Sentinel

Sentinel species - Polish mussels and Cassandra humans - [extinction]

Photo by Jelle Taman on Unsplash

---

The year is 1898 and Erwin’s world is a single cone of light surrounded by heavy, humid darkness. He slithers through the bowels of the earth. The soundtrack? Distant grunts, clangs, and the lilting titter of a canary, holding back the oppression of tons of soil and rock above Erwin’s head. The fragrance? The rank, sweaty smell of toiling men and boys; boys like Erwin, small and skinny, who wriggle through narrow openings to inspect what lies ahead.

His quarry? Coal. Unbeknownst to Erwin, he’s hunting for the remnants of ancient beasts compressed into the polluting fuel that drives the industry that chains him. All he knows is that more coal equals more days of food for his family. More coal equals more money from men in scrappy suits, who work for men in slightly less scrappy suits. A chain of suits ending in people Erwin will only see in pictures.

Erwin stops scrambling. Something is wrong.

The canary no longer sings.

Danger often arrives in silence.

---

Sentinel species

In 1895, the Scottish physiologist John Scott Haldane¹ first proposed using canaries in coal mines to detect carbon monoxide. The little birds, after all, are more sensitive to this toxic gas than humans are. By providing an early warning for the odorless and colorless gas, coal mine canaries are a prime example of a so-called sentinel species — animals, plants, or microbes that are more sensitive to environmental hazards than we are, and so can provide a heads-up.

Sentinels are useful because human senses are not that impressive.

Take our sense of smell, for example. Humans possess about five to six million olfactory receptors, which is respectable among primates. But humanity’s best friends have up to 300 million, which makes them anywhere from 10,000 to 100,000 times more sensitive to odors than humans.

In terms of hearing, humans detect sounds in the frequency range of about 20 to 20,000 hertz. That sounds good, but dogs up the ante to around 45,000 to 65,000 hertz, and they pale in comparison to bats that use echolocation at over 100,000 hertz to navigate and hunt insects in complete darkness. On the opposite end, elephants can detect infrasound below 20 hertz, which allows them to communicate across long distances with rumbles that are completely inaudible to us.

Human vision is… alright, particularly in daylight. We have trichromatic color vision², with three types of cone cells that allow us to see a wide range of colors, and our visual acuity is pretty decent. Still, we fall short in several areas. Eagles have a visual acuity four to five times sharper than humans, enabling them to spot prey from kilometers away. Then there’s the mantis shrimp, which boasts twelve to sixteen types of color receptors.

When it comes to taste, humans have between 5,000 and 10,000 taste buds, allowing us to perceive sweet, sour, salty, bitter, and umami flavors. This is relatively average compared to other animals. Catfish, for example, have over 100,000 taste receptors, many of which are located across their skin. They basically taste their surroundings.

Our comparatively most impressive sense, though not always getting the attention it deserves, is touch. Our fingertips and lips are especially sensitive. The nose of the star-nosed mole, however, is covered in more than 25,000 tiny sensory organs, making it the most touch-sensitive structure known in the animal kingdom. Elephant trunks also possess extraordinary dexterity and sensitivity.

Leaving the human sensory systems in the dust, many animals possess sensory capacities that humans lack entirely. Birds and sea turtles can detect Earth’s magnetic field to navigate across vast distances. Sharks and certain types of fish sense electric fields produced by other animals. Pit vipers and some species of beetles can detect infrared radiation (they ‘see’ heat).

And then there are mussels.

Polish mussels and Cassandra Humans

The Polish capital, Warsaw³, is home to over 1.8 million people. A sizeable portion of its water supply is protected by a diligent team of eight… mussels. These shellfish are extraordinarily sensitive to contaminants in the water, and when they close their shells, human specialists are alarmed for further testing.

Ignore the mussels at your own peril.

Wait. Did you think about Greek mythology too? What a coincidence.

Pantheons are wildly more interesting than one all-powerful god because pantheons of many gods are more, well, human. The members of a pantheon have personalities, pet peeves, and predilections. Take Apollo, the Greek god of healing, prophecy, and music, among other things. Widely considered the most beautiful god of the Greek pantheon, Apollo is gifted with the trait that all beautiful men possess: They hate rejection (because they’re not used to it). So, when the priestess Cassandra rebuffed the pretty god’s advances, he cursed her with the gift of truthful prophecy. The catch? No one would ever believe her.

Cassandra, of course, is but a character in a myth.

Are there real-life Cassandras, though?

Perhaps. In our glorious age of quantification, prediction markets are yet another attempt to forcefully push the complexity of reality into a single number. In broad strokes, prediction markets are (virtual) places where people can predict the probability of future events. Crowd-sourced Cassandra⁴, basically. But within the Cassandra crowd, there are individual Cassandras, known as superforecasters.

The Good Judgment Project, which began in 2011, sought to steer the crowd’s wisdom toward forecasting world events. Over time, it became clear that some people were consistently more accurate in their predictions than both the rest of the crowd and domain experts. In some cases, superforecasters were even more accurate than intelligence analysts with access to secret intel.

How do they do it?

To simplify, they intuitively collect a lot of data and then try to remove the biases from it. Superforecasters seem to be especially adept at removing noise from the signals they pick up. Part of this can be learned through training in probabilistic thinking, but part of it may be intrinsic. Superforecasters tend to be intelligent, politically knowledgeable, and open-minded. They’re not shy about changing their minds and updating their beliefs.

