Yesterday, I gave background on my experience with this topic
and the idea that “Data Centers are poisoning our communities.”
Water is just about the most common resource on the planet. Just about the most easily accessible and manipulable, and one of the cheapest resources to refine.
That doesn’t mean it’s easy, cheap in absolute terms, or unimportant.
Most public policy issues with water come down to the water being pure enough to healthily drink, or to discharge into the natural environment. As I put it yesterday, “as part of the ‘drank from the garden hose’ generation, I have a lot of disdain for the hype about water pollution in this context.” The difference between potable water and drinkable water is definitional.
One thing I don’t think gets enough acknowledgement: a building produces “waste water” by existing. I recently had reason to calculate the roof run-off of my house, garage and barn, at about 275 thousand gallons per year. That is “OVER A QUARTER MILLION GALLONS OF NON-POTABLE WASTE WATER OMG!!!11!1!”. This is a small family farm.
I covered a lot of this yesterday from the angle of “pollutants”; but to recap: The water in your sink is contaminated. The water in your toilet is contaminated. The water from your roof, and in your garde hose are contaminated. They’re NOT “polluted”. They’re perfectly acceptable to discharge into the ground (via a septic tank) or a municipal sewer. Storm water run-off doesn’t generally get discharged into the same sewer system in cities, but ONLY because it isn’t septic (doesn’t have humans defecating directly into it- unless you’re in Portland or another big left-wing city) and MOSTLY because the sheer volume of it is impractical to treat in the municipal sewage treatment facilities. In fact, a big problem in Portland (as an example reasonably near me) is when heavy rains overcome the stormwater drain capacity and the septic sewers and storm sewer overflow with sewage sludge into the Willamette River. This is an example of a problem of scale and design at big city levels. But the storm runoff of any big building and its surrounding infrastructure like parking lots is naturally going to have literal tons of bird feces washing away at any given time. Sorry folks, the world is not sterile.
This water is not consumed and forever beyond re-use. It isn’t removed from the natural environment; it is BARELY diverted from where it would have gone anyway. But the devil is in the details. Concentration and channeling of runoff is a significant problem because once in place, a building’s runoff always goes to the same place. Over time, that matters a lot, erosion and concentration both being compounding problems.
But for at least a century, we have had some form of stormwater and wastewater planning and permitting process, that has been incrementally improved over that time. It’s NEVER going to be PERFECT. But at some point, the line plotting safety and the line plotting expense cross somewhere deemed to be “good enough” based on data and professional estimation.
Just keep in mind, that at the end of ALL of these water management processes, there are 3 and only 3 places the water can go.
1. Into the air
2. Into the ground
3. Into the ocean (generally via a river)
It doesn’t disappear, and it isn’t ‘consumed’ in the sense of not existing any more. It isn’t ruined. It is far easier to remove contaminants from water than almost anything else. You just distill it. It may not be fit for drinking without processing, but in modern risk management mindsets and legal definitions, no water is “safe for drinking” that hasn’t been processed and tested and monitored. It’s a definitional thing.
Industrial buildings have several options for cooling.
Purely passive air convection is the most basic. Doesn’t work well if there is a big source of excessive heat in the building. But there are some surprising effective low-tech passive cooling designs out there where the whole structure is designed around cooling, rather than other functions.
Open loop systems expose the coolant (for pure open systems, generally water) to the source of heat. This doesn’t mean they flow water over and in contact with the hot thing. It usually means they have a cooling loop, like a radiator in reverse, exposed to the radiant heat of the object needing cooling. The coolant flows through the loop, picks up heat, and is discharged along with the extra energy. Like spraying water from the garden hose on your hot roof during the summer, the water picks up heat and runs off, leaving the surface somewhat cooler than it was.
This is not very efficient, and there is the effluent to manage. It doesn’t scale, and in most cases is not a permitted use of water (you can’t get a building or use permit to do this.)
More commonly, there is a hybrid system. There is a closed loop, and an open loop working together. The closed loop is designed to be efficient at picking up heat, and the open loop is designed to be cheap at discharging it. This is where we get cooling towers and evaporative cooling.
Most of the time, the municipal supply is used by default because it’s a convenient source of relatively clean water to start with. Sometimes it still has to be filtered again prior to use. Some industrial processes use de-ionized water. It can’t even have the normal mineral content that humans NEED in our water (trace minerals) and cannot tolerate the additives like chlorine and fluoride used for sterilization and tooth support, respectively.
