought
#64Using 40ml of water to ask chatgpt how many grains of rice 40ml of water would grow, and the answer is ~1.
Re: 'Thirsty' ChatGPT uses four times more water than previously thought
#651. ChatGPT is in the title purely for click-related reasons. This is about data center water usage, and applies to any application that uses a data center. 2. Seems like water should be more expensive in water-stressed areas and that data centers should pay a rate that includes the cost of the externality. Is it, and do they?
1. Yeah, let’s talk Microsoft Teams instead. I suspect that’s worse than AI training…
2. The article I’d write starts with incumbent uses of water. Residential use, especially in dry areas, has been pressured to progressively increase efficiency. It does take energy to treat and pump and treat water again. But some people always wonder if they’re saving water just so another part of town can add residents or sell water to a new data center that runs Microsoft Teams.
Evaporative cooling shouldn’t pay the (embedded) full price of treating sewage. For example, in California, that’s charged at a fraction. But it still covers the cost of treatment, the administration of water rights, and pumping. And that’s enough for corporations to choose to put AI training in other places.
Re: 'Thirsty' ChatGPT uses four times more water than previously thought
#66One avocado takes 320 liters, I'd rather give up on avocado than Claude.
How about T-shirts? Websites, ads and "fun fact" news articles tell me it takes on the order of 10000 liters for each item... which seems to be a bit high but I don't have the means to verify the number.
Re: 'Thirsty' ChatGPT uses four times more water than previously thought
#67Re: 'Thirsty' ChatGPT uses four times more water than previously thought
#68Re: 'Thirsty' ChatGPT uses four times more water than previously thought
#69Earlier quoted context omitted.
I really am not. I just drank 1 liter of water. ChatGPT consumed several orders of magnitude less doing much more than me. I also just heated up my meal, with meat and fries. ChatGPT didn't need that. All of that while walking around my apartment full of various "useless" stuff. ChatGPT doesn't care.
You're not counting the training of the LLM.
I am. That's amortized over millions of users and billions of invocations.
But who cares? I'd gladly pay for something to do in minutes what would've taken me hours. It doesn't need to be as efficient as my body in the same time frame.
Re: 'Thirsty' ChatGPT uses four times more water than previously thought
#70One avocado takes 320 liters, I'd rather give up on avocado than Claude.
Most people eat avocados not for their nutrition but because they are delicious. I would definitely suggest not buying and eating them if you don't find them delicious! There are much cheaper options.
Does Claude offer much beyond its utility as a tool? I mean, it's ok to like tools too. But I'm not sure I'd want to pay an ongoing cost for a tool just because I enjoy using it. It has to deliver utility too. And if it does then you shouldn't have to give up avocados, in fact you should be able to buy more!
If/when we to the Her level of AI then things will be interesting...