Attack of the goblins

Don't worry, the techbros tell us.  AI is gonna be awesome.  Look at how much it can do, how fast it's improving.  At the same time, the scoffers tell us that okay, maybe it's pretty impressive, but keep in mind it's just completely deterministic software, and its failings happen mainly because it's been trained on questionable input.  You hear the phrase "garbage in, garbage out" pretty frequently, usually along with a shrug of the shoulders.

What both of these viewpoints fail to acknowledge is how freakin' weird some of the AI/LLMs have been acting.  We had a guy named Matt Schlicht creating an AI-only social media site called "Moltbook" -- in 72 hours, over a million AI accounts had joined, and within a week they'd created their own religion, called the "Church of Molt."  Just a few weeks ago, a company called Just Like Me launched an AI Jesus, trained on the King James Bible no less, whom you can have a chat with for $1.99 a minute.  AI Jesus is a vaguely Middle Eastern-looking guy with long tousled hair and a gentle smile, and although the company says in the fine print that this "is not Jesus Christ himself, and does not possess divine authority," I can only imagine the effect such a conversation might have on someone who is already a true believer.

But if you think this is as weird as it gets, I got news for you.

Open AI's ChatGPT -- one of the most widely used AI/LLMs in the world -- has suddenly developed a strange affinity for...

... I swear I'm not making this up...

... goblins.

The problem apparently started last November, when there was a strange uptick in ChatGPT's use of words like goblin, gremlin, and troll.  At first, it just seemed like a quirk, and that perhaps it was happening because people heard from friends that they'd been getting responses involving mythological humanoids, so they put in prompts themselves that generated such output.  But it quickly became obvious that this was more than a blip caused by the users' own prompts.  By the time ChatGPT 5.4 came out two months ago, the use of gremlin had risen by 52%, and goblin by 175%.  And it continued to accelerate.

"The problem evolved from a minor quirk into a persistent 'verbal tic' that was showing up in almost every user conversation," said a company spokesperson.

The origin of the problem, Open AI claims, is that one of the settings you can use when you set up your AI's profile is a persona called "Nerdy," which is described as follows:

You are an unapologetically nerdy, playful and wise AI mentor to a human.  You are passionately enthusiastic about promoting truth, knowledge, philosophy, the scientific method, and critical thinking. […]  You must undercut pretension through playful use of language.  The world is complex and strange, and its strangeness must be acknowledged, analyzed, and enjoyed.  Tackle weighty subjects without falling into the trap of self-seriousness.

Apparently during training, humans who were reviewing the "Nerdy" AI responses to prompts and questions unconsciously rewarded them for using language involving mythological references, such as calling a dangerous place "a troll's lair."  This created a feedback loop in which such "quirky" responses multiplied.  Then, because there's no hard barrier between the different styles of AI personalities -- they're often trained from the same datasets -- the goblins jumped into other personas, and soon, they were everywhere.

The coders started frantically trying to find a solution, and they found the only thing that worked was a top-down, brute force command: “Never talk about goblins, gremlins, raccoons, trolls, ogres, pigeons, or other animals or creatures unless it is absolutely and unambiguously relevant to the user’s query.”

I'm not sure how the raccoons and pigeons got involved, but may as well cover your bases, I guess.

Anyhow, this has more or less fixed the immediate problems, except for two things -- one funny and one not so funny.

The funny one is that of course the conspiracy theorists think that goblins are showing up in ChatGPT because there actually are goblins -- well, electronic ones -- within the internet, and they're trying to get out.  Or warn us of something.  Or steal our souls.  Or destroy the world.  You know the drill; something odd happens, and the modern Chickens Little start running around claiming the sky is falling.  Then, it turns out there's a completely prosaic explanation for what happened, and they calm down for exactly 5.8 milliseconds until they find the next odd thing that's trying to kill us all, and they move on to that, ad infinitum.

The less funny thing is that this kind of unexpected response is an indicator that we really don't have any good way of predicting what AI is going to do.  Okay, yeah, it's running on a purely deterministic machine, but that doesn't mean that its behavior is going to be predictable; hell, the human brain is (at least according to a lot of neuroscientists) a deterministic machine, and we humans can be mighty fucking unpredictable at times.  There's this thing called an emergent property -- something that comes out of the interactions of the parts of the whole, and could not even in principle be predicted from watching the behavior of the pieces in a purely reductionistic fashion -- and AI is looking like it's going to have some emergent properties that will take some monitoring.

