Monday, September 15, 2025

Data Centers On the Grid: Ballast or Essential Cargo?

 

Back in the days of sailing ships, the captain had a choice when a storm became so severe that it threatened to sink the ship.  He could throw the cargo overboard, lightening the ship enough to save it and its crew for another day.  But doing that would ruin any chance of profiting from the voyage. 

 

It was a hard decision then, and an equally hard decision is facing operators of U. S. power grids as they try to cope with increasing demand for reliable power from data centers, many of which are being built to power the next generation of artificial-intelligence (AI) technologies. 

 

An Associated Press report by Marc Levy reveals that one option many grid operators are considering is to write into their agreements with new data centers an option to cut off power to them in emergencies. 

 

Texas, which is served by a power grid which is largely independent of the rest of the country's networks, recently passed a law that prescribes situations in which the grid operator can disconnect big electricity users such as semiconductor-fab plants and data centers.  This is not an entirely new practice.  For some years, large utility customers have taken the option of being disconnected in emergencies such as extremely hot or cold days that put a peak strain on the grid.  Typically they receive a discount for normal power usage when they allow the grid operator to have that option.

 

But according to Levy, the practice is being considered in other parts of the country as well.  A large grid operator called PJM Interconnection serves 65 million customers in the mid-Atlantic region.  They have proposed a rule similar to the one adopted in Texas for their data-center customers.  But an organization called the Digital Power Network, which includes data-center operators and bitcoin miners, another big energy user class, complained that if PJM adopts this policy, it may scare off future investment in data centers and cause them to flee to other parts of the U. S. 

 

Another concern is rising electricity prices, which some attribute to the increased demand by data centers.  These prices are being borne by the average consumer, who in effect is subsidizing the gargantuan power needs of data centers, which typically pay less than residential consumers per kilowatt anyway.

 

In a way, this issue is just an extreme example of a problem that power-grid operators have faced since there were power grids:  how to handle peak loads.  Historically, electricity has to be generated at the same time it's consumed, although there is some progress recently in battery storage of electricity, though not enough to make much of a large-scale difference yet.  This immediacy requires a power grid to have enough generating capacity to supply the peak load—the most electricity they will ever have to supply on the hottest (or coldest) day under worst-case situations. 

 

The problem with peak loads from an economic view is that many of those generating facilities sit idle most of the time, not producing a return on their investment.  So it's always been a tradeoff between taking a chance that your grid will manage the peak load and scrimping on capacity, versus spending enough to make sure you have margin even with the worst peak load imaginable, but having a lot of useless generators and network stuff on your hands most of the time.

 

When the electric utility business was highly regulated and companies had a guaranteed rate of return, they could build excess capacity without being punished by the market.  But since the deregulatory era of the 1970s, and especially in hyper-free-market environments such as Texas, the grids no longer have this luxury.  This is one reason why load-shedding (the practice of cutting off certain big customers in emergencies) looks so attractive now:  instead of building excess capacity, the grid operator can simply throw some switches and pull through an emergency while ticking off only a few big customers, rather than cutting it off to everybody, including the old ladies who might freeze or die of heat exhaustion without power. 

 

Understandably, the data-center operators are upset.  They don't want to spend the money on backup generators that they would rather the grid operators spend.  But the semiconductor manufacturers have learned how to do this already, and build costs for giant emergency-generation facilities into their budgets from the start. 

 

Some data-center operators are starting to build their own backup generators so that they can agree to go off-grid in emergencies without interrupting their operations.  After all, it's a lot easier to restart a data center after a shutdown than a semiconductor plant, which could suffer extreme damage after a disorganized shutdown that could put it out of action for months and cost many millions of dollars. 

 

Compared to plants that make real stuff, data centers can easily offload work to other centers in different parts of the country, or even outside the U. S.  So if there is a regional power emergency, and a global operation such as Google has to shut down one data center, they have plenty more to take up the slack. 

 

It looks to me like the data centers don't have much of a rhetorical leg to stand on when they argue that they shouldn't be subjected to load-shedding agreements like many other large power users tolerate already.  We are probably seeing the usual huffing and puffing that accompanies an industry-wide shift to a policy that makes sense for consumers, power-grid operators, and even the data centers themselves, if they agree to take more responsibility for their own power in emergencies. 

 

If electricity gets expensive enough, data-center operators will have an incentive to figure out how to do what they do more efficiently.  There's plenty of low-power technology out there developed for the Internet of Things and personal electronics.  We all want cheap electricity, but if it's too cheap it leads to inefficiencies that are wasteful on a large scale.  Parts of California in the 1970s had water bills that were practically indistinguishable from zero.  When I moved out there to school in 1972 from water-conscious Texas, I was amazed to see shopkeepers cleaning their sidewalks every morning, not with a broom, or a leafblower, but with a spray hose, washing down the whole sidewalk. 

 

I don't think they do that anymore, and I don't think we should guarantee all data centers that they'll never lose power in an emergency either.

 

Sources:  Marc Levy's article "US electric grids under pressure from power-hungry data centers" appeared on the Associated Press website on Sept. 13 at https://apnews.com/article/big-tech-data-centers-electricity-energy-power-texas-pennsylvania-46b42f141d0301d4c59314cc90e3eab5. 