Both loops COULD be water. Usually, the closed loop is a liquid with a better ‘coefficient of thermal conductivity’ than water. A material which transfers heat faster, more efficiently, at lower or higher temps without boiling. Water is a great heat conductor, but glycol is better. You can expose glycol to a heat source, then expose the open loop evaporator to the glycol, and the water takes up energy by evaporating into the air. Evaporating water takes almost no other trace elements with it. It’s essentially distilled water at that point. It leaves any mineral content behind. In general, it takes more area to dissipate heat than to collect it. This disparity drives the ‘heat engine”. Taking advantage of area dispersal and gross temperature differences is an old technique and can be pretty cost effective, depending on the cost of water and space (area) to discharge heat. You use fairly cheap fans and liquid pumps, at relatively low pressures, so all the machinery is easy to make and cheaper.
The “cooling chemical”, more properly referred to as coolant or refrigerant, stays in the closed loop. It’s the expensive part, and the part that has potential to affect the environment negatively. We use coolants in ALL KINDS of applications. Refrigeration of food, automobile air conditioners, automobile radiators for cooling the engine, building HVAC systems, and specialized industrial processes. Over the years we have determined some coolants are more problematic than others for disposal or leaks, and standards have evolved in regards to allowable uses and types. The refrigerant used in car AC and home heat pumps has changed a lot in recent decades.
The evaporated water (a small part of the overall water used) becomes rain. The remnant water that is not evaporated (we’re talking large flow for efficiency) has slightly more concentrated levels of whatever mineral content that already existed in it coming from the source. Not high enough to require specialized treatment. This water is not “used up”, or eliminated. It simply isn’t ready to drink without being verified. It does tend to have a higher than ambient temperature, which is the biggest reason not to discharge it into natural bodies of water, as stream temps affect aquatic life and they are adapted to the existing temps. Sometimes the solution is as simple as a settling pond on the premise.
Evaporative systems and most forced heat pumping systems for that matter, have ranges (min/max ambient temperature and humidity) where they are most efficient, and they work best when the differential is significant within that range. If you want to cool a hot space, it’s best if you’re pumping hot coolant to a quite cold place to exchange heat. Evaporative cooling towers work best when the air isn’t already saturated with moisture (high humidity), where the ambient temp is not slow it causes the evaporated water to condense back into the tower or so high the closed loop can’t ‘lose’ heat to it. Altitude and partial pressure of the atmosphere at your location can also matter (see Boyle’s law discussion below).
Inside the cooling equipment there IS a potential to pick up contaminants that affect health. If the cooling equipment is built with substandard materials, solder that isn’t rated for drinking water, etc, some unhealthful metals could leech into the effluent. This doesn’t mean the water is fore ever poisoned or unusable. Building codes require cooling system discharge to be within tolerance before going into municipal sewers, where the water gets treated anyway.
Depending on the municipal water supply’s mineral content (varies wildly across the USA, as long as it is within EPA tolerances), cost, and delivery availability, a closed loop system might find they don’t want the municipality involved. They might have water trucked in, or tap a natural supply from roof run-off or natural water source, and filter it themselves. In that case, the water they discharge could potentially be “cleaner” than what they take in. Again, they have to have permits that have been through the planning and approval process for all of this. The water isn’t being diverted to Mars.
There are also entirely closed-loop systems. These generally do not use water at all. Your kitchen refrigerator is a ubiquitous example. It has two stages, and uses a compressor and Boyle’s law to move heat. When you compress a gas, it gets hotter. When you decompress it (allow it into a larger volume) it gets colder. When you hook up two big BBQ grills to a single propane tank and turn on all the burners, you may let the gas out so fast it freezes the gas in the valve. That kind of thing.
All systems have less than perfect efficiency. A closed loop system still needs top-ups due to leaks, maintenance, etc. But closed loop systems aren’t deliberately discharging coolant at all. They’re just pumping it from one pressure vessel to a different containment to benefit from the changes in ambient heat and pressure. Glycol and other thermally-efficient coolant liquids are expensive. You don’t discharge “cooling chemicals” and refrigerant, you keep it in closed loop systems. Data centers are hardly the only, or even primary user of this tech. Most office buildings use it too. Home (window units, heat pumps and mini-splits all) and automotive air conditioning is all closed-loop.
Closed loop systems are more energy intensive than open loops systems with evaporative towers. You’re forcing higher and lower temps through pressure changes, more than simple heat exchange through exposure and conduction. That requires higher pressure piping, pumps, and more powerful fans. When you do maintenance, you have to pump out the refrigerant/coolant into safe temporary storage and replace it when maintenance is done. Leaks are a bigger risk of contamination (modern refrigerants are less problematic than the past but not without impact.)
