Hacker News new | past | comments | ask | show | jobs | submit login
Ant Colonies Retain Memories That Outlast the Lifespans of Individuals (smithsonianmag.com)
395 points by bangonkeyboard on Dec 12, 2018 | hide | past | favorite | 66 comments

These algorithms are similar to the ones I developed for the computer strategy for the Empire game. They worked, but I was unable to get them good enough to reliably defeat a human player.

The first strategy I tried was simply randomness. It was pretty ineffective. A rule based one was also ineffective, as it tended to get itself into a box. But I found that adding a dose of randomness to the rule based strategy worked well.

Most computer strategies of the day cheated, or altered the rules to give an advantage to the AI. Empire didn't. I did have thoughts about creating an interface so people could provide their own AIs and then the various AIs could battle it out.


There's been a lot of research on the Life game. But I bet Empire would be a much better research platform - the rules are simple, but the game play can get pretty complex. I had a lot of fun developing AI algorithms for it. I wish the ant algorithms were known at the time, I bet they'd help!

(One of the reasons the simple rules worked so well was the pieces had a hammer-paper-scissors relationship to each other.)

Hey Walter, did Sid Meier ever consult you while developing the Civilization series? I recognized a lot of similarities between Civ and Classic Empire when I went back and played the older games in the series.

I knew Walter developed some strategy game, but this is the first time I realized how big it was in the 80s. Definitely will try it out.

Also, I second the comment about using games like that in AI research. I don't have time now to explain it in detail, bit I believe it would be much more interesting/useful than playing Star Craft or any MOBAs. It has to do with the fact how short-term decisions affect long-term gameplay. Also, it would me much easier for non-gamers to understand what exactly researchers achieved.

Another game I would love to see played by AI is M.A.X. It's rules aren't any more complex than StarCraft, but the core gameplay is much deeper without diving into weir meta-gaming scenarios.

No, he never acknowledged Empire.

Do you mean he never acknowledged Empire as an inspiration? This article mentions him admitting that: https://arstechnica.com/gaming/2017/03/sid-meier-tells-civil...

Additionally, I'm like 75% sure I remember him mentioning Empire at a GDC talk.

I didn't know he had. Thanks for the link.

The game Armored Core: Formula Front for the PSP has a core mechanic that let’s players “program” their fighting robots in a destructured way. IIRC it’s about adding abilities and maneuvers that the AI can learn to use in combination while battling other AIs. It has a full campaign using this mechanic.

Battling Amiibos in the Super Smash Bros game for the Wii U could be compared to this in a way, too.

Walter, friendly note. On classicempire.com you have

> Empire is not a video arcade-type game. It is a thinking man's game, as it requires strategy and tactics instead of hand-eye coordination

You probably should say thinking person's game

Won't the term "person" unnecessarily alienate otherkin[1]?

Walter would appear to be safest with "thinking entity's game."

Then he has full coverage for men, women, intersex, asexual, androgynous, and non-human identities. His game will also still be relevant and inclusive once we've made contact with non-human extraterrestrials, and uplifted animals and machines to sentience.

[1] https://en.wikipedia.org/wiki/Otherkin

Weird, you're being a sarcastic shit but you don't exactly make a terrible point. OP is probably safe with "person", but if genuine concern came up I'm sure the two seconds it takes to change a word in an HTML file wouldn't derail his day.

Surely the "thinking" part could alienate AIs as well?

I just traversed the ant part of Wikipedia, and this is what I found.

> Queen ants have one of the longest life-spans of any known insect – up to 30 years.

> The fertile eggs become female worker ants and unfertilized eggs develop as males; if the fertilized eggs and pupae are well-nurtured, they potentially become queens.

> males are "quickly converted to sexual missiles."

> the queens often try to escape the males, allowing only the fastest and the fittest males to mate. Mating takes place during flight.

> One queen usually mates with several males. The sperm is stored in a special organ, known as a spermatheca, in the queen's abdomen, and lasts throughout her lifetime. This can be as long as 20 years, during which time the sperm can be used to fertilize tens of millions of eggs.

> The males have evolved for the single purpose of inseminating the queen.[3] During "the quick and violent mating," the male literally explodes his internal genitalia into the genital chamber of the queen and quickly dies.