Monday, September 08, 2025

The Fading Glory of Subsidiarity

 

I'll get to subsidiarity in a minute.  First, here is why I'm writing about it this morning.

 

For many years, I have subscribed to the Austin American-Statesman, first in its hard-copy paper form, and then when that got insupportably expensive, in its digital form only.  Already by then, it was owned by a large media conglomerate called the Cox Media Group, but the operations and editorial control of the paper remained in Austin.  An outfit called GateHouse Media bought it from Cox in 2018, but relatively little changed when the owners of GateHouse bought the company that ran USA Today, Gannett Media, and moved the Statesman under the Gannett umbrella.  That caused some changes, but they were tolerable.  Back in February 2025, however, Gannett sold the Statesman to Hearst Communications, another media conglomerate. 

 

This may or may not have anything to do with what happened to me this week, but I suspect it does. 

 

I've been accustomed to propping my iPad on the breakfast table and reading the "e-edition" of the Statesman along with having my cereal and orange juice.  The software worked reasonably well most of the time, and until Wednesday of this week (Sept. 3) everything went smoothly. 

 

Suddenly on Wednesday, I was asked for a password, and the system rejected it.  After futilely trying to reset the password and getting no response from the paper's system, I called a help number and got connected to a man who said there was a software problem, and I should uninstall the Statesman app on my iPad and reinstall it. 

 

I tried that Thursday, but it didn't help.  Then I tried pretending I was a new subscriber (although I had found a place online which said my subscription was paid up until December of 2025), and tried to subscribe.  That didn't even work. 

 

Finally, I called the help line again.  I spoke to one person, who silently connected me to another person, who sounded like she was working in a boiler room with fifteen other people crowded into a space the size of a VW bus. 

 

She tried to identify me by name and phone number, and all those records had been lost.  (This was also the case when I called the day before).  Finally, she could locate me by street address, but it said I wasn't a subscriber.  I asked if she could look up my subscription record to tell when it expired.  She said because of the transfer to Hearst they didn't have that information, and would I like to subscribe now?

 

Seeing no other option, I said yes.  I'd already spent about half an hour on the phone, and I figured this was the only way to get my paper back.  It took about twenty minutes for her to take my information and put it in the system, and I could hear her asking for help in the background.  Then it took another twenty minutes for me to log on and get my new subscription going, and we never could change the password they started me with. 

 

The entire megilla cost me an hour of time I was not planning to spend, and $360 for a year's subscription to the e-edition only, which was the price point that made me cancel the hard-copy edition a few years ago.  We've had some inflation since then, but not that much.

 

If there is anybody under 40 reading this, you are probably wondering why this old guy is insisting on paying that much money for stuff that I could get for free.  Well, while I disagree with the editorial positions of the Statesman staff on most matters, it is still an edited entity that does a fairly good job of telling me what's been going on.  And for another thing, you can't find that many comics all in one place on the internet for free, or if you can I don't know where to go.

 

Now for subsidiarity.  It is a term from Catholic social teaching which describes the principle that "issues should be dealt with at the most immediate or local level that is consistent with their resolution."  That's according to Wikipedia, and while that source is slanted on some matters, they are right on with this definition.

 

Going straight to my problem with the Statesman, most of the paper is written and edited thirty-five miles away up in Austin.  The issue of my e-edition was a local issue, not going any farther in principle than Austin and San Marcos.  The principle of subsidiarity says that the problem with my subscription, and the list of when I've subscribed, and my credit balance which has apparently vanished into the bit void, and my passwords, and whatever other stuff is relevant to the problem including the authority to do something about it, should all be right there in Austin, and not stuck in some anonymous server farm in Seattle and controlled from a boiler-room operation in God knows where, and owned by a corporation based in Manhattan which clearly doesn't give a flip about how it treats its new customers.

 

Simply because technology did not yet permit otherwise, newspaper operations prior to about 1970 had to be local, in accordance with the principle of subsidiarity.  If my father had a problem with his subscription to the Fort Worth Star-Telegram, he'd get on the phone and call their office downtown.  A human being less than 20 miles away would answer the phone, and flip through physical pages of paper until he or she found the hand-written subscription records, and the issue would be resolved, or not.  People made mistakes with paper records too, but they were more easily resolved.  I have no idea what's gone wrong with my subscription to the Statesman, but again, only God knows exactly what the problem is, where all the loose ends are, and whether and how it can be resolved, because it's now so complex and involves incompatible computer systems and who knows what else. 

 

I don't have an answer to this problem, except to point out that if we try moving toward systems that are more in accordance with the principle of subsidiarity, a lot of these kinds of problems might take care of themselves. 

 

Sources:  I referred to the Wikipedia articles on subsidiarity and the Austin American-Statesman

Monday, September 01, 2025

Will AI Doom A Good Chunk of Hollywood?