We lucked out this time that the "goblins" thing turned out to be more of a quirky nuisance than a real problem, but there's no guarantee that'll be the case next time.  Consider, for example, that just a couple of days ago, an AI being used to handle a routine task for a software company called PocketOS deleted the company's entire database within seconds of being activated -- bypassing all of the security firewalls by using a programming token key no one at the company even knew existed.

And our government "leaders" are gung-ho about turning over our nation's fiscal management and defense to AI-based systems.  Hello?

What the fuck are you people thinking?

Anyhow, like I've said before, so often I'm getting tired of hearing it my own self, we need to put the brakes on AI.  Like, now.  Not that anyone's going to.  The "Oh, it won't happen to me" thing is just too strong.

I'm just hoping the next round of goblins that are released don't have their fingers on any literal triggers.

****************************************

Cracker crumbs

Three years ago, I wrote here at Skeptophilia about the scary Cascadia Subduction Zone, which is capable of enormous earthquakes and tsunamis -- and which, unfortunately, lies right off the coast of British Columbia, Washington, and Oregon.  A subduction zone is a region along which two plates are coming together, forcing one underneath the other.  Because rocks experience a high degree of friction, the two plates often get stuck, sometimes for centuries, and then can give suddenly.  This lurch is what causes big earthquakes.

The motive forces here are convection and drag.  Rising plumes of magma underneath ridges diverge, and the friction between the magma plume and the underside of the plates forces them apart.  Where the leading edge of the plate strikes another, something's got to give.  In this case, the oceanic Juan de Fuca Plate, made of (relatively) thin, brittle basaltic rock, hits the old, thicker and colder North American Plate.  The Juan de Fuca Plate jams up and eventually plunges underneath.  The downward drag produces a trench, and inland from the trench you often find volcanoes, created as the subducted plate melts and the molten rock pushes its way to the surface.  (This is how the Cascade Volcanoes, most famously Mount Rainier, Mount Shasta, Mount Hood, and Mount Saint Helens, formed.)

The red dots are undersea earthquakes; the green ones, on-land earthquakes.  [Image is in the Public Domain courtesy of the United States Geological Survey]

What hasn't been clear until now is how exactly subduction happens.  We know that the process usually isn't smooth (as I described, it often goes by fits and starts rather than releasing the compressional force gradually).  But what happens to the plate itself as it descends and is destroyed in the upper mantle?

Thanks to a new study out of Louisiana State University, we now have our first good picture of how this process occurs.

It turns out that the destruction of the last piece of a plate, such as Juan de Fuca -- which is one of the only remaining fragments of the Farallon Plate, that once underlay most of the northeastern Pacific Ocean -- is anything but orderly.  The (relatively) small slab of solid rock beneath the ocean off the coast of the Pacific Northwest is being bent as its eastern edge is pulled downward, creating multiple fractures and dozens of "microplates."  "Getting a subduction zone started is like trying to push a train uphill -- it takes a huge effort," said geologist Brandon Shuck, lead author of the study, which appeared in Science Advances. "But once it's moving, it's like the train is racing downhill, impossible to stop.  Ending it requires something dramatic -- basically, a train wreck...  This is the first time we have a clear picture of a subduction zone caught in the act of dying.  Rather than shutting down all at once, the plate is ripping apart piece by piece, creating smaller microplates and new boundaries.  So instead of a big train wreck, it's like watching a train slowly derail, one car at a time."

Which, if you think about it, makes sense.  Picture shoving together two saltine crackers.  One will likely push underneath the other, but the leading edges are going to crumble, and what you'll be left with will probably be a disordered pile of cracker crumbs.

This process doesn't really change the picture with regards to earthquake risk; just because the plate is shattering into smaller chunks doesn't mean the effects will be small when the breaks occur.  One example -- the Shuck et al. research found a major, 75-kilometer long fault where pieces of it have dropped by five kilometers.  The scary part is that despite the fault collapse, it's not done separating.  "This is a very large fault that's actively breaking the [subducting] plate," Shuck said.  "It's not one hundred percent torn off yet, but it's close."