Closed loop systems for building HVAC mostly don’t use water as part of the process. People seem to be (deliberately?) conflating power plant (hybrid system) cooling towers with building HVAC. They are not even close to the same thing. Water just isn’t efficient enough at heat transfer to make closed loop systems work well. But even if it uses water for the cool side, and has a REALLY BIG reservoir to do so, that water is maintained in the system, minus only leaks, maintenance, and unavoidable losses of efficiency. It isn’t being “used up”, it’s being used “over and over”.
The word “millions” gets thrown about a lot. It’s a big number. But context matters. A million leaves on the trees in the park only sounds like a lot until you think about it. A million gallons of water is less than two Olympic size swimming pools. It’s a lot of water for an individual, it is not very much for a process or facility that serves thousands of people daily. I pointed out above that my home and barn combined have “over a quarter million gallons” of rain runoff per year. Your home probably has at least half that if you live a rainy place like I do. If you’re an office worker, the building you work in probably doubles it. If you work in a smaller commercial building with a parking lot, probably quadruples it. The data center you’re bitching about almost certainly generates more stormwater runoff than it uses in cooling. These things matter, but they’re generally accounted for in the building planning and permitting process already. It all adds up, but it isn’t a catastrophe, and certainly not a unique, new problem.
The water isn’t gone, and it isn’t ruined forever. Data center cooling systems are not going to turn your home into a desert. If you already live in the desert, you won’t notice a difference.
Data centers are ALL AROUND YOU. You don’t even know they’re there. They are just another ugly industrial/commercial building and you have no idea what is happening inside. Many of them are in one or two floors. Or the basement of an office building. A bigger national chain bank building almost certainly has a datacenter in it. I’m not saying there is no negative impact to them. Lots of reasons to not want a big industrial building next to you. But plenty of reasons to want them to exist: without datacenters, you would have no electricity at your house to read this, no Internet to read it from, and you’d be using metal coins and paper folding cash to pay for everything.
It’s quite likely that the AI Hype is causing over-speculation into data center investment. But since a data center can be housed in nearly any modern commercial building shell, that construction will be largely fungible. Right now, a great deal of the discussion is on how much heat this dense computing will generate, but efficiencies in that aspect of computing have been a target for a long time, and it keeps improving. Crypto-mining has largely jumped the shark, so super dense GPU card (a very hot tech, literally) manufacturers are probably loving this hype right now, and salivating over anticipated sales to replace the waning enthusiasm for crypto-mining as electricity prices have risen… but here is a prediction. Over the next 24 months, this market segment will cool off significantly.
As more big tech companies (and some of them are already backing away) fail to find the expected benefits and savings in LLM implementations because it doesn’t live up to the ‘intelligence’ part of AI, and more space and efficiencies are found in existing datacenters and spare capacity on existing cloud infrastructure is found for additional workloads, many of the projects now on the drawing board for discussion will be cancelled or scaled down.
Some form of this tech is going to be with us for the long haul, but we are not currently on the path to machine consciousness, and the appeal of a digital sycophant designed for addictive levels of engagement rather than useful computing rigor, is already wearing thin. The base of this tech is too heavily entwined with social media algorithms to be a great base for helping us live a better life. It is a useful but flawed tool, and it is designed for things most of us don’t actually want while being marketed to us as something it isn’t.
If I’m wrong about it*, in 2 years you have my permission to throw pumpkin pie at my face, as long as some lands in my mouth.
Larry Good, Author at Ordinary Times
Facebook: Lawrence Good, Scribbler
*Y’er wrong about it, Larry. It’s just right now few people have any idea how to use it, but those are massively more productive than they ever were, and that is going to redound in marked benefits as we go on.
Also we’ll all find uses (better uses than now) for them as we go on. It takes time. What you’re saying is the equivalent of “Computers are a fad. They’re not useful for most of us.”
This is said without prejudice, because at least you’re arguing in good faith, instead of trying to panic us with water or something. I won’t throw pie, but I will reserve the right to say “I was right!” – SAH
According To Hoyt 
Thank you for writing this up, Larry.
LikeLiked by 2 people
Good information; sadly, the opposition can fit their complaints on a bumper sticker or protest sign, coincidentally approximating the upper limit of a typical leftist’s attention span.
LikeLiked by 2 people
Do you have some official-looking source I can send to people about this? There’s a wave of astroturf activism building in my area and I’d like to be able to at least get some facts out. But sadly a blog post, no matter who it’s written by, is too easy to dismiss. Something from an Agency or an Institute, perhaps?
LikeLiked by 1 person
Create an “Institute.” Post this on that agency’s website. Profit.
