Monday, June 26, 2023

An AI Church Service: Whom Will You Serve?


Every couple of years, the Evangelical Church in Germany, one of the world's largest national Protestant church federations, holds a convention in a different location.  This year, it was in Fuerth, outside of Nuremberg in Bavaria, and the event that attracted more media attention than anything else was a church service produced almost entirely by the AI chatbot ChatGPT.  The service was the brainchild of 29-year-old Jonas Simmerlein, who is identified as a "theologian and philosopher," not a computer expert.  This may explain some of the complaints heard from the congregation about the rather wooden delivery style of the four avatars—two women and two men—who delivered the prayers and the sermon and selected the music. 


Some of the younger attendees among the 300 or so present said on the other hand that the service was surprisingly good, even though "a bit bumpy at times."  The service was not interactive in any real sense, so when the congregation found some of the avatar's words inadvertently humorous, as when it said with a deadpan expression, "to keep our faith, we must pray and go to church regularly," it kept right on going despite the laughter. 


To be fair, some of the problems people noted could be fixed by more resources.  Hollywood studios have access to animation techniques that reliably pull emotions from millions, so the expressionless delivery style is not a fundamental issue, but simply reflects Simmerlein's inexperience with the process.  But suppose the avatars had been much more emotionally engaging, and the whole service felt like real people were up there talking to you?  Would that have been better?  It all depends on what you think a church service is for.


If you view attending church as just one of a number of ways to spend your time searching for happiness, and if you judge a service by how happy it makes you, then the emotional buzz it produces is all-important.  By this criterion, going to your favorite band's rock concert is probably going to do you more good than any number of church services. 


But if you view church attendance as intentionally placing yourself in the presence of a supernatural God, the picture changes.  If a human being is not running the service, some branches of Christianity regard it as worthless, because only a human priest can do certain things that have to be done for a mass to be a mass, for example.  In the Catholic mass, a duly ordained priest must consecrate the bread and wine to change it into Christ's body and blood.  No AI system can do that.  And when the members of the congregation take the elements, as they are called, the priest is acting as Christ for them, which is why he has to be male. 


This odd little AI experiment highlights an issue that we are going to face more and more as time goes on, and as so-called generative AI develops more abilities that mimic human ones.  There are all kinds of work for human beings in the world, and none of them, perhaps, is entirely without some negative aspect.  Even Michelangelo got a tired back while working on the Sistine Chapel.  But if he had been able to call upon some wizard to do the work for him, would he have done so?  I seriously doubt it.  An artist uses tools, but the quality of the art depends not primarily on the tools but on how well the artist uses them and transmits the vision in his or her mind to canvas, paper, or sound.  Art, or a church service, produced almost entirely by a machine is going to be missing the essential ingredient that makes it a human achievement. 


But not everybody will miss that ingredient.  In the dystopian novel Nineteen Eighty-Four, if my memory serves, one of the characters made a living by operating a "pornography machine" which cranked out dirty books for consumption by the masses.  The work was entirely physical, like running a printing press, and involved no mental effort at all.  Despite the completely mechanical source of the books, they proved to be very popular, as pornography usually does.


The vast traffic in pornography on the Internet, except for the fact that it is digital instead of mechanical, fulfills George Orwell's prophecy of 1949, and shows what can happen when modern technology is twisted to deliver what some people consider happiness.  The connection between Internet pornography and church services is not obvious, but if you place them both on a spectrum of things designed solely to make the user happy, the question is simply how effective each one is in achieving the goal of happiness.  Some people may like church services run by AI-generated avatars, and others may like AI-produced Internet porn—it just becomes a matter of taste.


We have Jonas Simmerlein to thank for creating an extreme example of the wrong kind of church service.  The right kind is truly a service, but not primarily to the individuals involved.  If God truly is the Creator of the universe and everything good in it, we owe Him everything, in principle, and an hour or so of devotion and worship every week seems like only a pittance.  If there is some way (right now I can't imagine how, but maybe there is some way) to incorporate AI into a worship service that is truly and obviously devoted to God, maybe that can happen.  But turning the whole thing over to a machine is no different than showing a video of a recorded service—in fact it's a good deal worse, because a video of a live service at least shows real people doing and saying things that have a chance of pleasing God.