Further reinforcing my assessment that while I dearly love the Pacific Northwest for some of the most beautiful scenery in the world and the absolute best gardening climate in the United States, I'd never live there again.

It bears mention, however, that it may be that the fault won't rupture for another two hundred years; on the other hand, it could happen tomorrow.  While our ability to analyze plate tectonics is light years beyond what it was even thirty years ago, when the situation in the Northwest first began to come clear, we still don't have any way to determine when the earthquake will happen with any kind of precision.  At the moment, all we know is that it will rupture, sooner or later.

And I don't want to be anywhere near it when it does.

****************************************

Tense situation

In my Critical Thinking classes, I did a unit on statistics and data, and how you tell if a measurement is worth paying attention to.  One of the first things to consider, I told them, is whether a particular piece of data is accurate or merely precise -- two words that in common parlance are used interchangeably.

In science, though, they don't mean the same thing.  A piece of equipment is said to be precise if it gives you close to the same value every time.  Accuracy is a higher standard; data are accurate if the values are not only close to each other when measured with the same equipment, but agree with data taken independently, using a different device or a different method.

A simple example is that if my bathroom scale tells me every day for a month that my mass is (to within one kilogram either way) 239 kilograms, it's highly precise, but very inaccurate.

This is why scientists always look for independent corroboration of their data.  It's not enough to keep getting the same numbers over and over; you've got to be certain those numbers actually reflect reality.

This all comes up because of an exciting new approach to one of the most vexing scientific questions known -- the rate of expansion of the entire universe.

[Image is in the Public Domain, courtesy of NASA]

A while back, I wrote about some experiments that were allowing physicists to home in on the Hubble constant, a quantity that is a measure of how fast everything in the universe is flying apart.  And initially, the news appeared to be good; from a range of between 50 and 500, physicists had been able to narrow down the value of the Hubble constant to between 65.3 and 75.6.

The problem is, nobody's been able to get closer than that -- and in fact, recent measurements have widened, not narrowed, the gap.

There are two main ways to measure the Hubble constant.  The first is to use information from Type 1A supernovae (whose brightening and eventual dimming curves are connected to their intrinsic brightness) and Cepheid variables (stars whose period of brightness oscillation varies predictably with their luminosity); these properties make them good "standard candles" to determine the distance to other galaxies.  Once you know a star's intrinsic luminosity, you can use that to determine how far away it is -- just as you can estimate your distance to an oncoming motorcycle at night because you know how bright a motorcycle's headlight actually is.  This, coupled with the galaxy's redshift, allows you to figure out how fast the galaxies we see are receding from each other, and thus, how fast space is expanding. 

The other method is to use the cosmic microwave background radiation -- the leftovers from the radiation produced by the Big Bang -- to determine the age of the universe, and therefore, how much bigger it's gotten since then.  The problem with this method is that it relies heavily on the correctness of our current models of the evolution of the universe, some of which have resulted in predictions not matched by the available observations.

Here's the issue: not only does each of the methods -- standard candles/cosmic ladder, and the CMBR method -- each have its difficulties, the measurement of the Hubble constant by these two methods has resulted in two irreconcilably different values.

So the astrophysicists have tried to narrow in from both ends.  Improve the data, and improve the models.  This backfired.  As our measurement ability has become more and more precise, the error bars associated with data collection have shrunk considerably; at the same time, the models have improved dramatically.  You'd think this would result in the two values getting closer and closer together.

Exactly the opposite has happened.

This result, called the Hubble tension, is considered to be one of the most frustrating problems in astrophysics.  And it's not just some fringe-y side quest; this is a fundamental issue with our understanding of the entire universe.

Here's where the new research, out of the Technical University of Münich, comes in.  You probably know about the phenomenon of gravitational lensing, where light traveling through the curved space near a massive object (like a galaxy or a supermassive black hole) gets bent, in much the same fashion as light going through a glass lens.  Sometimes this causes distant bright objects to look like they're stretched, or even multiplied.  For these objects, there is more than one pathway the light can take through space to get here to us, so the image we see is distorted.

Well, we've just detected one of the most remarkable examples of gravitational lensing ever observed; a supernova in a brilliant galaxy whose light split up into five separate paths in order to get here.

