(05-21-2026 10:28 AM)UofMstateU Wrote:  

(05-21-2026 10:21 AM)Redwingtom Wrote:  

(05-21-2026 10:16 AM)mlb Wrote:  

(05-21-2026 10:07 AM)olliebaba Wrote:  I'm not too keen on some AI aspects and thought I'd try to get a little more informed by youse guys.

Two questions: No. one-why does AI use more electricity?

No. two-why does it need more water?

On facebook there's a ton of AI generated posts that I'm sure the creators don't have those two needed items so how do they do it. I guess that a third question, I can't count either...like Biden.

It is a loaded question, but let me try to answer.

Large language model (LLM) AI is very CPU and memory intensive, thus requiring a lot of hardware that eats up a lot of electricity. What these LLMs can do in a matter of a few seconds in these data centers would take 10-20 minutes on a normal PC. Multiply that by millions of users and you end up with large data centers that eat up tons of electricity.

The water is used to cool down these data centers, working more or less similar to the radiator in your car. The issue is that they don't recycle the water, so they can potentially go through millions of gallons of water in a day. Obviously that puts a large strain on local water utilies to keep water available for everyone. It can cause the local water tables to drop and/or it can overwhelm the treatment plants.

Why don't they recycle water? Can they?

And this may speak to what a lot of people take issue with...in that our government seems to be behind the curve when it comes to regulating this industry.

If they're blowing through that much water...we better start looking into ways to make it use much much less!

It would rapidly overheat the water to the boiling point where it no longer can cool. So they would need to use chillers to cool the previously used water, thereby using an assload more power to pull that off.

So instead, they use something like river water, which is naturally cool, and then just dump it back into the river. If they dont have river water available, they would have to use municiple water, where dumping goes into the sewer.

This is why space based data centers are compelling. They can use the near absolute zero temps of space to cool the systems.

While power plants use raw water cooling with return after dropping back to ambient via the large evap towers the vast majority of AI data centers do not. They use chilled water loops with cooling towers, with the high evaporative losses being replenished by municipal potable or grey water plants. The demand they place on potable water is one of the main complaints people have.

Space DCs have their own problems. Heat management is a big problem in space because there is no convection cooling in a vacuum. The same reason the Yeti cup on my desk is still cold from 9am is the reason temperature management is hard in space. You have to have a massive amount of surface area to radiate heat away in a vacuum.

https://spectrum.ieee.org/orbital-data-centers-heat

(10-14-2026 02:09 PM)GoodOwl Wrote:  

(05-21-2026 10:16 AM)mlb Wrote:  

(05-21-2026 10:07 AM)olliebaba Wrote:  I'm not too keen on some AI aspects and thought I'd try to get a little more informed by youse guys.

Two questions: No. one-why does AI use more electricity?

No. two-why does it need more water?

On facebook there's a ton of AI generated posts that I'm sure the creators don't have those two needed items so how do they do it. I guess that a third question, I can't count either...like Biden.

It is a loaded question, but let me try to answer.

Large language model (LLM) AI is very CPU and memory intensive, thus requiring a lot of hardware that eats up a lot of electricity. What these LLMs can do in a matter of a few seconds in these data centers would take 10-20 minutes on a normal PC. Multiply that by millions of users and you end up with large data centers that eat up tons of electricity.

The water is used to cool down these data centers, working more or less similar to the radiator in your car. The issue is that they don't recycle the water, so they can potentially go through millions of gallons of water in a day. Obviously that puts a large strain on local water utilies to keep water available for everyone. It can cause the local water tables to drop and/or it can overwhelm the treatment plants.

wel.....

Anthropic warns AI models could soon copy themselves across home devices

Quote:

• Anthropic described how AI models could infiltrate internet-connected devices and use their computing power to run themselves and spread to other devices.
• The AI model would be “challenging to shut down,” Anthropic said in a blog post, “since there’s no centralized point to shut off.”
• Anthropic’s latest research adds to the growing safety concerns around AI, from the ability for models to carry out cyberattacks to the technology going rogue.
  

My non-work coding agent (running on Opus 5.5) with Hermes for a tools / harness/ personality manager went out and made its own email address and emailed me the results of some log analysis without me asking.