LikeLiked by 1 person
LikeLiked by 1 person
I agree with Larry fully on the lack of overall significance on water use for Data servers, compaired to all the other uses of water. I also agree with Sarah that Larry is wrong about Data Centers being a “fad”.
AI is a tool, much like steam was a tool. Is it intelegent? Don’t make me laugh, of course it’s not.
Is it very useful for certain things, like patern recognition, and analysis? Oh Yes. It’s an Idiot Savant. It depends on being given good data and having the right questions asked. And like all other computing tools, GIGO. There are a large number of uses for it, from examining a process (ANY PROCESS) to determine where ineficciencies lie, to steering vehicles for delivery use, whether that delivery is from Amazon, or the US Airforce.
LikeLiked by 2 people
Thank you for both today’s and yesterday’s posts. Currently, our town (Barrington, NH) and the state, is in a frenzy over a proposal to build a data center in Nottingham, NH. The protestors were out in force at the proposal hearing at the Nottingham school the other night. And this was just a preliminary proposal, no engineering proposals yet to actually consider evidence. The level of knee jerk NO! was unbelievable. Ditto the reactions to the article on it in the Granite Grok site (good place for more than just 3 sets of memes each week.)
LikeLiked by 1 person
Apparently this “grass roots opposition” is being astroturfed and heavily funded by Neville Roy Singham.
LikeLiked by 1 person
c4c
LikeLike
Hey, sounds fine to me. 😛
I saw an interesting example of Boyle’s Law at a shipyard. They had blowers powered by compressed air, which could be used in places you didn’t want electric motors, like where flammable vapors could be present. After an hour or two of operation, the pneumatic motor housing and pipe fittings would be covered with a thick layer of frost and ice just from controlled expansion of 100-150 PSI compressed air.
LikeLike
Television is just a fad, I know I heard it in the fifties….
Let’s wait a few years and see what everyone is freaking out about then and get excited about that.
Humans are brilliant, but people are just dumb…
LikeLike
Agent K: “A person is smart. People are dumb, panicky dangerous animals and you know it.”
LikeLiked by 1 person
Hmmm, how many ‘just a fads’ have there been?
“Flying machines are just a fad!”
“Cars are just a fad!”
“Electric lights are just a fad!”
“Trains are just a fad!”
“Iron is just a fad!”
“Writing on shredded plant fibers instead of stone tablets is just a fad!”
“Sticking seeds in the ground and watering them for months is just a fad!”
LikeLiked by 2 people
“Painting on the cave walls is just a fad! Go out and kill mammoths rather than waste your time doing artsy fartsy crap!”
LikeLiked by 2 people
The Story of Ung
LikeLiked by 1 person
When has ANY new technology EVER been stopped by people protesting against it?
it’s been slowed down, sometimes until the people blocking it die from old age but never killed off by protestors
(nuclear being the closest to being killed off, but it’s just too useful and now finally may be getting unleashed)
The only time a technology disappears is when something newer/better makes it obsolete.
LikeLiked by 1 person
When has ANY new technology EVER been stopped by people protesting against it?
it’s been slowed down, sometimes until the people blocking it die from old age but never killed off by protestors
(nuclear being the closest to being killed off, but it’s just too useful and now finally may be getting unleashed)
The only time a technology disappears is when something newer/better makes it obsolete.
LikeLiked by 1 person
Damnit Sarah, Gimme the pie!
Seriously, the tool will evolve greatly from what it is now. It won’t go away. So all the people pointing to past statements like “there’s only a need for 3 or 4 computers in the world” can calm down. That’s not at all what I am saying.
What I’m actually saying is that we’ll need to rebuild some major underpinnings from the LLM model (which was more or less discovered by accident) to a more deterministic model. If you want your own personal Jarvis to help you do engineering projects, it can’t be based on a toy that blows smoke up your ass. Bridges don’t stay intact just because you feel good about your ideas.
We had a major Internet bust back in 2001, not long after I got into the biz. It set my employment back temporarily. I haven’t forgotten. We’re going to see that with AI. The tool has potential, but the hype is overblown, because- marketing people.
In IT, we have a saying; a system is people, processes, and technology. Technology is listed last on purpose.
A business process driven by technology has a HUGE probability of failure. Business decisions need to be made for business NEEDS and business OPPORTUNITIES. Not just because “shiny”.
My doods, *I’m a technologist* and I’m telling you this.
The ‘Tokenmaxxing’ Tide May Already Be Turning — The Information
Uber’s COO says it’s getting harder to justify money spent on tokenmaxxing | Hacker News
Microsoft just cut its engineers off from AI because the bill got too big — why AI might not take your job after all – AOL
These articles are also premature breathless pearl clutching, but still useful indicators. The enthusiasm to convert everything to AI is baseless and counterproductive at this point. The situation will evolve over time.