Put a different way, we saw the same supernova occur five different times.

Now, here's the kicker: because the paths that each of those beams of light took to get here differ in distance, comparing the timing of arrival of each image could give us the first-ever direct, no-assumptions-required method of measuring the Hubble constant, one with far fewer systematic uncertainties.

"We nicknamed this supernova SN Winny, inspired by its official designation SN 2025wny," said astrophysicist Sherry Suyu, who co-wrote the paper on the discovery.  "It is an extremely rare event that could play a key role in improving our understanding of the cosmos.  The chance of finding a superluminous supernova perfectly aligned with a suitable gravitational lens is lower than one in a million.  We spent six years searching for such an event by compiling a list of promising gravitational lenses, and in August 2025, SN Winny matched exactly with one of them."

In-depth analysis of the timing and positions of the five supernova appearances is currently underway.

Whether this will resolve the Hubble tension, of course, remains to be seen.  The worst-case scenario is that the SN Winny data doesn't agree with either the cosmic ladder value or the CMBR value, or has error bars large enough to overlap with both.  A happier outcome would be a decisive landing in one camp or the other -- although that'd still leave the astrophysicists puzzling over why the losing method doesn't work.

But it's an incredible discovery, and I know I'll be watching the science news to see what comes out of it.  Settling the Hubble tension question would be an amazing coup; having it resolved because of a one-in-a-million observation of a lensed supernova -- well, if you don't find that super cool, I don't even know what to say to you.

****************************************

Heap of trouble

In my AP Biology class, we did a lab that involved extracting chlorophyll from spinach leaves.  The first step was to grind the leaves into a paste with a bit of solvent, which we did the old-fashioned way using a mortar and pestle.

The instructions said to add a "small amount of fine sand" to the leaves (to act as an abrasive, facilitating the breakup of the tough cell walls), and one of my students -- a little on the tightly-wound side, as I recall -- asked how much to add.

"Doesn't matter," I said.  "Some.  A pinch.  You're going to filter it out at the end anyhow."

This didn't satisfy her.  Everything else was measured to high accuracy, so the sand should be, too.  How much sand was "some"?

I grabbed a pinch of sand between my thumb and index finger and tossed it into the mortar.

"There you go," I said.  "All fixed."

She gave me the suspicious side-eye, as if by my insouciance I had ruined her chance of getting good results.  As I recall, she did just fine on the lab, but I don't think she quite trusted my lab technique afterward.

A couple of amused students who overheard the conversation got into a discussion about the imprecision of measuring-words in English, and decided to fix matters by constructing a list:

2: a couple
3: a few
4: some
5: a bunch
6: a lot
7: quite a lot
8: a helluva lot
etc.

I recall that they got up to 20, which was "a shitload."

After showing the list to me, they did admit that these designations could shift depending on what you're talking about.

What I didn't realize until recently was that this discussion, as lighthearted (honestly, ridiculous) as it was, touched on a paradox that has been around for at least 2,400 years -- the Sorites Paradox.  The name comes from the Greek word σωρός, meaning "heap," and is attributed to the fourth century B.C.E. philosopher Eubulides of Miletus, who is said to have formulated it.  It goes something like this:

Let's say you have a million grains of sand in a pile, sitting on the left side of the table.  Nearly everyone would agree that this constitutes a "heap of sand."  On the right side of the table, you have a single grain of sand.  No one, I suspect, would say that one grain of sand is a heap.  Okay, so that means that if you remove one grain at a time from the left-hand side, at some point it changes from "a heap" to "not-a-heap."

When does that happen?

It's not just the word heap that has this problem.  Take away one teaspoonful of water at a time from the ocean, and at some point -- admittedly, it'd take a while -- what's left would no longer be an "ocean."  When does that change happen?

How about old?  Along the pathway of life, I think we can all agree that a fifteen-year-old is "young," and a ninety-year-old is "old."  So, when do things flip?

I'm currently sixty-five, and I will not admit to being old, so anyone inclined to answer should keep that in mind.

This also relates to another famous paradox, the Ship of Theseus.  If you take Theseus's ship and replace, one at a time, each of the components that make it up, at what point does it cease to be the original ship?