The Fuerth ChatGPT church service is a first, all right, but it may turn out to be one of those things that we have to try in order to find out that this is not really the direction we want to go.  Of course, there will always be an audience for such things.  In fact, one of the fears of AI developers is that AI will come up with false religions to deceive the masses into doing things like Jim Jones did with his Peoples Temple cult, from which we get the grim phrase "drink the Kool-Aid."  Let's hope that most of us, anyway, have more sense than that, and that AI developers put safeguards in place to make such things impossible.


Sources:  The Associated Press carried the article "Can a chatbot preach a good sermon?" on June 10, 2023 at 

Monday, June 19, 2023

The Davenport Apartment Collapse


Around 5 PM on Sunday evening, May 28, if you had been standing near the corner of West 4th and Main Streets in downtown Davenport, Iowa, you would have seen The Davenport apartment building yield up about a third of its facade to gravity.  With a loud rumble and a cloud of dust, more than a dozen apartments in the building collapsed, taking three lives and trapping Quanishia Berry in her fourth-story apartment for over twenty-four hours before she was rescued.  Like many engineering disasters, this one was the outcome of a series of actions, or inactions, that could have prevented it.  But now that it's too late for the victims, we can at least learn something from the sequence of events that preceded the tragedy.


The Davenport was built in 1907 and at six stories, was probably one of the most prominent structures in the city at that time.  The residents living in it just before the collapse complained of odd creaking noises, cracks in the floor underneath carpets, and window frames pulling away from walls. 


The Davenport's owner, Andrew Wold, paid a structural engineer in February to conduct an inspection of the building.  The engineer's report stated that although bricks in the facade were cracked, the flaws were not an "imminent threat to the building or its residents."  In later reports, the last one dated four days before the collapse, the same engineer said that now, several patches of bricks "appear ready to fall imminently, which may create a safety hazard to cars or passersby."  Evidently, a portion of a load-bearing wall behind the brick facade had collapsed, transferring loads onto the facade itself, which was not intended to carry such loads.  This would explain the bulging and other issues noted by the residents.


Some residents became so alarmed by the deterioration that they broke their lease and moved out before the collapse.  But others took the word of the landlord that there was no immediate danger. 


It is always a judgment call to determine when a building's condition is so bad that it must be evacuated.  Wold had obtained estimates for shoring and supporting parts of the building, but at $50,000 he regarded them as too costly.  The owner of the masonry repair service who provided the quote said that he wouldn't allow his workers on the site without such repairs.  So here was a technically competent person saying he wouldn't even let his employees near a place where residents were living twenty-four hours a day. 


How could things have gone differently?  There is evidence that Trishna Pradhan, the city's chief building official, thinks she might have been able to prevent the tragedy by issuing an evacuation order, something she was presumably authorized to do.  The evidence is that a few days after the collapse, she resigned her position.  Another official gave the reason for her resignation as having "wrongly categorized an inspection of the building."  While orders to evacuate a structure immediately are serious things and involve potential legal liability, failing to issue such an order can have worse consequences, as it did in this case. 


A building is an extremely complicated structure, and predicting down to the minute when something as large as The Davenport will collapse is essentially impossible.  But when large cracks appear in significant places and the facade begins to bulge visibly, even the average non-technical person would agree that something bad may be about to happen. 


One article on the collapse interviewed engineers who cite modern technologies such as vibration sensors that can detect abnormal shifts and movements that occur before serious structural failures.  These can be installed in suspect structures and monitored, but then the question becomes, "How much movement is too much?"  These methods are still experimental, and the expense of installing and monitoring them is one more reason to simply leave things as they are.


