News Just Five ChatGPT Queries Can Use 16oz of Water, Say Researchers

[Geef]

stillsuits would need to have some kind of battery-powered airconditioner to keep their occupants from boiling.

I remember seeing in one of the 'Making of' videos about the original Dune that some of the characters almost passed out while wearing them and they could only wear them so long.

---

There is another character who ran around in a rubber suit all day long. NSFW, rated R batman.

 

[bit_user]

I remember seeing in one of the 'Making of' videos about the original Dune that some of the characters almost passed out while wearing them and they could only wear them so long.

Perhaps that's because they were neither real Fremen nor wearing real Stillsuits.
; )

Now, where'd I put my nice cup of spice coffee?

 

[NatalieEGH]

Data centers are causing environmental stress not just due to power demands, but water usage too, with AI services like ChatGPT ramping up consumption.

Just Five ChatGPT Queries Can Use 16oz of Water, Say Researchers : Read more

Please provide data in terms other than 500ml of water? Calories, kilocalories, joules, ergs, ... would provide actual data. This information has no meaning as presented except to shock and amaze the uneducated. This statement could have multiple meanings.

For the less scientific:
1 ml = a volume of a cube 1cm x 1cm x 1cm.
A calorie is the amount of heat required to raise the temperature of 1ml of water by 1 degree Kelvin. (I am most comfortable in calories as that is the units I used in college and later in research.)
1ml of water at 273.55 degrees Kelvin has a mass of 1 gram, for most uses this is a good approximation of the mass of water in a liquid state.

Are you stating 5 queries will raise the temperature of 500 ml of water:
1. 1 degree Kelvin or 500 calories?
2. from room temperature or about 300 degrees Kelvin to boiling at about 373.15 degrees Kelvin or about 37Kcalories?
3. from 300 degrees Kelvin to where water will dissociate 3270 degrees Kelvin or about 15,000K calories?
4. sufficiently to actually convert the entire mass 500 grams to energy or as much as a 10.75M ton nuclear blast? You do indicate the water is destroyed.

Are you suggesting in the cooling process, the water is becoming polluted to the point is must be cleaned before it can be used again and therefore cannot be released as steam because of all the contaminants that would be released along side?

While I generally find the articles on Tom's Hardware informative and useful, I find this article a waste of time on the part of the author and on the readers. If the author has information he/she feels the audience needs to know, please rewrite the article in a more meaningful manner.

 

[NatalieEGH]

Did it ever occur to you that a closed-loop water cooling system requires additional energy to remove heat from the water (or re-condense it, if it's steam)? That involves additional energy (not to mention equipment costs). So, it's essentially a water vs. energy tradeoff. Take your pick.

And no, you can't get that energy from your turbine - the steam in your imaginary system would be far too low-pressure to convert to an appreciable amount of kinetic energy.

May I suggest and in-ground system for radiating the heat.. The lack of actual numbers provides little to no information on the amount of heat generated or the size of the heat pump system to store/transfer to the earth that much heat (I hesitate in using the term geothermal as that is usually reserved for extracting heat from the earth). If such a system were instituted and there were valves to shunt the heat in different paths, perhaps in the Winter, pipes could be routed fairly close to the surface to keep the parking lot cleared, keep the ground warm enough for plants to grow and bloom across the Winter, ... I do not suggest for building heating for reason of possible danger if too high of pressures happened.

 

[bit_user]

Please provide data in terms other than 500ml of water?

If it's 100 ml of water that's being consumed per query, then why is that an inappropriate metric? It would directly characterize the environmental impact, which is the point.

For more detail, you should look up the paper they're referencing. Unfortunately, this article is referencing it indirectly, via an article by the AP, so a bit of searching will probably be necessary.

Are you stating 5 queries will raise the temperature of 500 ml of water:
1. 1 degree Kelvin or 500 calories?
2. from room temperature or about 300 degrees Kelvin to boiling at about 373.15 degrees Kelvin or about 37Kcalories?
3. from 300 degrees Kelvin to where water will dissociate 3270 degrees Kelvin or about 15,000K calories?
4. sufficiently to actually convert the entire mass 500 grams to energy or as much as a 10.75M ton nuclear blast? You do indicate the water is destroyed.

Obviously, it's being raised more than 1 degree K. Furthermore, you don't need to raise water to the boiling point, in order to evaporate it.

Are you suggesting in the cooling process, the water is becoming polluted to the point is must be cleaned before it can be used again and therefore cannot be released as steam because of all the contaminants that would be released along side?

No, just that it's being removed from the watershed from where it was taken.

While I generally find the articles on Tom's Hardware informative and useful, I find this article a waste of time on the part of the author and on the readers. If the author has information he/she feels the audience needs to know, please rewrite the article in a more meaningful manner.

While you raise some interesting questions about how many BTU they're dissipating and the efficiency of evaporative cooling, I think your line of questioning goes well beyond the realm of common sense. The article's headline figure (if accurate) is directly relevant to the environmental impact of ChatGPT-class search queries - no further analysis is actually needed. In other words, while I respect your inquisitiveness, you're overthinking it.

Furthermore, I don't recall Mark Tyson ever replying to forum comments. If you want answers, you're probably going to have to search out the paper (and then feel free to share with us what you found).

Do note that since the figure allegedly includes water losses during power generation, we won't be able to link it back directly to datacenter cooling. However, perhaps the paper contains a breakdown, in which case I'll be interested to know the stats on the latter.

 

[bit_user]

May I suggest and in-ground system for radiating the heat.. The lack of actual numbers provides little to no information on the amount of heat generated or the size of the heat pump system to store/transfer to the earth that much heat

We'd need more information to evaluate the potential of such a solution. However, it would certainly be much less efficient than evaporative cooling, in which case the point about water vs. energy/infrastructure tradeoff applies. Such a system would have to be designed so that the steady-state thermal gradient remains sufficient to continue dissipating all of the generated heat.

If such a system were instituted and there were valves to shunt the heat in different paths, perhaps in the Winter, pipes could be routed fairly close to the surface to keep the parking lot cleared, keep the ground warm enough for plants to grow and bloom across the Winter, ... I do not suggest for building heating for reason of possible danger if too high of pressures happened.

People have utilized datacenter waste heat for industrial scale eel farming, among other things:

Japanese Data Center Starts Eel Farming Side Hustle

Where there's an eel, there's a way.

www.tomshardware.com

Sadly, I think there's more heat to dissipate than opportunities to harness it productively.