And, moving on to army ants.

> A nest is constructed out of the living ant workers' own bodies to protect the queen and larvae, and is later deconstructed as the ants move on.

> As many as 150,000 to 700,000 worker bodies cover and protect the queen, linking legs and bodies in a mass that measures a meter across.

> In the morning, the bivouac dissolves into raiding columns that form a fan-shaped front. These raiding columns can travel up to 20 metres per hour with lead workers laying a chemical trail for other workers to follow. Smaller workers lead the column, while larger, formidable soldiers protect the flanks.

> When the queens emerge, the workers in the colony will form two 'systems' or arms in opposite directions. These queens that are hatched will move down either the arms and only two queens will succeed, one for each branch. The remaining new queens will be left in the middle and are abandoned to die.

> The whole colony of army ants can consume up to 500,000 prey animals each day, so can have a significant influence on the population, diversity, and behavior of their prey. [...] About five species hunt in higher trees, where they can attack birds and their eggs, although they focus on hunting other social insects along with their eggs and larvae.

>> The fertile eggs become female worker ants and unfertilized eggs develop as males

One interesting consequence of this is that male ants have no fathers. All of their genetic code is inherited from their mother, and there is no variation in what they pass on. This lack of variation in what fathers pass on means that two ant sisters share 75% of their chromosomes instead of the 50% you see in most other animals, which some researchers believe is the underlying cause for the high level of cooperation in ant species (eusociality).

More reading: https://en.wikipedia.org/wiki/Haplodiploidy

You can also, if you're in a certain quirky intellectual mood, think of an ant colony as a single organism, since all the ants in it have 75% identical DNA.

Its body just happens to be physically disjoint.

And if you’re in that quirky intellectual mood, could you also think of all of humanity as being a single organism, since we all share about 99.9% of our DNA? Certainly a thought I find hard to shake observing cities from the window of an airplane..

I'm not someone who studies this stuff, but I'd wager that the 99.9% thing isn't quite the same context as the 75% thing, and two ants in the same colony are more similar than two humans.

Not to detract from your comment about all humans being one organism and whatnot (hell, maybe ants and our bodies operate on a different level of consciousness that as specialized intellectual creatures we typically override woooah mann). I mostly just wanted to fish for more info on this 75% thing from someone who studies DNA and can confirm my hunch - and maybe has more info on correlation between DNA similarity and cooperative behavior.

  > I'm not someone who studies this stuff, but I'd wager that the 99.9% thing isn't quite the same context as the 75% thing,
  > and two ants in the same colony are more similar than two humans.
Exactly. There are several very distinct contexts for discussing percentages of genetic similarity, and although it's usually clear from context which one people are talking about, making it explicit would prevent nonsensical conflicts like this 99.9% vs 75% one.

Its not 99.9% vs 75%

It is more like 75% closer to 100% than 99.9%

That's exactly my point. They represent totally different things.

One difference is that thinking of an ant colony as a single Superorganism is a way for biologists to talk about Darwinian evolution of ants. That is, they frame the colony as the unit of evolution.

> Certainly a thought I find hard to shake observing cities from the window of an airplane..

What about maps? Reading maps of highly-populated and industrial locations reminds me of histology. Perhaps cities are the organs of humanity and the roads are the blood vessels...

Zoom out a bit further and you'll see the universe has the same structure. Or in a bit further and have a look at neurons.

Does collaborative behavior play a role then? For example, I would not consider a herd of cloned sheep to be one organism. Let's call it a grey spectrum then?

Which came first, the chicken or the egg?

Your analogy only works if you somehow engineered the sheep to reproduce like ants and waited for many successive generations. If the reproductive change stuck, the behavior of the sheep would presumably evolve in a way that they began behaving like a single organism, including individual sheep exhibiting extremely self-less behavior, such as how individual ants from a colony or individual cells from your body behave.

Exactly. Cloned sheep are still sheep, optimised by for a life in which they have no clones.

Whereas ants are instead optimised for this colony life. As are leaves of a tree, for instance, who are 100% willing to sacrifice themselves.

Hofstadter raises this in "Godel, Escher, Bach" discussing the characteristics of emergent intelligence.

Also true for bees!