And that was the main sin of omission committed here.  No one wanted to be the person who stuck his or her neck out and said, "Okay, this place is about to fall down, and we need to get the residents out before it does."  Only a few occupants made this decision on their own, which is not to blame the ones who stayed.  After all, they were told that there was no imminent danger, and up to 5 PM on May 28, that statement was correct, depending on how you define "imminent." 


That word did come up in the engineering report four days before the collapse, and should have motivated someone—the landlord, the city—to act.  But yet another decision point was passed, and events moved on to their tragic conclusion.


Whoever takes over from Trishna Pradhan as chief building official is going to be a lot more careful about classifying the results of building inspections in the future, one would hope.  And as the technology for sensing vibration in buildings gets cheaper and more available, it might be a good thing if buildings with structural problems were required to have such sensors installed.  The most valuable thing about that would be that we could get lots of data from normal wear and tear, and could get a better idea about what happens with regard to vibrations as a building gets closer to falling down.  Electronic infrastructure sensors like these are available for new buildings, but it would be a challenge to install them in old structures, although not impossible.


I can't close this blog without remarking on the likely class of residents who occupied The Davenport in its final days, and whether their complaints might have received more attention if they had been, frankly, richer.  But rich people don't live in 116-year-old run-down buildings in the middle of deteriorated city centers.  Just this morning I was reading in the Old Testament book of Proverbs, and came across one that seems to sum up what happened here:  "The poor man utters supplications, but the rich man answers roughly."  Next time, the rich man needs to listen more carefully, and take more thought for the safety of his fellow human beings.


Sources:  I referred to the AP article "Many warnings before collapse" at and the Wikipedia article "Davenport apartment collapse."  The Proverbs verse is Prov. 18:23 from the New American Standard Bible.

Monday, June 12, 2023

Prudence, Justice, Democracy, and Technocracy


In the June 12 issue of National Review, writer Noah Rothman takes aim at the web of intrusive regulations that, as the Lilliputians tried to bind Gulliver with hundreds of strings, try to micromanage our personal lives with regard to the kinds of technologies we use around the house.  In recent years, the following well-established items have become targets for government regulation above and beyond the usual safety concerns:  incandescent light bulbs, gas stoves, gas water heaters, air conditioners, lawn mowers, and even paper bags and plastic straws. 


Any attempt to elucidate a coherent motivation behind these regulations, which emerge from all levels of government (city, state, and federal), tends to end in confusion.  Various reasons are given ranging from suspicious increases in childhood asthma for gas stoves, to air and noise pollution from gasoline-powered lawn mowers, to alleged environmental harms from plastic straws.  None of the regulators claim that any of these technologies are bad because they are directly responsible for the deaths or injuries of those who use them (although people have died in fires caused by gas stoves and cut off their fingers with lawnmowers).  The harm is always more remote:  vague and poorly-implemented studies showing correlation between asthma and the use of gas ranges, increased levels of environmental pollution that are so small as to be negligible, or the ever-present excuse of climate change. 


As Rothman points out, neglected in all these calculations by the regulators are the inconveniences, or worse, placed on the consumer by the byproducts of regulation.  LED lighting costs more than old-fashioned incandescent lighting.  This is a clear burden on poor people.  Some kinds of cooking are hard or impossible to do with electric or induction ranges, which also cost more to operate in many locales than gas ranges do.  In the name of various vague political causes, then, the regulators ask the much larger number of ordinary citizens to endure a set of individually slight inconveniences, which however add up to a considerable burden if totaled over the entire U. S. 


If we look at this situation from an ethical point of view, two cardinal virtues come to mind:  justice and prudence.  Justice sees that each person gets what is coming to him or her, in either a nice way or a not-so-nice way.  Prudence is a sort of all-around virtue of balancing various factors so that justice is served.  In a democracy, both of these virtues should be possessed by the body politic so that whatever is democratically enacted embodies them. 


