Review

tl;dr: The more knowledge you  have, the smaller the button you need to press to achieve desired results. This is what makes moth traps formidable killing machines, and it's a good analogy for other formidable killing machines I could mention. 

Traps

I was shopping for moth traps earlier today, and it struck me how ruthlessly efficient humans could be in designing their killing apparatus. The weapon in question was a thin pack in my hands containing just a single strip of paper which, when coated with a particular substance and folded in the right way, would end up killing most of the moths in my house. No need to physically hunt them down or even pay remote attention to them myself; a couple bucks spent on this paper and a minute to set it up, and three quarters of the entire population is decimated in less than a day.

That’s… horrifying.

Moth traps are made from cardboard coated with glue and female moth pheromones. Adult males are attracted to the pheromones, and end up getting stuck to the sides where they end up dying.[1] The females live, but without the males, no new larvae are born and in a few months time you’ve wiped out a whole generation of moths.[2] These traps are “highly sensitive” meaning that they will comb a whole room of moths very quickly despite being passive in nature.

Why are moth traps so effective? They use surgically precise knowledge. Humans know how to synthesize moth pheromones, and from there you can hack a 250-million-year-old genetically derived instinct that male moths have developed for mating, and then you set a trap and voilà. The genetic heuristic that worked 99% of the time for boosting reproductive rates in moths can be wielded against moths by obliterating their reproductive rates.

Moth traps aren’t even the pinnacle of human insecticidal war machines. Scientists have, after all, seriously considered using gene drives to eliminate an entire species of mosquitoes with a single swarm and some CRISPy cleverness.[3]

The smallest button

Moth traps and gene drives work by understanding something so well that when you use brute force (because everything is brute force) to do something, you do it in the most optimal and surgical way. Intelligent design means humans can engineer very, very effective traps that harness the smallest buttons you can push in order to get a desired result.

Evolution can also produce sexually deceptive traps that take advantage of insect brains. This is because genes that contribute to pushing a particular button that makes reproduction more likely, are more represented in the environment, so most genes in living beings today are already vetted for their capacity to harness niche buttons in the universe.

The blind idiot god can’t hope to compete with intelligent design however, so we can expect humans to win the find-the-smallest-button arms race against their evolution-derived enemies (like moths, mosquitoes, or viruses).

Brute force

Brute force always works. If you stuff enough moths into my house, my measly passive traps won’t be sufficient. In fact, if my house were big enough and there were enough moths, the males that were somehow not attracted to my sticky female pheromones but found females anyway would be the only ones to pass down their genes. With enough moths and enough time, the blind idiot god of moth evolution would find a way to elude my traps by pressing an alternate small button to those specific pheromones, in order to power its reproduction. This type of brute force, which grants a stupid and blind enemy the power of adaptation, can be found in battles with cancer, viruses, or pesticides.[4]

The only counter to this brute force is more brute force, in the form of chemotherapy, gene drives, or pesticides 1 level of magnitude deadlier than the last. One fell swoop instead of targeted attacks.

Why is this important?

Being intelligent lets you engineer things. It lets you find the smallest buttons to press to bring about results, and often those buttons are insidious and lay on a whole other dimension than the ones your enemy operates on. Moths can’t understand “synthetic pheromones” and they can’t understand “I’m killing you because the vast fields of cloth you call food are my clothes, and I don’t want holes in them”. Individual moths cannot compete with my moth traps, and the only hope moth genes have of surviving in the long term involves brute force in the form of swarm-worthy numbers and long timescales.

Given how fast human progress is going, it won’t be long before we have more efficient moth traps that can respond to adaptation, or before we find a reliable “one fell swoop” solution (like gene drives for mosquitoes, chemotherapy for cancer, or mass vaccination for smallpox).

Moth traps operate on a dimension above individual moths. In turn, an artificial superintelligence would operate on a dimension above individual humans. Tiny buttons equivalent to “moth pheromones” would be available to it. One fell swoop solutions like “gene drives” would also be available to it. Humans are currently in the phase in between “moth traps” and “gene drives” in which they are not effectively omnipotent relative to moths, but they are very close. An AGI would bridge that gap much faster if only because transistors are faster than synapses.