But, as it was for the Cassandra from myth, some people (myself somewhat included) are skeptical about the superforecaster concept. Claims that supported the initial burst of ‘superforecaster’ into the cultural ether appear to have been overly generous, and the always controversial Nassim Nicholas Taleb and colleagues have criticized the concept by claiming that the “superforecasting enterprise“ doesn’t seem to translate well to real benefits for decision makers.

Cynical me would agree and say that it’s (mostly) a marketing trick. Then again, ignoring Cassandra can have disastrous consequences.

Like extinction.

[extinction]

I’ve written about human extinction before, but let’s tailor it to this post and reframe it as p(doom), or the probability of humanity going extinct.

The idea/meme of p(doom) has known a resurgence since the advent of generative AI. All and sundry were suddenly eager to put a number on humanity’s chance of survival when faced with the (false⁵) promise of artificial general intelligence or AGI. Of course, p(doom) is meaningless without a time horizon. P(doom) without a time horizon is always one. Humanity will go extinct, even if only at the heat death of the universe.

But p(doom) by 2100, for example, is another matter. That’s tangible and not that far off. So what do our superforecasters say? In 2022, the Forecasting Research Institute asked superforecasters and domain experts exactly that. What are the odds of humanity going extinct by 2100?

• Experts: 6%

• Superforecasters: 1%

The biggest risk, as seen by both the experts and superforecasters, was AI (0.38% for the supers, between 2 and 4.75% for different domain experts), followed by engineered pathogens, nuclear extinction, non-anthropogenic risks (giant meteors and the like), and natural pathogens. The order of risks from most to least probable to kill us all was the same for the superforecasters and experts, but in all cases, the supers were more optimistic than the experts.

A few thoughts:

• Experts are probably biased toward their fields as the main risk and they probably overestimate the risk.

• Both experts and superforecasters are not immune to hype, as seen by the number one spot for ‘AI’⁶.

• Where is climate change? Habitat destruction? The list of possible risks in itself probably also reflects a certain bias.

In the Greek myth, Cassandra warned the Trojans about the soldiers hiding in the giant wooden horse. She was ridiculed. During the fall of Troy, she retreated to the temple of Athena and clung to the goddess’s statue, only to be dragged away by the mighty Greek warrior Ajax. After the war, Cassandra was taken as a concubine for the Greek king Agamemnon. Cassandra, as well as Agamemnon, were killed later by Agamemnon’s wife, Clytemnestra, and her lover, Aegisthus.

If you put on your Cassandra hat, what is your p(doom)? And what do you think will be humanity’s downfall?

---

The year is 2098 and the sky is iron.

At least, that’s what it feels like to Rewin, the gray sky, heavy and oppressive with smog weighing down on his skinny shoulders. He quickly shuffles into the pod of the vacuum train and tunes in to the orb’s song through the infoplant grafted into the bone behind his left ear.

The golden orb, he knows, is the size of a large city and it hovers beyond the iron sky. But what the orb is, no one knows. It appeared before Rewin was born, in 2065, and remains stubbornly impervious to any attempts at contact or analysis. Even though it has become a thriving field of research, no scanning method gets through. All the orb does is hover and sing. That’s how Rewin sees it anyway. There are as many theories as there are questions. Language, some say. Mathematics, others claim. A warning. A beacon. A hello. A get lost.

The more alarmist observers see the orb as a canary in a dark forest. Maybe it guards Earth from other ‘things’ out there. Maybe it’s a sentinel that sings as long as Earth is safe. Maybe it’s a scout that signaled home.

Whatever the case, it’s hard to escape the notion that —

Rewin taps his infoplant. The canary’s song stopped.

This is the day the world ends.

---

Thanks for joining me in another Subtle Sparks exploration, this time one that took us from coal mines to ancient Greece and ended with human extinction. If you enjoyed the journey, click a button; it helps.

1. This guy experimented on himself a lot, breathing toxic gases ‘for science’. He also ran experiments on his son (like throwing him in the ocean in a homemade diving suit). That son, J.B.S. Haldane, would grow up to leave a big mark on evolutionary biology.

2. Some people may be tetrachromats, who boast a fourth type of color receptor, somewhere between green and red. There has been one confirmed case, but because tetrachromacy seems to be linked to the random inactivation of one of two X chromosomes, it’s been speculated that some women who are carriers of the genes for a specific form of color blindness may be tetrachromats too. (Bonus fact: the distant ancestor of all vertebrates was probably tetrachromatic.)

3. For my US readers, Minneapolis also uses mussels to monitor water quality.

4. Prediction markets are predicated on the ‘wisdom of the crowd’ idea, which states that the collective judgment of a group tends to be the most accurate. It’s a popular idea, but it’s not without its caveats. It works better the more diverse the group is, and it tends to be true mostly for questions that have a single correct answer.

5. I don’t think we’re as close to AGI as some people in the media seem to think.

6. It has become fashionable to cast ‘AI’ as the next big doom scenario for mankind. I personally don’t lend much credence to the more extreme ‘evil (or indifferent) superintelligence’ scenarios. However, I can see the current generation of generative AI as enabler for bad stuff, like providing factually wrong and biased information used by policy makers, deepfakes and misinformation that rile up the public into voting for more extreme politics, a boost for the fossil fuel industry, etc.