Before the era of micro-managing regulations, all of the technologies mentioned above were adopted not because they were politically correct, but because enterprising inventors and manufacturers developed them in response to needs and wants of the public.  And the public went for them without any help from government.  In the name of prudence, certain basic safety regulations and voluntary schemes such as the Underwriters Laboratory label were put in place to ensure that direct harm from these technologies was kept to an acceptably low level, but that was it as far as regulation is concerned.


However, there is another philosophy of government that is more or less opposed to democracy, and that is technocracy:  rule by experts.  In this view, the great unwashed masses are too ignorant to choose domestic technologies wisely, and need to be guided by their elitely-educated masters into the path of enlightenment.  Exactly which path that is depends on the expert, so in practice the regulation that is enacted depends on just who grabs the levers of power in the increasingly complex web of bureaucracies that we live under today.  This probably explains the geographically diverse nature of these regulations:  what's fine in Nevada may be banned in California and vice versa, depending on which technocrat is running the show. 


Both prudence and justice are violated when an ideologue imposes his idiosyncratic whims on a large group of citizens who have been minding their own business.  Whatever allegedly meritorious political goal is served by the regulation is usually far outweighed by the sum of small (or not so small) inconveniences suffered by those who are regulated.  Voluntary suffering and privation, as many citizens did to sacrifice consumer products during World War II, is one thing.  But to have it imposed on you in the absence of a national emergency is both unwise (imprudent) and unfair (unjust).


Rothman ends his article in the hope that voters will turn to the ballot box to alleviate some of the ills imposed by bureaucratic micro-regulation of technologies which have proved both popular and safe, by and large.  Unfortunately, the same diffuseness that marks the regulations (small harms caused to a lot of people) means that such matters are unlikely to become dominant issues in political campaigns.  I doubt that even the most progressive candidates have made banning gas stoves a major plank in their platforms. 


The most we can hope for in this regard is that one party or the other will take up the cause of pushing back against rule by technocrats in general, and a vote for that party will work against the creeping regulations that threaten to deprive us of things that work just fine, thank you.  While this might work on the national level, the entrenched nature of local politics, especially in large cities and certain states, means that there are only limited hopes in this direction. 


Sometimes other types of public-relations actions can have effects all out of proportion to their actual size.  I think of Mahatma Gandhi's famous Salt March of 1930, which publicized the Indian populace's opposition to the British monopoly on salt production.  Salt was something everybody used and had to pay for, and the British unfairly prevented Indian citizens from making it themselves, no doubt using reasons of sanitation and quality to justify their injustice.  After the arrest of some 60,000 protesters, Gandhi was able to negotiate with the British leaders for the right to make salt by ordinary citizens, and the march was a success.


I don't know how effective a lawnmower march on a state capital would be, but I'm sure it would garner some publicity.  When ordinary democratic means either fail or are unavailable, more unorthodox ways should be tried.  But the best thing is to re-establish both a working democratic system, and a populace who will not stand for any nonsense.


Sources:  The article by Noah Rothman "The War on Things That Work" appeared in the June 12, 2023 issue of National Review, pp. 21-24.  I also referred to an article on the Gandhi-Irwin Pact at


Monday, June 05, 2023

Train Crash in India: An Avoidable Tragedy


On Friday evening, June 2, the Coromandel Express carrying about 1250 passengers was traveling through a stretch of track just southwest of Balasore, about 130 miles southwest of Kolkata, India.  A freight train full of heavy iron ore was parked on a siding that paralleled the main line.  Somehow, the express train was signaled to travel past the siding, but the actual switch over which the train passed was set to shunt it into the siding, forcing it to head straight for the stationary freight train.


The express train derailed, throwing about a dozen passenger cars over both the siding and the mainline tracks like a spilled box of toothpicks.  To make things worse, another express train, the Yesvantpur-Howrah Express, arrived before emergency signals could reach it, and crashed into some of the derailed Coromandel cars, leading to further derailings and deaths.