Eliezer Yudkowsky on X: "(via Dank EA Memes) https://t.co/5y7jdn1drg" / X
  1. ^

    It’s a pretty horrible way to die, too. When I checked in on the trap at some point, there were about a dozen moths there, all of them alive and squirming, wings permanently damaged with no hope of escape. Insect sentience is still up for debate but it’s still not very fun to watch.

  2. ^

    Read more here. (The Wikipedia page ends with a hit list of insects that pheromone traps are effective with.)

  3. ^

    The case for using this seems strong.

  4. ^

    Cancerous cells that somehow evade your immune system (or chemotherapy rounds) survive to reproduce; viruses that bypass human vaccines survive to reproduce; insects that resist certain pesticides survive to reproduce.

    Numbers and time are the weapons with which humanity’s unagenty enemies fight. (Evolution is kind of agenty and it is an emergent property of numbers and time. An emergent intelligence, if you will.)

New Comment
17 comments, sorted by Click to highlight new comments since:

>the blind idiot god of moth evolution would find a way to elude my traps by pressing an alternate small button to those specific pheromones, in order to power its reproduction. This type of brute force, which grants a stupid and blind enemy the power of adaptation, can be found in battles with cancer, viruses, or pesticides

Interestingly some cockroaches have evolved to perceive sugar as bitter, due to its use as bait in traps: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC8003998/

Interesting, thanks for posting that! One of the reasons I like this forum is because there are people running around on here who've read papers like "Salivary Digestion Extends the Range of Sugar-Aversions in the German Cockroach" and you get to talk to them for free. 

So if I understand the abstract and skimmed paper so far, we're seeing more saliva-based aversion to pure glucose because pure glucose is a superstimulus (the roaches still accept "complex glucose"), and human trap designs are fond of superstimuli, as cheap ways to radically increase the probability your trap works, so the traps are selecting for pure glucose aversion. Given how short insect reproduction cycles are and how many there are anyway, we'll probably observe this kind of evolution everywhere, as well as every time we switch traps. 

This reminds me of an insight I found while staring at the ground, a spider was frantically crawling on the soil, under leaves and sticks and then suddenly through an opening a spider wasp sweeps in, attaches itself to it and kills the spider within a second.

Crawling insects are 2D navigation specialists, often prey to flying insects, 3D navigation specialists. In this sense a geometric dimension is used by the 3D specialists to hunt down crawling insects. Merely raising a dimension doesn't give you the trap, though, because a high perception and energy cost is incurred to maintain that 3D navigation ability. With this perception and energy cost, flies are more vulnerable to the traps of spiders. These costs could probably be modeled as extra latent space dimensions alongside the spacial dimensions involved.

Web-weaving Spiders are 2D specialists that slice a 3D space with a sticky web, capturing 3D specialists without expending energy to navigate their prey's space.

Just like it takes work for the manufacturer of moth traps to source the cardboard and obtain the knowledge of moth pheromones, it takes work for a spider to actually make the web (with patterns related to the silk's transparency and web design encoded in DNA). That being said, I don't know of spiders that coat their webs with prey insect pheromones to further attract their prey (and one is left wondering, would that be too small of a button as-to significantly deplete a spider's food source?).

There are predator insects that mimic the flashing patterns of certain firefly species, but they have to have more "skin in the game" and can only trap-then-eat one firefly at a time, there is less of an asymmetry in this instance.

Effective traps are time <-> space translation constructions. The button of the construction is in latent space.

Hello! To support your point, I think entomology is particularly fascinating as a window into how evolution works because of how many niches there are in the micro world. "Amazon rainforest" is more or less a single biome for macroscopic humans. But for insects, a whole new set of dimensions are navigable and there are substantial differences between, say, tree X, pond Y, canopy W or river V.  When you're small, things aren't just bigger; there's also a lot more variety because what look like subtle differences to us (like whether a tree is wet or not) is a huge difference for small buggers (water is sticky at small scales, and can drown insects). 