Ants really do provide a peek into the diversity of evolution and often how little we know about the world around us. I buried the dream of becoming a myrmecologist myself, but there is so much fascinating stuff to share.

Ants can dive [0] and swim [1], and in mangrove forests adapted to survive extended periods with low O2 during flooding [2].

Some species have established ritual fights to solve territorial conflict [3] and while they are known to be fiercely territorial, there are species who defy the rule and fuse nests after intense fighting [4], or form mega colonies without hostility to strangers across globes - one of these mega colonies being located along the coast of California [5].

0: https://en.wikipedia.org/wiki/Camponotus_schmitzi

1: https://www.youtube.com/watch?v=JJw5hBB8-Os

2: https://onlinelibrary.wiley.com/doi/abs/10.1046/j.1440-6055....

3: https://myrmecos.wordpress.com/2008/10/12/question-ritualize...

4: https://phys.org/news/2016-03-mortal-enemies-allies-ants.htm...

5: http://news.bbc.co.uk/earth/hi/earth_news/newsid_8127000/812...

If you want to become addicted to ants, you need to visit the Ants Canada YouTube channel.


I went down the rabbit hole one day and watched a few videos. It's a YouTube channel about ants, with 4K video and 2.5 million subscribers. When you watch a video, you'll know why.

This is a lovely article, and I feel there's a lifetime of lessons to be learned from any colony behaviour. I've recently submitted a thesis on a swarm intelligence algorithm which exhibits some of these qualities, and the parallels between the two was a core component of the contribution.

This effect, where something is globally stable, even though all the individuals that make it happen are unstable is sometimes described as "a forest whose contours remain the same, as the trees all change".

Consider this algorithm for an ant colony foraging for a good source of food (this behaviour has also been observed in ants searching for a new nest site):

  for each ant in the swarm
    if the ant is unhappy
      run to a random ant
      if the random ant is happy
        follow it to its location
      if the random ant is unhappy
        select a location at random
    if the ant is happy
      the ant continues to search its current location

   for each ant in the swarm
     the ant searches a tiny of its location, at random
     if the ant finds food
        the ant becomes happy
      if the ant doesn't find food
        the ant becomes unhappy
If you run this algorithm you will find a "clusters" of ants form, which is a number of ants who share the same foraging location. Importantly, and this is mathematically proven, the largest cluster will form in the location with the best probability for finding food. This algorithm works even when the locations change over time and, as in the article, even when the ants which found the location are replaced with ants who have simply followed other ants to get there.

The aspect which captured my attention is that there are tiny changes to individual lines of the algorithm which implement diverse behaviours such as hill climbing, optimise for exploitation or exploration, and global optimisation.

The algorithm is called Stochastic Diffusion Search and I'm in the process of polishing a Python library which implements it and its many variants for a Show HN :) The repo is here https://github.com/AndrewOwenMartin/sds some info and an explanatory animation here http://www.aomartin.co.uk/sds-animation/ and an beta version is already on PyPi here https://pypi.org/project/sds/.

Contact me (email address on my profile) if you're interested in using this algorithm or contributing to the library, it needs snappy C implementations, and a better explanatory animation!

These emergent behaviors from relatively simple rules fascinate me. Have you seen https://pixling.world ? It’s basically a tool to build systems with rules for both a grid (environments) and agents on that grid (pixlings). It’s quite advanced - every agent can have a neural network that is mutated to choose what rules to run instead of you hard coding it, and agents can split which causes a slight mutation in the new agent (evolution).

Seems as good a time as any to mention this excellent book: https://mitpress.mit.edu/books/ant-colony-optimization

Really interesting stuff. There is code online somewhere. It's possible ant colony optimisation is dated now, I don't really know, but it's still fascinating.

It reminds me my old days playing with NetLogo. Multi-agent modeling and complex system.

Could you model this as a cellular automaton? Kind of reminds me of that. Sounds cool.

Each ant, or "agent", just needs to store its "activity" (whether its happy or unhappy) and its "hypothesis" (its current location). You don't even need to store the hypothesis of inactive agents, so it's pretty minimal as it is, it's not clear how you would make it a cellular automaton.

I have some ideas, but what did you have in mind?