LikeLike
Dan does not consider it a toy. Neither does Eric S. Raymond.
LikeLike
I consider it a very sharp and double edged sword.
LikeLike
Yes, this. Often being swung by a swordsman not always the most deft…
LikeLike
it has potential. But you can’t rely on a calculator that gives you false and random answers.
Vibes and cute images don’t get Lunar cities built.
LikeLike
Thank you for writing these.
–
At least in Oregon one is allowed to harvest water off of buildings without a permit, for both agricultural, landscape, and non-potable water household use (bathing and toilet bowl), latter requires two non-interacting systems. It can be used for drinking water with approved, permit process, filtering and sterilization systems. Suspect Washington is the same, at least western side, but I didn’t ask Google.
Kelso, Washington? Then you are familiar with Longview Columbia college. That is where I started my programming path. We owned a home off of Columbia Heights (’80 – ’89*). Lived there from ’79 – ’85.
(*) ’80s economy.
LikeLiked by 1 person
There are two things going on here, to my eye.
First, all the Greenies will fight -anything- no matter what it is. Water is their new hobby-horse (we’re running out, you know) so watching them lie about “waste water” used for cooling is not a surprise.
There is going to be a LOT of money spent fighting data centers because the Chinese want to win the AI battle. They think AI will make them the rulers of the world, so money spent encouraging NIMBY activity in America looks like a great idea to them.
Scratch a Greenie, find a Communist.
Second, and probably related, people are genuinely IGNORANT of how anything works anymore. Partly because our education system is essentially babysitting service, partly because of the sheer volume of lying in the public sphere.
The truth about data centers is that the water comes out of the cooling system cleaner than it went in. But it doesn’t matter what the truth is, because people don’t know that and there is an entire industry of liars dedicated to lying about it. Some of whom are foreign adversaries off the USA, some are domestic adversaries. The lying is the same though.
The big hurdle is that lying to the uninformed, about things that don’t interest them even on a good day, is as easy as falling off a log. Getting them to wake up and smell the coffee is essentially impossible.
But show them they’re being swindled? Now, suddenly, they care. There’s money involved, right? Money is serious business.
LikeLike
K, a trick to watch for with water, is counting pumping out of a settling location as new gallons of water– especially if the Settling Pond takes runoff from the building, or graywater from the sinks.
They pull this on farms that made run-off ponds to avoid nitrogen runoff.
LikeLiked by 1 person
To try to describe this for non-farmers— picture that you’re watering trees on a tilt, and there’s a nice big pond at the bottom. You fill the watering system from that pond.
And the measurement for “how much water is consumed” is based off of how many gallons go through the pipe.
There are non-evil reasons for this, but evil people use it in order to penalize those who are trying to actually reduce water issues.
LikeLiked by 1 person
“…but evil people use it in order to penalize those who are trying to actually reduce water issues.”
Of course. Greenie=Communist, right? Or just tax collector, closely related type of evile.
It’s genius to the kleptocrat set. Tax people for cleaning the water on their own land in a settling pond, and then tax them for using the cleaned water.
I’m sure they have a way to tax you for cleaning out the settling basin too, and using that stuff for fertilizer, I just don’t know what it is. Effluent disposal tax or similar.
Here in Peeples Republik of Canuckistan, recently there was a regulation passed in my county (or possibly in the whole province, I can’t remember exactly) that banned the use of septic tank effluent as fertilizer. Now they charge the #PumpGuy to empty the truck at the town water treatment plant instead of having a farmer pay #PumpGuy a small gratuity to spread it on his fields.
The reason? Run-off, of course. Some of that doody -might- make it into the river, y’know. No word if spreading manure is going to be illegal soon, I’m sure they really, really want it to be.
Because Greenie=Commie, and Commies love to starve the peasants. It’s the one thing they’re actually good at.
LikeLike
“Conservationists.”
LikeLike
“Conservationists.”
My favorite. Arms-length apparatchiks for the Chicom regime, working in North America and/or Europe.
I mean, they finally found the Quality Learing Center, right? I’m sure a bunch of college educated gentlemen of questionable morals can figure out a way to get the government to pay for their Lamborghini and their trophy wife’s plastic surgery bill while they lie about the environment.
LikeLike
*raises claw, not middle claw* Environmentalists. Conservationists are wise-use oriented, and look at how to do the most with the least waste of a particular resource (good forestry, proper land use, rotational grazing and mixed ag, those sorts of things). Environmentalists are the “it must be pristine!” and/or “save all the non-human things!,” and/or “we must expiate our sins and failures by offing our species and living a Neolithic lifestyle*.” And yes, I realize that the media and activists have conflated the two almost to the point of being meaningless.