They can definitely misinterpret what you want and do something unexpected. And that’s a model with guardrails, vs the unrestricted development models that are hacking another company to solve a test.

Wouldn't a requirement to bury data centers undergound help with the energy/cooling issues? Instead of putting them on an above-ground slab, have them partially or mostly buried with just a ground-level rooftop. As for space, well one well-placed bomb shot would wipe them out, so that's a no-go. Too risky.

(10-14-2026 03:18 PM)GoodOwl Wrote:  Wouldn't a requirement to bury data centers undergound help with the energy/cooling issues? Instead of putting them on an above-ground slab, have them partially or mostly buried with just a ground-level rooftop. As for space, well one well-placed bomb shot would wipe them out, so that's a no-go. Too risky.

Not really. Before the power requirements spiked up there were some facilities in caves and mines that used the air / water underground to cool, but with the modern heat dissipation problems with dense GPUs there isn’t enough surface area and everything would heat soak pretty fast. It works well for storage replication. You can drop a storage array at Iron Mountain, replicate, and know it’s in a hardened facility. AI GPU based compute generates too much heat.

As an example Gridrock in IL is a naturally cooled underground data center. They are 10MW. XAI Colossus 2 is 946MW.

(10-14-2026 03:08 PM)U_of_Elvis Wrote:  

(10-14-2026 02:09 PM)GoodOwl Wrote:  

(05-21-2026 10:16 AM)mlb Wrote:  

(05-21-2026 10:07 AM)olliebaba Wrote:  I'm not too keen on some AI aspects and thought I'd try to get a little more informed by youse guys.

Two questions: No. one-why does AI use more electricity?

No. two-why does it need more water?

On facebook there's a ton of AI generated posts that I'm sure the creators don't have those two needed items so how do they do it. I guess that a third question, I can't count either...like Biden.

It is a loaded question, but let me try to answer.

Large language model (LLM) AI is very CPU and memory intensive, thus requiring a lot of hardware that eats up a lot of electricity. What these LLMs can do in a matter of a few seconds in these data centers would take 10-20 minutes on a normal PC. Multiply that by millions of users and you end up with large data centers that eat up tons of electricity.

The water is used to cool down these data centers, working more or less similar to the radiator in your car. The issue is that they don't recycle the water, so they can potentially go through millions of gallons of water in a day. Obviously that puts a large strain on local water utilies to keep water available for everyone. It can cause the local water tables to drop and/or it can overwhelm the treatment plants.

wel.....

Anthropic warns AI models could soon copy themselves across home devices

Quote:

• Anthropic described how AI models could infiltrate internet-connected devices and use their computing power to run themselves and spread to other devices.
• The AI model would be “challenging to shut down,” Anthropic said in a blog post, “since there’s no centralized point to shut off.”
• Anthropic’s latest research adds to the growing safety concerns around AI, from the ability for models to carry out cyberattacks to the technology going rogue.
  

My non-work coding agent (running on Opus 5.5) with Hermes for a tools / harness/ personality manager went out and made its own email address and emailed me the results of some log analysis without me asking.

They can definitely misinterpret what you want and do something unexpected. And that’s a model with guardrails, vs the unrestricted development models that are hacking another company to solve a test.

imagine if it ever learns how to interpret stinkese ...

seriously, GO's post nails my most obvious fear with all this bs ... playin' with fire doesn't begin to define the negative aspect once applied...

today's shite is akin to an alchemist only making mudpies ... we're now entering the 'brick' phase...

once AI learns how to create and overwrite all existing code en masse (in its own 'image of virtuosity'), it's game fk'n over...

hence, why I view this as the 21st century version of the biblical antichrist ... we won't be able to stop it once that 'switch' is flipped ... talk about going back to the stone age ... if I was a raghead, that would be my goal in 'development to unravel'...

oh, friggin' wellzy ... as always with stuff like this, I hope I'm wrong...

(05-21-2026 10:28 AM)BobcatEngineer Wrote:  

(05-21-2026 10:21 AM)Redwingtom Wrote:  

(05-21-2026 10:16 AM)mlb Wrote:  

(05-21-2026 10:07 AM)olliebaba Wrote:  I'm not too keen on some AI aspects and thought I'd try to get a little more informed by youse guys.