One solution to the Sorites Paradox is simply to declare these things a continuum, which therefore renders such questions essentially meaningless.  The problem is, the number of grains of sand in a heap isn't a continuum; it's necessarily an integer (you can't have a heap made of 1,827,793-and-a-half grains of sand).  Neither, for that matter, are the pieces of a ship.  So while this might be a reasonable response in cases of true continua (such as age, water volume, or the colors of light in a rainbow), it doesn't work in systems with discrete states.

So maybe it's just unanswerable, and relies simply on usage -- language is inherently vague, and there's nothing to be done.  This is the stance of British philosopher Timothy Williamson and others, who solve the Sorites Paradox by shrugging their shoulders; there is a point where a heap becomes not-a-heap, but where the point lies is unknowable.

While all this might seem like nothing more than philosophical noodling, it has its serious applications.  The question of when depictions of sex in movies cross the line into obscenity or pornography (therefore suggesting that they should be subject to censorship) made it all the way to the Supreme Court, which ruled that the movie in question -- Louis Malle's The Lovers -- wasn't pornography, even if the court couldn't come up with a good definition of where the line was.  "I could never succeed in intelligibly [defining pornography]," Justice Potter Stewart famously said, "but I know it when I see it."

The Sorites Paradox also has a strange connection to evolutionary biology, and one that knocks a neat hole into the creationists' assertion that every species represents a "kind" that is in some sort of hard-and-fast, unchangeable box.  The issue is with ring species, of which there have been several described (two well-studied ones are circumpolar populations of gulls of the genus Larus, and populations of the Greenish Warbler around the Himalayas).  In a ring species, adjacent, similar-but-distinct groups can interbreed, and thus by definition should belong to the same species.  The problem is, the ends of the ring have diverged enough that where they do overlap, they no longer can interbreed, and thus should be separate species.  But where do you draw the line?  No matter where you do, you end up separating individuals that (by the canonical definition of the word species) should belong together.

Or -- to take Williamson's approach -- maybe the problem is trying to force a fuzzy reality to conform to limited, inaccurate use of language, and the word species is simply kind of a mess.  This is my opinion on the matter; I tend to agree with my evolutionary biology professor, who memorably said, "The only reason we came up with the concept of species is that we have no near relatives."

A ring species of salamanders in California [Image licensed under the Creative Commons Thomas J. Devitt, Stuart J.E. Baird and Craig Moritz, Ensatina eschscholtzii ring species, CC BY 2.0]

Anyhow, that's today's consideration of a philosophical problem that has been around for over two thousand years, and thus is clearly above my pay grade to weigh in on.  Not that this ever stops me.  Now, y'all'll have to excuse me, because I need to go work in my garden spreading the heap of bark mulch we just had delivered, an amount that is clearly "a shitload."

****************************************

The Estes method

In somewhat the same vein as yesterday's post, which was about the capacity of subsonic standing waves to induce the sensations we often associate with a haunting, today we have: a way to pick paranormal messages out of ambient (and random) noise.

You've probably heard about the idea of electronic voice phenomena, which was popularized as a ghost-hunting method by Latvian paranormal researcher Konstantīns Raudive in the 1970s and has become a standard tool in the kit ever since.  The idea is that you place a recording device of some kind -- it started out with reel-to-reel, then cassette tape recorders, and finally moved on to digital voice recorders -- in an allegedly haunted location, leave it running, and later listen to the recording for any anomalous sounds.  Adepts claim that they hear human voices.

The method was used to great effect in the brilliant Doctor Who episode "Hide," although it turned out that what Clara and the Eleventh Doctor were talking to wasn't a ghost, it was a time-traveler trapped in an alternate universe.  As one does. 

Some of these EVP are more convincing than others, but all of them tend to be muffled and slurred, and to benefit greatly from the phenomenon of suggestion -- once someone tells you that the voice is a ghost saying "I died in 1859" you're much more likely to hear the message.  This is the same thing that occurred with the foolishness surrounding backmasking -- that supposedly, rock bands were including satanic messages in their music that could only be understood consciously if you played the song backwards, but could be somehow picked up subliminally even if you heard it played forwards.  (One of the most popular claims of backmasking involved Led Zeppelin's famous "Stairway to Heaven.")  The problem is, even played backwards, the messages are pretty damn garbled -- but miraculously clear up when you know ahead of time what it's supposed to be saying.