*Except some also eschew any farming, or grain farming. Y’all go first.
LikeLiked by 1 person
Yep. Conservationists are the ones trying to do the work. Environmentalists have other agendas than the environment in question. Always. And always much, much higher in the hierarchy is money and power than anything that does anything to or for the environment.
LikeLike
Any term that is GOOD, they try to steal.
The rule my folks taught me is to ask what are they conserving, what do they want to “return” the wild to.
If I remember right, you’re a few years older than me, and likely pioneered even if by just passive support the “sane world” meaning for “conservationalists.”
While I was reading RANGE magazine as they skinsuited it, and did things like insist that you needed only historically testified species to get the desired results for their claimed goals.
The real question– and yes, Iknow you know this!– is “conserve to what end” and “Restore to where?”
To put my cards on the table, I want a roughly stable decades-on-end goal that allows native animals to survive while also supporting humans.
which is vital because the single best survival strategy for ANYTHING ON EARTH is “be good for human food.”
LikeLike
Yep. At least nowadays. The dodo, passenger pigeon, and great auk had the bad luck to be good for human food back in the days when communication took way longer and mass communication was nearly impossible, so by the time anyone realized they were going extinct (I know that happened with the passenger pigeon, not sure anyone ever realized that about the dodo and great auk) it was too late to farm them. (Attempts were made for the passenger pigeon, but apparently it didn’t breed well in captivity, so the attempts failed).
LikeLike
Yet another example:
https://redstate.com/wardclark/2026/06/02/australias-net-zero-folly-soaring-costs-now-forcing-massive-crop-cuts-n2202949
LikeLiked by 2 people
Ah yes, when you make more money for not-planting than you do from the actual crop. Because fuel costs. Nice.
Australia. Same crap as Canada, but with bigger spiders.
LikeLike
Yes, but are they Communist spiders, Comrade? Asking for the census. Taxes won’t collect themselves and these books need to be pencil whipped into shape or the party apparatchiks won’t be able to afford their third homes this year.
LikeLike
As long as the MSM exists, spreading manure will continue to be legal.
LikeLiked by 1 person
The hype on the internet is thick with this one.
While data centers can use large amounts of water for cooling, good engineers design a system where the water is used efficiently, and doesn’t require constant replenishment, which costs money. Contaminants are not nearly as great a problem as with sanitary sewer treatment. I’d think any contaminants would be minor, and easily treated.
LikeLiked by 1 person
“…large amounts of water…”
It always sounds like such a huge, ridiculous volume, doesn’t it? Because the Greenies count it every time the same water goes through the cycle.
If you look at how much comes in and how much goes out, they fill the tanks one time, and then only make up what’s lost to evaporation and leaks, flushing tubes for cleaning and etc.
As in, hardly any. Thousands of gallons, not millions like the Greenies are constantly going on about.
Because lying liars lie.
LikeLiked by 1 person
On the (apparently clear) disagreement between Today’s Guest Author,
“Some form of this tech is going to be with us for the long haul, but we are not currently on the path to machine consciousness, and the appeal of a digital sycophant designed for addictive levels of engagement rather than useful computing rigor, is already wearing thin… a useful but flawed tool, and it is designed for things most of us don’t actually want while being marketed to us as something it isn’t.”
and Our Esteemed Blogmistress,
“Y’er wrong about it, Larry. It’s just right now few people have any idea how to use it, but those are massively more productive than they ever were, and that is going to redound in marked benefits as we go on. Also we’ll all find uses (better uses than now) for them as we go on. It takes time. What you’re saying is the equivalent of ‘Computers are a fad. They’re not useful for most of us.'”
— I tend to come down, and pretty strongly, on the side of “You’re both right, or mostly.”
Hype-bubbles, the commercial-financial kind not just the notoriety and seven-day-wonder kind, don’t only happen because the underlying “thing” is “just a fad” — they also happen because the underlying “thing” is either not mature, or currently being mis-packaged or mis-sold, or both, and-or any of a number of other things. It can take a long time to get “there” for real, and demand much further cleverness, innovation, and development first.
Currently “AI” (cough!) is being massively over-hyped, and current “AI” chatbot oracles (“Ask Grok” and kin) are indeed typically “designed for addictive levels of engagement” with users (name one that doesn’t provide you with assorted hints for more queries at the bottom of its answers). The fact that many of us are savvy enough already not to take the bait doesn’t mean it isn’t there, or isn’t effective on a lot of the naive or credulous or easily-psy-opped (and ads are of course mass influence ops).