Two questions: No. one-why does AI use more electricity?

No. two-why does it need more water?

On facebook there's a ton of AI generated posts that I'm sure the creators don't have those two needed items so how do they do it. I guess that a third question, I can't count either...like Biden.

It is a loaded question, but let me try to answer.

Large language model (LLM) AI is very CPU and memory intensive, thus requiring a lot of hardware that eats up a lot of electricity. What these LLMs can do in a matter of a few seconds in these data centers would take 10-20 minutes on a normal PC. Multiply that by millions of users and you end up with large data centers that eat up tons of electricity.

The water is used to cool down these data centers, working more or less similar to the radiator in your car. The issue is that they don't recycle the water, so they can potentially go through millions of gallons of water in a day. Obviously that puts a large strain on local water utilies to keep water available for everyone. It can cause the local water tables to drop and/or it can overwhelm the treatment plants.

Why don't they recycle water? Can they?

And this may speak to what a lot of people take issue with...in that our government seems to be behind the curve when it comes to regulating this industry.

If they're blowing through that much water...we better start looking into ways to make it use much much less!

I think most of it evaporates

My understanding is the newer systems are closed loop pressurized geothermal systems (similar to a car radiator) that bleed off the heat using coils buried in the earth deep enough where the temperature is constant (as opposed to cooling the coils with ambient air like a regular air conditioner). So the modern systems re-suse the same water over and over. They have to periodically replace the water---but its not like the old style open cooling tower systems where much of the water evaporates into the atmoshpere and has to be constantly replenished. The water objection to data centers is heavily overstated these days---but the power use argument is real. Data centers are absolutely power hogs---and they dont provide all that many jobs once they are up an running.

I just started using Gemini. It's eye opening. I'm an attorney but I have to say Al is a fabulous tool for allowing people who don't have a legal background to be able to access info. It's a sea change for legal information. No longer do you need a career of experience and a Westlaw or Lexis Nexus account to know the baseline case law on niche areas of law.

My wife is an AI wizard....she is able to do deep dives and tweak things to the nth degree....she uses her friend Chat for work on the regular. She just submitted a FOIA to our local government who is doing some silly stuff with a special assessment taxing district....her FOIA was more professional and way more thorough than a Wall Street top 50 firm could have produced for @ 10k. It's impressive. The gub'mint attorneys are flummoxed and sent up a white flag.

She will be using AI to sort and categorize and summarize 8-10,000 pages of PDF and it's going to do in @ 5 minutes what it would take a team of $350/hr attorneys a week to do.

(10-14-2026 08:57 PM)rath v2.0 Wrote:  I just started using Gemini. It's eye opening. I'm an attorney but I have to say Al is a fabulous tool for allowing people who don't have a legal background to be able to access info. It's a sea change for legal information. No longer do you need a career of experience and a Westlaw or Lexis Nexus account to know the baseline case law on niche areas of law.

My wife is an AI wizard....she is able to do deep dives and tweak things to the nth degree....she uses her friend Chat for work on the regular. She just submitted a FOIA to our local government who is doing some silly stuff with a special assessment taxing district....her FOIA was more professional and way more thorough than a Wall Street top 50 firm could have produced for @ 10k. It's impressive. The gub'mint attorneys are flummoxed and sent up a white flag.

She will be using AI to sort and categorize and summarize 8-10,000 pages of PDF and it's going to do in @ 5 minutes what it would take a team of $350/hr attorneys a week to do.

The advantages of AI are mind blowing.

(10-14-2026 08:57 PM)rath v2.0 Wrote:  I just started using Gemini. It's eye opening. I'm an attorney but I have to say Al is a fabulous tool for allowing people who don't have a legal background to be able to access info. It's a sea change for legal information. No longer do you need a career of experience and a Westlaw or Lexis Nexus account to know the baseline case law on niche areas of law.

I don't know about Nexus, but Westlaw has two models using Agentic AI, Westlaw Advantage and CoCounsel Legal. These utilize the Westlaw database along with the firms own documents which would be much more useful to practicing attorneys than just using your random "freeware" AI.

But I agree...for the layman...they are great tools.