This is to say that there aren't just two dimensions in the small world; some spiders operate on one dimension, other spiders operate on an other dimension, and of course 12,000 different species of ants all have their own way of integrating whatever niche dimensions they operate in. 

You can continue your way downward, by the way: the world of unicellular organisms is incredibly dense and varied, and there are way over a million clearly identifiable bacterial species alone. 

So when you describe the nature of traps, there are probably thousands of effective space translation constructions, beyond just 2D and 3D.

I mention the Amazon specifically because that part of nature is a hellish death zone, where insects genocide each other every other Tuesday while odd inventions like door-shaped-ants, zombie-ants fungus and worryingly intelligent trial-and-error capable spiders like Portia (technically not native to the Amazon) pop up all the time before getting out-tactic-ed by some other horridly violent species. There is a lot of small-button-pressing going on here, and new tactics that let you explore a new dimension are so effective that "one fell swoop" strategies are common. 

It's like in Worm where every time the protagonists are up against a person with a new power, they nearly die, because the effect of surprise is just that powerful. When you live in a world with superpowered individuals, capability distribution between humans is uneven; the same can be said for the level of variance between species of ants, for example. In both cases, individuals have access to space translations very different from those of their enemies, which is why surprise and the inability to adapt are common observations. 

Hi! Thank you for this walk in the space of insects. The spacial complexity implicit in pockets of miniature space seem akin to fractal dimensions or hausdorff dimensions.

Exploring insect space reminds me of my experience exploring non-self-similar 3D fractals, such as hybrid variants of mandleboxes and menger sponges. In these 3D fractals there are fields of complexity and patterns bound to specific scales, and zooming into surfaces would reveal more highly complex space on that surface; branches would "pop up", and I could rotate my camera around these branches, then zoom into those. And I could repeat this until my zoom level exceeded the floating-point precision offered by by program I was using.

I also see what you mean by the implicit time / energy constraints when moving on surfaces giving extra dimensionality, from the perspective of an insect. Depending on the insect's locomotion abilities certain surfaces would allow it to move way faster and easier than others, and the differences can be quite stark.

This reminds me of another insight, while thinking about brain-machine interfaces, regarding to how brain neurons are organized: neurons on the cortex / surface of the brain of are highly connected and have a high fractal dimension (~ 2.8 according to wikipedia and the paper it references). As you go deeper into the brain, towards the corpus callosum, this complexity is reduced... axons are longer and tend to be covered with myelin sheath, which increases the conductivity of these connections for longer-running connections. So from the perspective of a neuron in the cortex, neurons way further away can appear closer since the charge and sensitivity requirements between topologically distant neurons are similar to the connections of its neighbors.

As for Worm, I have not read it yet, but the ecological feedback loops being described here is very fun to think about.

---
refs:

3D fractal rendering software: https://mandelbulber.org/
https://en.wikipedia.org/wiki/List_of_fractals_by_Hausdorff_dimension
https://www.sciencedirect.com/science/article/abs/pii/S105381190300380X?via%3Dihub
 

That is a pretty good point,  who's to say there aren't 'predators' of some sort in higher spatial or temporal dimensions?

Wouldn't that obviate the concern about 'AI' though? Maybe they will be our best friends against the 4D spider wasps?

I have thought about this. I still don't know if that obviates the concern about an advanced super-intelligence. Why would it not collaborate with the 4D spider wasps?

Maybe they will feel sympathy for their fellow 3+1 dimensional brethren.

I'm not sure how useful it is to speculate about 4 dimensional spider wasps. Seems like the same type of reasoning that leads people to seriously debate about being in a simulation; the answer to that is "no experiment you can perform inside a simulation can prove that you are in one". In my conception of "dimension", you're essentially stuck where you are and would be operating on literally zero evidence if you thought about 4 dimensional spider wasps. There's no point wondering "I wonder whether the simulators skipped lunch today and are therefore particularly grumpy today", because you'd be doing your rationality more harm than good by coming up with elaborate theories about concepts that aren't concrete at all. 