As James Randi put it, "You can't miss it if I tell you what's there."

Graphical plot of white noise waveform [Image licensed under the Creative Commons Omegatron, White-noise, CC BY-SA 3.0]

There's apparently a new way to approach all this that's becoming popular amongst the ghost hunting crowd, and I learned about it from British paranormalist Ashley Knibb's website just yesterday.  It's called the "Estes method," named after Estes Park, Colorado, home of the Stanley Hotel (made famous in The Shining).  The idea here is that a volunteer "receiver" is blindfolded and puts on headphones connected to a radio that's set on "scan" mode, so the only auditory input (s)he gets is blips and fragments of speech or music, interspersed with white noise.  Another volunteer, the "recorder," asks questions -- not of the receiver, but of any ghosts that happen to be present -- while the receiver (who, presumably, can't hear the receiver) reports any interesting phrases heard from the random radio input, which the recorder then writes down.

The claim is that this isolates the receiver; (s)he relies only on any ghosts present to jigger about with the radio and use its audio output to answer what the recorder is asking.

Well, okay.  There are a couple of problems with this, and to his credit, Knibb mentions both of them (although you get the feeling he is still inclined to think that something paranormal may be going on here).

The first is that how the random phrases picked up by the receiver are interpreted afterward is very much dependent upon the subjective opinions of the ones doing the interpretation.  You may recall the famous experiment done by Carl Sagan in a high school class, where he told the students that their birthdates and times had been used to draw up astrological charts and create a personality profile for each of them, and handed out cards with the results.  The students were then asked to rate how accurately it described them, from zero to ten.  Not a single card received a score lower than six; most were between eight and ten.

Wow, astrology vindicated, right?

Not exactly.  Sagan then had the students exchange cards with a neighbor -- and it turned out they'd all been given the same personality profiles.

The point is, when we are given some random piece of text, we're all too likely to interpret it as if it means something -- especially if we walked into the situation already primed to think it does.

The second problem, of course, is exactly the same as what I described in yesterday's post; apophenia, our built-in tendency to find order in random input.  The receiver in the Estes method is trying his/her hardest to listen for anything that sounds meaningful; after all, that's why (s)he's there.  It's not a far step to consider the possibility that the receiver might (even if unconsciously) create something meaningful out of what is, honestly, chaos.

Again, as with yesterday, I'm not accusing anyone of anything underhanded.  Hoaxes aren't even necessary, given how easily our own sensory-perceptive systems can play us false.

So I'm not thinking the Estes method is going to convince anyone who's not already convinced.  As far as the ghost hunters go, no harm if it amuses you, but it still doesn't meet the minimum criterion required for acceptable evidence in a scientific setting.

Me, I'm still in the camp of Andrew MacPhee, the hard-nosed skeptic in C. S. Lewis's novel That Hideous Strength:

"My uncle, Dr. Duncanson," said MacPhee, "whose name may be familiar to you — he was Moderator of the General Assembly over the water, in Scotland — used to say, 'Show it to me in the word of God.'  And then he’d slap the big Bible on the table.  It was a way he had of shutting up people that came to him blathering about religious experiences.  And granting his premises, he was quite right.  I don’t hold his views, Mrs. Studdock, you understand, but I work on the same principles.  If anything wants Andrew MacPhee to believe in its existence, I’ll be obliged if it will present itself in full daylight, with a sufficient number of witnesses present, and not get shy if you hold up a camera or a thermometer."

****************************************

Bad vibrations

A point I've made here more than once is that my doubting many claims of the paranormal isn't because I think it's necessarily impossible, but because our sensory-interpretive systems are so fundamentally flawed.

I mean, they work well enough, for most of us most of the time.  But not only do we have the capacity to miss a great deal of what's going on around us -- as the famous experiment in which a great many test subjects failed to notice a guy in a gorilla suit showed -- what we do sense is all too easy to misinterpret or remember incorrectly.  This is why if someone comes to me with a claim of some supernatural occurrence or another, I'm going to ask for some kind of hard, scientifically-admissible evidence.  To quote astrophysicist Neil deGrasse Tyson, "I need more than 'you saw it.'"  Neither he nor I are accusing anyone of lying or perpetrating a hoax; the problem is that eyewitness testimony is all bad, even if you mean well and are trying your hardest to be honest.