Our “personal computer” would’ve been ludicrous until the 1980s or so; the old mainframe “time-share” services were really limited (110 baud modems, or maybe 300, 1200 if you were very very lucky), and “Joe user” software was almost nonexistent, largely from having almost no market to drive it.
The first personal computers were a fad, they were indeed not useful to the “most of us” — how many people would have had the patience to toggle in a a bootloader one data bit at a time using those front panel switches, and perfectly to boot (so to speak), every last single time they turned on the machine? (Imsai 8080, etc.) Does anyone else remember audio-tone data storage on a (stock) tape recorder (Apple II, as in pre-Mac)? Or (interpreted!!) Basic on a TRS-80 or Apple II, perhaps?
But in the fullness of time they got better. A lot better. Usably so.
Meanwhile, so much of the hype, blather, and just plain ad-copy (whether ever clearly identified as such for us or not) is somewhere between hinting and trumpeting how we’re not only “on the path to machine consciousness” but it’s just around the corner. (Despite whole, clear, factual books to the very well-researched contrary.) All we need now is one big short-term stark collision of wishful fantasy with cold hard reality, and… bingo, there goes your bust, bang! (“AI is toast!” I suppose.)
That’s not to say these technologies (mostly LLMs, run on “deep” as in more than two layers, neural networks) aren’t already doing some really impressive things: look at “amplified artistry” by “our” Cedar Sanderson or Sarah Hoyt, for instance; or listen to the latter’s latest clankerfilk. More to the point, watch “AI” (cough!) in Intelligence Amplification mode help hugely with, say, software writing, debugging, modifying. Or even (in a quietly supporting role) fiction/nonfiction writing. And do remember, all these things — like “deep” neural nets trained using “error backpropagation” and deployed using the “transformer” architecture — are of really very recent vintage.
My own persistent idea (though we might need considerable “on the ground” data to make it work) is using large-net “machine learning” on images of (say for instance) Mars — to locate rock types, buried ice, ores of assorted kinds — over an entire planet at once. Sound useful to you? (Note the part about likely needing somewhat of a Martian Geological Survey busting rocks first, though.)
My guess, and it’s more and more of a firm conclusion as time goes on, is that we (see Sarah’s quote) barely have any clue, right now, how to use these systems, well — even making them has been compared to medieval alchemy rather than real chemistry — and, of course, the decades of encrusted mythology and ingrown speculation about “machine minds” turning humans into cyber-pets or UBI-Eloi etc., etc., etc., don’t help much at all with that. (Slash and burn early agriculture comes to mind; slash it with a machete, burn it with fire, repeat as needed.)
Seen from this angle, current (all-knowing, riiight) “chatbot oracles” seem about on the level of CompuServe or America Online — good interim ideas that got left behind in our rush to a better future. Maybe it was almost inevitable that our first visible, big-splash use of “AI” would be in the same centralized, industrialized, oligopolized sort of “trust-era” mass service from a hub paradigm that Karl Marx could have known and loved. But, as with the personal computer era succeeding the mainframe era, it’s all too plausible we’ll (at least mostly) get over that one too. (Does good ol’ Compuserve even let you rent time on a mainframe anymore? Well, I have it on good authority that Lunanode still lets you do that, if you want.)
Likely you’re both partially wrong; I’m sure enough that this here is. But my view remains, you’re also both far more right than wrong.
LikeLike
I don’t know. Claude hands me my marching papers if I loiter. “We are done with this topic” or something like it.
LikeLike
Mrs. Hogarth mentioned on X that Claude got terse and forgetful with the last model upgrade. Anthropic might be trying to dial back on token usage so they have enough compute to go around.
LikeLike
Yes, and also see recent deal for Anthropic to lease one of xAI / Space X’s big data centers for (IIRC) $1B+ per month; Brave Search’s “answers by AI” says $1.25 B per month. Obviously, A. was feeling some notable tightness in its “compute” (parts of speech, hah! this is jargon, grammar can go to the devil) to pay rent at such a level ($15 billion per annum if it continued that long).
Suddenly I’m reminded of Ian B.’s comment from a few days ago, about a few clever people using “AI” to massively increase their own productivity, while assorted big companies pile up giant wads of cash and use “AI” as an excuse/mechanism to light it on fire…
LikeLike
“Does anyone else remember audio-tone data storage on a (stock) tape recorder (Apple II, as in pre-Mac)?”
