Recently, geologist Casey Luskin interviewed Eric Cassell, author of Animal Algorithms: Evolution and the Mysterious Origin of Ingenious Instincts (2021) on one of the central mysteries: How do animals “know” things that they can’t have figured out on their own?
Consider, for example, butterflies migrating over several generations from Canada to Mexico and back. No single butterfly makes the whole trip there or back.
How can animals do math they know nothing about? How can a great deal of information be packed into a brain with comparatively few neurons? We are slowly learning about some of that.
Eric Cassell is an expert in navigation systems, including GPS, whose experience includes more than four decades in systems engineering related to aircraft, navigation and safety. He has long had an interest in animal navigation. His model for animal navigation is the natural algorithm: The animal’s brain is “programmed” to enable navigation.
Here’s “Animal Algorithms Webinar: One of Nature’s Biggest Mysteries” (January 20, 2022). A partial transcript and notes follow:
Casey Luskin: So, uh, surprisingly, I didn’t fully appreciate the title of your book, Animal Algorithms, until yesterday when I was re-reading the book in an airport. (04:59): We’re talking about literally programming an animal brain.
You define an algorithm as a formal procedure for any mathematical operation, especially a set of well-defined rules for solving a problem in a finite number of steps. And what you’re saying is that animals are literally programmed with algorithms, hard-coded presumably into their brains to allow them to perform complex calculations… all kinds of behaviors vital for their survival. And you said in your book that those who program computer algorithms for a living are especially well-situated to appreciate how complex an animal algorithm must be to function.
I don’t program computers for a living, but I did write something like 30,000 lines of Python code during my PhD… algorithms are complicated and they really are a set of ordered instructions. If you get one thing wrong, they don’t work. So I want to learn more in this conversation about why you think animal behaviors are like algorithms and what you think the implications are for the ability of these behaviors to evolve by Darwinian evolution. (05:48)
My first question for you is what got you started on this Animal Algorithms book project?(06:25):
Eric Cassell: I only started actually writing a book a few years ago, but the origins of it go back a lot farther when I was working on aircraft navigation systems and then just happened to be reading several articles, about animal migration and navigation. I was amazed at the ability of animals to navigate so accurately and repeatedly…
And it was also about the same timeframe that I was studying biology in an undergraduate program. (07:06) And then I sort of put two things together in terms of, okay, this is a pretty sophisticated ability that animals have, and then started asking questions about how did such an ability even evolve in the first place and how do you explain how they’re actually implemented? So that’s kind of the origin of it. As time went on, I started looking at other kinds of animal behaviors and was finding similar kinds of behaviors that are complex, and to me, difficult to explain from a Darwinian evolutionary, point of view.
Note: Darwinian evolution is a model of evolution in which adaptations in life forms are driven by natural selection acting on random mutations. That is, the random mutations that are most successful enable the life form to survive and pass on genes, which code for the adaptation. The model, while widely accepted in the culture, has become controversial of late because it is unclear how adaptations that would require many co-ordinated changes at once can develop as a series of simple survival-and-reproduction tests.
Casey Luskin: How do you think that this field of work prepared you to study topics like animal navigation and migration. And how do the systems that animals use for navigation stack up against our human technology? (07:56)
Eric Cassell: There’s definitely algorithms involved, but from a bigger perspective, when you talk about sophisticated navigation systems like those used by, modern aircraft, these are highly engineered systems. Typically, you know, commercial aircraft have hundreds of thousands, if not millions of lines of code in their navigation and flight control systems. So they’re highly complex. (08:25)
And there’s a number of aspects of the engineering that’s involved when you build systems like that. One is, they’re highly integrated. So you have to match up not just the software, but the hardware, the sensors, the flight control system, et cetera. And then all of this has to be done in a coherent manner. (09:00)
And it turns out, surprisingly — not just to biologists, but to those of us that are engineers as well — when you start examining some of these systems in animals, they exhibit the same kind of principles that we use in developing man-made navigation systems. My favorite example is actually a desert ant that resides in deserts in Africa. These ants actually employ several different types of navigation centers. They use a sun compass, a polarized light compass. They have an odometer, they do chemotaxis, in other words with sensing chemicals. And then they use all that information too, in an integrated manner. (09:30)
Eric Cassell: And they actually do what biologists call path integration, what us engineers would typically call inertial navigation — basically the same thing, where they integrate all this information and then are able to navigate accurately. But one of the unique things about it is, whenever they go on a forging excursion from their home nest, they can go on a very circuitous path away from the nest, turning a number of times in different directions. But then when they go to return home, they are able to actually compute a direct path from wherever they are back to their home nest. Again, it’s based on this inertial navigation type of system. It’s very surprising that such a system exists in an ant. (10:26)
Casey Luskin: That’s absolutely incredible, Eric. You’re talking about path integration. This brings back a little bit of anxiety, doing my field work during my geology, degree in in South Africa. (11:25)
You would wander off looking for rocks and totally go off the path, not knowing where you are. And then you’ve got to find your way back to the car. But ants are able to do these calculations automatically. I mean, things that we human beings struggle to do, and they’re just programmed to do it. (12:10)
Eric Cassell: And, as you imply, we are just now catching up technology-wise with what many animals have been doing for thousands or millions of years. Our human ability to navigate long distances didn’t really become a reality until sometime in the 1700s with the development of accurate clocks that could be used on ships. Prior to that, ships had a lot of trouble navigating very long distances because they couldn’t determine longitude in an accurate way. (12:49)
There’s a famous example, I think that I cite in the book, about a fleet of British ships that got lost, ran aground, and the ships sank, because they didn’t have accurate navigation information. So that launched a big project by the British Navy to try to develop, a better system. It really wasn’t until the 20th century with the development of aircraft that we developed much better navigation systems and it’s taken a long time to get there.