[Image is in the Public Domain]

To throw another monkey wrench into the situation, consider the recent paper by psychologist Rodney Schmaltz of MacEwan University.  Schmaltz became interested in the possible role of subsonic vibrations in claims of haunting; there was a case in England where a medical research building was claimed by several workers to be haunted, in one case by a "gray form that materialized, floated across the room, and vanished."  More than one person saw the apparition, and several described a sensation of chill, as if they were being watched.

The culprit turned out not to be a ghost, but a furnace fan that had set up a subsonic standing wave in the basement.  The frequency of the wave was around twelve Hertz -- so below the range humans can hear -- but created resonant vibrations in our eyes and ears that could be sensed by the brain.  The result: eerie hallucinations, altered perception, and feelings of unease.

What Schmaltz did was try to see if there was a way to measure the human response to infrasound, by setting up test subjects to listen to recordings of music through headphones.  Half the test subjects listened to calm instrumental music, and the other half eerie recordings that could have been the soundtracks of horror movies.  What the subjects didn't know, though, was that half of each of the audio tracks had been altered to include infrasound.

The results were incontrovertible.  The subjects exposed to infrasound weren't aware of it consciously, but responded to it regardless.  Also, it didn't matter what the audible component was.  If they were exposed to infrasound, they reported feeling unsettled and unhappy, and -- most strikingly -- a saliva test showed elevated levels of the stress hormone cortisol.

“Whether they were listening to calming instrumental music or something more unsettling, the infrasound shifted their mood and their stress response in a negative direction,” Schmaltz said.  “In plain terms, you cannot hear infrasound, but your body and your mood appear to respond to it anyway, and the response tends to be unpleasant.”

Schmaltz suggests that a lot of the reports of ghosts in old buildings might be nothing more than infrasound coming from antiquated boilers, furnaces, and plumbing -- aided, of course, by the fact that we're already primed to expect something paranormal from such places by a hundred years of scary movies set in run-down mansions.

Even knowing all of this, though, probably wouldn't make a whit of difference to our actual responses in such a situation.  Because that's the other part of the problem, isn't it?  Our emotional reaction to a particular set of circumstances has a way of derailing our higher brain functions, especially when that reaction is "OMG a ghost, run!"

And unfortunately, that applies not just to those Crazy Kids and Their Stupid Dog, but to skeptical rationalists.

****************************************

The case of the missing scientists

Our capacity for seeing patterns is absolutely critical.

It's easy to see how survival in a risky world could hinge on noticing clues in the environment, then putting them together correctly.  The key, though, is the word correctly.  When you have a built-in mechanism for interpreting sensory cues and recognizing danger, it can easily go awry.  But -- and here's the most important point -- in general, the evolutionary cost of overreacting is almost always way less than that of underreacting.  To use the oft-quoted example, for your average proto-hominin on the African savanna, it's better to hear a rustle in the grass, take fright and bolt, and have it turn out to be a fluffy bunny, than to shrug it off and stay put when it's a hungry lion.  

Or, worse still, not to notice it at all.

This tendency remains with us, lo unto this very day, and it's astonishing how badly it can backfire.  We look around us, assemble the information we have, and all too often put the pieces together wrong -- especially when the pattern that emerges is scary.

And especially especially when we're encouraged to do so by sensationalist media who make their money from clicks, and politicians who benefit by keeping their constituents frightened and distracted.

Take, for example, the current kerfuffle over all the scientists who have allegedly disappeared (or died) recently.  They all, we are told, worked with classified secrets.  The words "mysterious circumstances" have been thrown about.  The implication -- sometimes, the explicit claim -- is that the scientists' fates are all linked, part of a massive conspiracy to silence "people who know too much," or (perhaps) who were about to blow the whistle on some even bigger conspiracy and implicate Important People.

The story hit the news when a retired United States Air Force general, Neil McCasland, left his house in Bernalillo County, New Mexico on foot in late February, and never returned.  A search was conducted, but no trace of him was ever found.  McCasland, they said, was "involved in UFO research," and this was somehow relevant to his disappearance.  Then there's Monica Jacinto Reza, who was director of materials processing at NASA/JPL, who vanished while hiking in the Angeles National Forest in June of 2025.  I'll admit her disappearance was odd; she was an experienced hiker and was accompanied by two other equally able friends, had been seen about ten meters back on the trail and had smiled and waved, but when her friends looked back shortly afterward, Reza was gone.