Take a look at this visualization https://www.youtube.com/watch?v=VsHc91IhzdI

There's a repo in the description. Might not be specifically relevant to your project though

This is an example of an "extended phenotype" as written about in https://en.wikipedia.org/wiki/The_Extended_Phenotype by Richard Dawkins (invented the word meme).

Great concept, although harder to grasp than The Selfish Gene.

It is a bit like treating the ants as stateless functions, and storing the state in the world?

I mean, I feel like the wording "retain memories" might imply a greater degree of consciousness/agency than what was actually observed. The "memory" appears to be some completely tangible phenomenon, as per their comparison of trees "retaining memory" in the form of a tree growing bark around where a branch broke off. "Ants use the rate at which they meet and smell other ants, or the chemicals deposited by other ants, to decide what to do next." , etc. That said, it's still pretty interesting, and I think we can make some comparisons about how humans share knowledge/experience and what kind of tangible markers we use as cues for our own behaviour.

The concept of memory isn't related to consciousness; after all, we talk about computers having memory all the time.

Rather than say everything short of humans doesn't really "remember", I find it much more interesting to think how similar our concept of remembering (reinforced neural connections) is to phenomena we don't think of as alive like rivers or pheromone trails.

I'm not sure if I'm countering your point or supporting it, but I think a better example than trees is to point out how human cultures bear (and arguably, exist as different entities because of) after effects of events beyond the lifespans of those impacted.

Example: the American trend of "you must finish all the food on your plate" has long outlived its origins, and has been continued on not because people consciously choose to continue so much as they dont consciously choose to STOP passing it on. Any number of social events that are not universal (tipping, suburbs vs downtown as being prosperous, etc) can be examined this way, and our reasonable but unthinking reactions to events help define our cultures. If we arent being conscious of these choices, how is this memory any different than the ants?

I think the wrong comparison is being made - humans to ants. The memory in this case doesn't belong to the ant but to the colony, which could be considered an organism in its own right. Similar to how large companies have "memory" (and behaviour) that outlast the term of employment of individual employees.

In certain traffic conditions, bottlenecks can persist for hours without any apparent cause (i.e., there is no crash or stopped car blocking the road). Even though the cars entering and exiting the region are changing all the time, the bottleneck persists. You could say that traffic has "memory". Ant colonies can exhibit "memory" in a similar way.

The traffic example would work better if the bottlenecks persisted even after an extended period when all cars were removed from the highway.

The "memory" of the ant colony doesn't persist when all ants are removed, either. You can't have memory without having some substrate.

The article seems to be making the point that the concept of memory lies on a spectrum, with the tree having a memory of having had a branch is at one end of the spectrum. The ant colony stuff is somewhere in between. In many senses what they are speaking of is what we refer to as "culture," as in things that are learned and passed down. When a memory of a path to a food source is passed down, it is simply because the senior ant walked the path while another followed. The senior ant wasn't consciously "teaching" the younger one so much. Regardless, all of these things lie along a spectrum.

tl;dr - ants can be compared to neurons (they even use similar mechanisms) and ant colonies can be compared to consciousness (although they lack human-recognizable self-awareness). I think it's not useful to compare humans with ants and human culture with ant colonies (it's on another scale, although there are some similarities, too).

- - -

This type of memory can be compared to the building blocks of our memory when we look into our brains. Each neuron is basically an agent that acts on its own (comparable to an ant) - if you look into neuron behaviour, there are some pretty interesting things (e.g. they try to get to the locations where most of the activity is happening to get more energy and weaken connections based on activity - synaptic plasticity). The algorithms both employ are comparable - I see similarities between ant colonies and human brains.

The human consciousness emerges from this phenomenon - the biggest difference is that all our neurons work together to create a consciousness that is self-aware (actually, it's only a small part of the brain that creates "me"). Understanding ant colonies can be beneficial for understanding human consciousness.

https://youtu.be/TTFoJQSd48c (I don't care about the computer analogies, but the concept of individual agents acting together in the brain gets across)

Human memories are a tangible physical phenomenon as well.

>> Apr 23, 2014 - Almost 25 years since the Iron Curtain came down, deer roaming the Czech-German border still balk at crossing areas where electric fences one lay


I also remember a story from the US. A high-security prison was removed, its razor wire fences torn down. Years later, GPS data showed an outline of the prison. The deer were still afraid of a fence that they had never themselves seen.