As of today (Sunday June 4), the total fatality count stands at 275, with about 800 injuries. A preliminary investigation has pinned the cause to faulty signaling and switching.  Obviously, if the signals had matched the position of the switches, the Coromandel Express would have been ordered to stop.  But somehow, perhaps in a last-minute error, someone or something changed the switch from the main line to the siding, leading to the disaster.


Though the deaths of 275 people are horrific, this is not the worst train disaster in India's history.  In August of 1995, two trains collided near New Delhi and 358 people died.  This leads to the question of whether an Indian passenger is taking his life in his hands anytime he travels by rail.  Surprisingly, the answer is "not necessarily."


There are several ways to measure the relative hazards of travel in different modes and locations.  From an individual's point of view, the number that is most meaningful is probably the passenger-miles traveled for each fatality.  The farther you can go without dying, the better.  Surprisingly, by this measure it is about twice as dangerous to travel by rail in the U. S. as it is to travel by rail in India.  The fatality rate in the U. S. is one for every 3.4 billion passenger-miles, and in India it is one for every 6.6 billion.  Yet when even one passenger gets killed in a train wreck here, it's big news.


The difference can be partially explained by the fact that intercity passenger travel by rail in the U. S. is something only a few people do, whereas most of the middle and poorer classes in India travel by rail.  The passenger-miles traveled in India every year is about 35 times that in the U. S.  So even though the absolute number of train-related passenger fatalities in the U. S. is much smaller, we take the train even less than they do in India, and so our ratio of deaths to passenger-miles is worse.


Another difference is the density of people per train.  I haven't taken a ride on Amtrak lately, or a regional line such as the Long Island Railroad, but it would surprise me if you could find a regularly scheduled train anywhere in the U. S. that carries 1200 people at a time.  The more people on a single train, the more people can be hurt or killed if that train gets into trouble.


For a long time, railroads have been trying to devise foolproof systems of track switching.  The so-called interlocking switches were originally devised in the 1850s.  The idea was to guarantee that no combination of switches could be set that would lead to a collision.  At first the systems were entirely mechanical, but by the 1890s electric-motor-activated switches were introduced, leading to electromechanical and later entirely electronic systems.


But even the best-designed systems can fail.  Unless the trains themselves are controlled or monitored remotely (an element of what in the U. S. is called positive train control), a determined or confused individual may change a switch at the wrong moment, for example, just as an express train is heading toward the switch that leads to a siding.  While this will automatically change the signals ahead of the switch, if the train has passed all the signals already, it's way too late to do anything about it. 


Something like this may have happened in the Coromandel Express tragedy.  Once the relevant parties are interviewed and records examined, a clearer picture of the cause should emerge.  But at this point, that is small solace for the thousands of people who have lost relatives or loved ones in an accident that did not have to happen. 


The tolerance for violent technology-related death varies from culture to culture.  Here in the U. S., we have more or less passively accepted a rise in auto-related deaths over the last few years, despite numerous technological improvements such as safety belts and air bags that make accidents survivable which formerly would have been surefire killers.  In 2020 and 2021, the fatality rate increased about 17%, partly no doubt due to COVID-19.  But in 2022 the rate stayed about the same, which seems to indicate we changed something for the worse during the pandemic and haven't changed back.  Whether the reason is increased drug and alcohol use, more reckless driving by depressed individuals, or some combination of factors, an extra 4,000 or so deaths per year has not really seemed to concern anybody, except when it's you or someone you love.


At any rate, I hope that this latest train accident will lead to further improvements in the already comparatively good safety record of Indian railways.  Considering the huge numbers of people they transport every day, they are doing a better job by some measures than we are doing in the U. S.  But there is always room for improvement, and as technology and organizational integrity improve, maybe this can be the last major train wreck in India for a long time.


Sources:  I referred to an NBC News article on the wreck that appeared at  I also referred to statistics on comparative train-passenger fatality rates worldwide at and the Wikipedia article "Interlocking" and