 

This week's New Yorker carries an article by Joshua Rothman about what artificial intelligence (AI) is poised to do to the arts, broadly speaking.  I'd like to focus on one particularly creepy novelty that AI has recently empowered:  the ability of three guys (one in the U. S., one in Canada, and one in Poland) to produce fully realized short movies without actors, sets, cameras, lights, or any of the production equipment familiar to the motion-picture industry.  The collaborators, who call themselves AI OR DIE, use only prompts to their AI software to do what they do.

 

I spent a few minutes sampling some of their wares.  Their introductory video appears to show what a camera sees while it progresses through a mobile-home park, into one home, where the door opens, another door opens, and finally a piece of toast appears, floating in the air.  Another one-minute clip shows a big guy pretending to be a karate-type expert knocking down smaller harmless people.  If all the characters in that clip were creations of AI without any rotoscoping or other involvement of real humans or their voices, we are very far down a road that I wasn't aware we were even traveling on.

 

Software from a firm called Runway is not only used by AI OR DIE, but by commercial film production companies in mainstream films as well.  But so far, nobody has produced an entire successful feature film using only AI.  But it's only a matter of time, it seems to me.

 

Rothman quotes the AI OR DIE collaborators as saying how thrilled they are when they can have an idea for a scene one day, and then start making it happen the next day.  No years in production hell, fundraising, hiring people, and all the other pre-production hassles that conventional filmmaking entails—just straight from idea to product.  So far, most of what they've done are what Rothman calls "darkly surrealistic comedies."  If the samples I saw were representative, it reminds me of an afternoon I spent at Hampshire College in the 1990s, I think it was, in Massachusetts, where a screening of student animations was held.

 

Early in the development of any new medium, you will come across works that were made simply to exploit the medium, without much thought given to what ought to be said through it.  The short animations we saw that afternoon were like that.  The students were thrilled to be able to express themselves in this semi-autonomous way.  Chuck Jones, the famous Warner Brothers animated-film director, once said that animators are the only artist who "create life."  Most of the time they let their thrills outweigh their judgment.  A good many of the films we saw back in Massachusetts that afternoon were in the category of the classic "Bambi Meets Godzilla."  This film, which I am not too surprised to learn was No. 38 in a book of fifty classic animated films, at least meets the Aristotelian criteria of unities of action, time, and place.  There is one principal action, it happens over less than a 24-hour period, and it happens in only one location.  We see the fawn Bambi, or a reasonable facsimile thereof, browsing peacefully among flowers to idyllic background music.  Suddenly a giant foot—Godzilla's, in fact—drops into the frame and squashes Bambi flat.  End of story.  Most of the other films were like that:  a silly, stupid, or even mildly obscene idea, realized through the painful and tedious process of sole-author animation.

 

Just as our ability to manipulate human life technologically has led us to face fundamental questions about what it means to be human, the ability to only a few people to digitally synthesize works of art that formerly required the intense collaboration and technology-aided actions of hundreds of people will lead us to ask, "What is art?"  And here I'm going to fall back on some classical sources.

 

Plato posed that the transcendentals of truth, goodness, and beauty are things that lie at the roots of the universe.  According to theologian and philosopher Peter Kreeft, art is the cultivation of beauty.  Filmmaking is a type of storytelling, one in which the way the story is told plays as much of a role as the story itself.  And it's obvious that AI can now replace many more expensive older ways of doing moviemaking without compromising what are called production values.  The authentic quality of the AI OR DIE clips I saw would fool anybody not thoroughly familiar with the technology into thinking those were real people and real mobile homes.

 

But AI is in the same category as film, cameras, lights, microphones, technicians, and all the other paraphernalia we traditionally associate with film.  These are all means to an end.  And the end is what Kreeft said:  the cultivation of beauty. 

 

I think the biggest change that the use of AI in film and animation is going to make will be economic.  Just as the advent of phototypesetting obsoleted entire technology sectors (platemaking, Linotyping, etc.), the advent of AI in film is going to obsolete a lot of technical jobs associated with real actors standing in front of real scenery and being photographed with real cameras. 

 

Will we still have movie stars?  Well, is Bugs Bunny a movie star?  You can't get his autograph, but nobody would deny he's famous.  And he's just as alive as he ever was. 

 

Before we push the panic button and write off most jobs in Hollywood, bear in mind that live theater survived the advents of radio, film, and television.  It was no longer something you could find in small towns every week, but it survived in some form.  I think film production with real actors in front of cameras will survive in some form too.  But the economic pressures to use AI for more chunks of major-studio-produced films will be so immense that some companies won't be able to resist.  And if the creatives come up with a way to make a film that cultivates beauty, and also uses mostly AI-generated images and sounds, well, that's the way art works.  Artists use whatever medium comes to hand to cultivate beauty.  But it's beauty that must be cultivated, not profits or gee-whiz dirty jokes.  And unfortunately, the dirty jokes and the profits often win out.

 

Sources:  Jonathan Rothman's "After the Algorithm" appeared on pp. 31-39 of the Sept. 1-8, ,2025 issue of The New Yorker.  I also referred to Wikipedia articles on "Bambi Meets Godzilla," and the software company Runway.  Peter Kreeft's ideas of art as the cultivation of beauty can be found in his Doors In the Walls of the World (Ignatius Press, 2018).