I don't see how we can talk about '4 dimensional spider wasps' in any productive way, much less wonder what side the 4 dimensional spider wasps would take in an alliance, or whether AI would want to help us in our fight against the 4 dimensional spider wasps. 

If there's something I'm not understanding or if you guys have a clear image of what a 4D spider wasp is and I don't for some reason, please tell me. I'd be genuinely curious if '4D spider wasp' was an actually useful concept. 

The LessWrong Review runs every year to select the posts that have most stood the test of time. This post is not yet eligible for review, but will be at the end of 2024. The top fifty or so posts are featured prominently on the site throughout the year.

Hopefully, the review is better than karma at judging enduring value. If we have accurate prediction markets on the review results, maybe we can have better incentives on LessWrong today. Will this post make the top fifty?

Hey, that comic’s originally from https://www.buttersafe.com/2011/01/27/traps/ Might be nice to put a link to the original artist :-)

Interesting, thanks for sharing that! I tried searching it up on knowyourememe.com but nothing turned out for some reason. 

Given how fast human progress is going, it won’t be long before we have more efficient moth traps that can respond to adaptation, or before we find a reliable “one fell swoop” solution (like gene drives for mosquitoes, chemotherapy for cancer, or mass vaccination for smallpox).

We do, in fact, already have several foolproof methods of moth elimination, involving setting the ambient temperature to several hundred degrees, entirely evacuating the air from the space, or a small thermonuclear warhead. The reason that we don't use these methods, of course, is that there are things we're trying to optimise for that aren't merely Moth Death, such as "continuing to have a house afterwards". This is probably also an analogy for something.

Hello! Indeed, and those are all one fell swoop solutions, and they are messy and destructive. What I meant by "reliable one fell swoop solution" and by the examples of gene drives, chemotherapy and mass vaccination, is that you could target moths on a dimension so specific to them that only they get damaged. The three examples demonstrate this, as gene drives target specific species without interfering with others, chemotherapy is targeted to the tumor (chemo is the most iffy of the three), and mass distribution of smallpox vaccines only affects smallpox. 

If you understand your enemy well enough you can fight it on a battlefield so specific to it that no collateral damage occurs. This is what I referred to as "finding the smallest button". 

The point is to build a moth trap, which is very precise, on a scale that can actually achieve Moth Death. The three solutions I'm positing against moths are "pure brute force" (thermonuclear warhead in your house), targeted subtle button-pushing (moth traps), or a combination of the two, which could very well be gene drives (which would end the 250 million year old bloodlines of moths forever without so much as touching anything else we care about). I'm saying that we don't have that last option just yet but if we really wanted to, current technology could probably let us (which is kind of insane). It's that third solution I was getting at, not the first.

The fundamental problem with this (and the big question of returns to intelligence) is that the buttons have to exist. And that is not guaranteed at all scales and for arbitrarily difficult tasks. It is unlikely that human intelligence is peak gains in general, but I don't think that we couldn't at least gesture in the direction of possible mechanisms for things to happen. It's a complicated question, but IMO to answer it it's more useful to ask "what kind of world do we live in?" than about intelligence in itself. The problem is that known hard thermodynamic limits on such stuff are usually very high, and practical limits are much harder to guess.

This idea reminds me a lot of how social media came to power. The goal of social media, contrary to how it advertises itself, is to show you ads. Social media is looking for a smallest possible button it needs to push to show you ads. Additionally, all successful social media have used machine learning to help determine the smallest possible button to show a user as many ads as possible. I believe what this post predicts has already begun to come to pass.

I think this could generalize to "low Kolmogorov complexity of behaviour makes it easy (and inevitable) for a higher intelligence to hijack your systems." Similar to the SSC post (I forgot which one) about how size and bodily complexity decreases likelihood of mind-altering parasite infections.