An interesting definition of 'memory' (and 'history') includes the 'physical record' that we can decipher.

Anthropologists study 'memories' left behind by ancient cultures that created artifacts. And the Earth keeps 'memories' of extinct animals in fossils ... and of geological events in rocks and landforms. DNA and epigenetics keep a 'memory' of lessons learned. Darwin 'remembered' evolution by the study of species.

There's a valuable lesson to be learned here ... what record will 'digitization' leave behind?

Carpenter ants built a nest in my apple tv, it took me 3 days to clear them out. I hope they don't remember it.

Fascinating. First thing that comes to mind is Family Systems theory of psychology. The idea is the family unit itself is treated as an entity that is sick, rather than any single “identified patient.” I like it because it provides a useful model for understanding transmission of observable patterns like child abuse or other behavioral problems across generations. If ants can do it for certain types of behavior, it certainly seems plausible that “human colonies” aka “families” can as well.

The book A Deepness in the Sky (https://www.amazon.com/Deepness-Sky-Zones-Thought/dp/0812536...) goes really deep (no pun intended) on this idea! I highly recommend it as some really solid sci-fi. It was full of big new ideas I had never conceived of before and often found myself sitting down just to think.

Likewise Children of Time (https://www.amazon.com/Children-Time-Adrian-Tchaikovsky/dp/1...), another piece of solid sci-fi that explores some ideas about the development of "intelligence".

Likewise the short story "Glacial" in Alastair Reynolds's Galactic North book explores this concept. https://en.wikipedia.org/wiki/Galactic_North#"Glacial"

CRAB: It's too bad, Achilles, that you weren't here last week, when Dr. Anteater and Aunt Hillary were my house guests. I should have thought of having you over then.

ACHILLES: Is Aunt Hillary your aunt, Mr. Crab?

CRAB: Oh, no, she's not really anybody's aunt.

ANTEATER: But the poor dear insists that everybody should call her that, even strangers. It's just one of her many endearing quirks.

CRAB: Yes, Aunt Hillary is quite eccentric, but such a merry old soul. It's a shame I didn't have you over to meet her last week.

ANTEATER: She's certainly one of the best-educated ant colonies I have ever had the good fortune to know. The two of us have spent many a long evening in conversation on the widest range of topics.

ACHILLES: I thought anteaters were devourers of ants, not patrons of ant-intellectualism!

ANTEATER: Well, of course the two are not mutually inconsistent. I am on the best of terms with ant colonies. It's just ants that I eat, not colonies-and that is good for both parties: me, and the colony.

- Douglas Hofstatder, "Godel, Escher and Bach"

Doesn't almost every community of living creatures retain memories that outlast the lifespans of individuals?

Nice article. It’s fascinating how the experiences of individuals can shape the behaviour of the colony in future. (I am interested in studying this in future.) I think the ants as simple input and output devices i.e. provide stimulus, get some results. But, the interaction of ants provides the output to the complex situations.

When one observes the colony while doing any colony related function, one can see how each individual is assigned with specific function e.g. foraging, nursing etc. And suppose if you remove these ants, there will be a new set of individuals doing that function. Somewhat similar to the neurons of brain. (One has to compare the extent of similarities.)

PS: I used to do research with ants. My master’s thesis was on colony cohesion. What are the factors important for the cohesion? I have looked at few aspects (physical) of this.

Is it surprising? We do the same and call this culture. A set of behaviour rules that we inherit from previous generations, don't really need to understand but can still use it to increase our survival and which slowly adapts to big time environment changes.

This reminds me of the parable of the 5 monkeys: http://wiki.c2.com/?TheFiveMonkeys

This isnt very surprising me considering we humans--and probably most other living species--are pretty good at this. Memory seems to be an emergent property of a network.

I’ve heard the concept of a “hive mind” before but this is extraordinary.

So basically it's a massively legged delay line memory

ants have culture, ants have memes.


Applications are open for YC Winter 2021

Guidelines | FAQ | Support | API | Security | Lists | Bookmarklet | Legal | Apply to YC | Contact