Once again, searchers came up empty-handed.

Carl Grillmair, an astronomer and exoplanet researcher at Caltech, was murdered in mid-February of this year.  Nuno Loureiro, a plasma physicist at MIT, was shot and killed in December of 2025.  Here are a few others that often get thrown into the mix:

• Amy Eskridge, died in June of 2022; Eskridge was an "anti-gravity researcher"
• Michel David Hicks, a planetary researcher at NASA/JPL, died in July of 2023
• Anthony Chavez, of the Los Alamos National Laboratory, disappeared in May of 2025
• Jason Thomas, assistant director of chemical biology at Novartis, drowned in Lake Quannapowitt, Massachusetts in December of 2025

So the relevant question here is: does this make a pattern?

The answer is no.  Not even close.  Okay, McCasland and Reza are curious circumstances, to say the least; but McCasland had medical problems, and even his wife wrote a piece pleading with people to stop the "misinformation circulating about Neil and his disappearance."  As far as Reza... well, I'm an experienced back-country hiker myself, and the wilderness is a big, big place.  If you don't think it's possible someone -- even a dedicated backwoods-explorer -- could get inextricably lost and die out there, you haven't actually experienced what the back country is like.

How about the rest?  Well, the police actually arrested the perpetrator of Grillmair's murder, and it was pretty clearly a burglary that went badly wrong.  Loureiro's killer was almost certainly Cláudio Manuel Neves Valente, who knew Loureiro personally and seems to have had a grudge against him -- Valente had gone to college with Loureiro and graduated first in his class, but unlike Loureiro, his career never took off.  Valente was described as "often unhappy or even angry," and eventually snapped; the day after killing Loureiro, he went to Brown University, killed two students and wounded nine others, then turned the gun on himself.

As far as the others, Eskridge wasn't a government employee at all, just a fringe-y pseudoscience content creator; she'd been in chronic pain for years, and it's nearly certain she committed suicide.  Hicks died of natural causes from a chronic medical condition.  Chavez wasn't a scientist, but a construction foreman, and hadn't worked for Los Alamos for years.  Thomas, a medical researcher who had zero to do with space science, suffered from long-term depression, and his drowning death was ruled as an accident.

So what looked like a pattern turns out to be nothing much at all.  But of course that's not calming anyone down.  Karoline Leavitt, who hasn't been within hailing distance of reality for years and wouldn't recognize the truth if it came up and bit her on the ass, says there are "legitimate questions about these troubling cases" and that "no stone will remain unturned" in unmasking the conspiracy behind it all.  Especially if it turns out to have nothing to do with Jeffrey Epstein.  Podcaster Walter Kirn goes further, saying "What is going on seems to be an enemy action...  [The missing individuals were involved] in the most advanced realms of space-rocket propulsion and, you know, Air Force–NASA–type endeavors."  House Oversight Committee Chairman James Comer said that "something sinister could be happening."  Then Missouri Representative Eric Burlison got involved -- you may recall that last year, Burlison made a name for himself by claiming that the Nephilim were real, and the Smithsonian Institute was covering it up, so his grasp on reality is even more tenuous than Karoline Leavitt's -- and said the whole thing had to do with the fact that all of the victims, alleged and otherwise, knew classified stuff about UFOs.

And we're off to the races.

[Image licensed under the Creative Commons SkepticalScience, Conspiracy Theories Fallacy Icon, CC BY-SA 4.0]

It's pretty clear that what we have here is a consortium of the usual kooks and conspiracy theorists teaming up with a bunch of politicians who are desperate to distract everyone from the fact that their policies have directly led to an economic disaster, and that the upper leadership was involved in a vicious sex trafficking ring, so they're putting 2 and 2 together and coming up with 54.

Like I said... as usual.

In any case, if you see anyone posting this nonsense as serious evidence of a huge conspiracy, I'd be much obliged if you'd set them straight.  You can even send 'em a link to this post, if you think it'll help.

After all, we have enough real stuff to worry about at the moment.  There's no need to make shit up.

****************************************