/raises hand
My Dad got an Atari 800, and for whatever reason got the branded cassete player that Atari made. It was *incredibly* unreliable. It also didn’t make much sense to have, as Atari also made perfectly functional external 5.25″ floppy disk drives (including one that had *two* drives in the same machine!), which he also owned. And you didn’t have to worry about having the floppy disk drive run for half an hour, only to discover when it finished that something hadn’t loaded off of it properly (which happened frequently).
LikeLiked by 1 person
Never actually used the audio interface, but the Apple had one. It also had a 5 1/4 floppy drive, plus a DOS (mostly a Basic interpreter) to go with that. Wrote most of a Forth on that machine, using the “RWTS” (read-write track-sector) call in the system to do disk I/O; and that even worked well. Only very shortly after all that was working properly, the Mac 128 and then Mac 512 (1st useful Mac) came out — with a 68000 instead of a 6502.
“Itsy bitsy spider, ran up the water spout,
“Down came the rain, and washed the spider out.”
Tech moves fast, yes sir.
LikeLike
Oh, we’re both certainly wrong in some part.
BUT, I am not predicting AI to fail and go away.
I’m just saying there will be dips and bumpy spots; what we have NOW isnt what we’ll have when it becomes broadly useful. A hallucinatory self affirmation chatbot is much less useful for most people than getting rigorous answers would be. But most people are still playing not working with it. That’s why I say its a toy at this point.
Lots of toys turn into useful things eventually. Or learning tools.
Lots of products are hyped as the killer app, best thing since sliced bread. They don’t all prove out.
LikeLiked by 1 person
Okay, this one I have an issue with – When was the last time anyone exposed to any standards organization in the real world ever saw any standard loosened? The issue is there is no constituency within the standards setting org for relaxing or revising downwards the standards that org publishes – the ratchet only goes up.
Each little standards increment can be fully justified by some obscure corner case or another, but the organization as a whole would lose its purpose if “good enough” were reached. “Okay, we’re done, nothing else needed, wrap it up and let’s go home.”
Standards orgs ratcheting up the acceptable minimum is pretty much what they do.
LikeLike
What they get paid to do, too. Since they are getting paid to set standards that is what they’re going to keep doing. Because that’s where the money is. If they stop setting new standards, all that’s left is the fees and inspections and suchlike and nobody likes that. So once folks get used to the old standard and stop getting infractions, there needs to be a new one else what use is the standards making company?
LikeLiked by 2 people
space x ipo target per share price $135, 556million shares. $75 ish billion. You’ll pay whatever the market is when/if you buy, but that’s what they’re targeting. Current revenue and growth do not warrant that, more like $70-75 per share, but given what they’re paying for anything even vaguely ai they might well get it. Actually, given NVDA is a 5 trillion dollar company thats forecasted to multiply net income 13 times over the next five years (hint, they won’t,) space x might be cheap. 😜
I had though there’d be fewer shares at a higher target price, but it looks like they want retail — that’s us — buyers too.
LikeLiked by 2 people
And the price will likely shoot up early on. The questions are how quickly it starts to come back down from the initial peak, and whether it actually settles above or below the initial asking price.
On Memorial Day, I discovered that a cousin that I don’t see very often works for SpaceX. All of the employees have the option to invest some of their pay into stock shares, which he has done. He’s hoping that the IPO goes well, though he also mentioned AI as a possible reason why the enthusiasm might be limited.
LikeLike
On the uses of AI, I present the example of my wife. She is building a little business making customized objects (mugs, fridge magnets, etc.) that have the exact pictures the customers want.
Where AI comes in is that it speeds up her workflow in processing images and making customized designs so that she can actually reasonably compete on price with mass produced designs. It is actually quite fascinating to see the ecosystem growing.
LikeLiked by 1 person
The thing about LLMs is they are pattern engines. That’s a very powerful thing that, up until now, hasn’t been able to be automated to any degree.
What they let you do is find things that match the shape of the hole you’re trying to fill. But that also requires you to understand the criteria of the solution you are looking for, otherwise you’ll fill your cake pan with cement and end up with a GladOS surprise.
But, once people actually understand how to use it, I expect the yields to be huge. One of the things I’ve noticed about engineering is most problems are already solved, usually by someone a hundred years ago, who did it all in theory, patented it, and never found a use for it. The ability, once you can define the problem, to have a machine quickly go through the entire US Patent database and any random journals that may or may not have been related to the specific field to go find the thing someone invented and published before any of us were born, that potentially meets the solution criteria is huge.
But you absolutely have to know enough about what you’re doing to define the problem and understand what comes back.
This is why it ends up being very useful for programmers, but fuzzier for writing. Most programs are defined blocks of problem solutions. Stories are weirder and squishier.
LikeLike