Note: “The impact of the lack of accurate knowledge of longitude was indicated by the untold number of ships and seamen lost at sea due to poor navigation. The most famous such incident occurred in 1707 when four British warships ran aground and sank, killing about two thousand sailors and soldiers.” (Animal Algorithms, p. 50, citing Dava Sobel’s book, Longitude)
This disaster, known as the Isles of Scilly Isles Shipwreck, spurred a search for a means of determining longitude: “Early sea navigators could find accurate latitude – their ship’s position north or south of the equator – by observing the sun. Longitude – an east-to-west measurement – was calculated by estimations of speed and course from a given position at a given time, known as ‘dead reckoning’. This method risked cumulative error, compounded by variables of wind, current and tide and the impossibility of keeping accurate time at sea, where motion and temperature could affect timepieces.”– Historic England
Eric Cassell: And finally, with GPS, we have a system that actually starts to mimic what many animals have been doing for a long time.
Casey Luskin: Well, you talked in your book about, examples of jets landing at the wrong airports. And you would think that would never happen in our modern day of technology, but yet it is so vital to have accurate navigation information. And I think — and we’ll get into this more, Eric, in the conversation — but I think flying a plane is actually a really good analogy to the way Darwinian evolution works. My dad was a pilot. I’ve flown with my dad many times over the years, and I’ve actually been in a plane when it was losing, when it lost power. And it’s a pretty scary experience, let me tell you. (13:35)
You’ve got to get everything right in a plane or it falls out of the sky. I’m not telling you anything you don’t already know but it’s just like Darwinian evolution. If you have the whole system working, but one component fails in an organism, then it dies. And if it dies, it doesn’t pass on its genes to the next generation, and it’s an evolutionary dead end. (14:06) …
Your book is titled Animal Algorithms and I want to know exactly what you mean when you say that animal behaviors are like an algorithm. (14:45) In your book, you distinguish animal behaviors that are programmed from those that are learned like, say, many primate behaviors. And you say at the very beginning, you’re not talking about these learned behaviors, you’re talking about preprogrammed behaviors. What do you mean when you say that these behaviors are programmed and what makes them like an algorithm? (15:10)
Eric Cassell: First of all, they’re “programmed” in the sense that they’re innate. Most of these animals can actually perform these behaviors the moment they’re born. And so they don’t require development. Now, some animals refine the behavior over time. So there is a little bit of a learning process there but the basic behavior is there the moment they’re born, so that it’s innate and not learned. (15:41)
And then the “algorithm” part of it… there’s actually a number of different aspects of this. As you mentioned earlier, typically when we think of an algorithm, it’s more of a mathematical algorithm. That is the case for many of these behaviors, particularly the ones involving navigation, where the animal is actually computing a path. That involves trigonometry or some other form of mathematics. But other types of algorithms involved here are more along the lines of decision-making.
So for example, social insects actually are programmed to make decisions about the behavior that they’re going to perform. In other words, are they going to take care of the queen or take care of the other insects? A lot of different behaviors go on in social insect colonies. The algorithms are actually a process where they sense the condition that the colony is in and then make a decision about the optimum behavior to perform at any particular time. So that is a pretty sophisticated algorithm. (16:19):
Casey Luskin: Along these lines, you define a key concept in your book, a complex program behavior or CPB. What do you mean by a complex program behavior? (17:08)
Eric Cassell: First of all, it’s not just simple in the sense of a stimulus and a response. These behaviors are more complex than that. They’re programmed (and) they’re purposeful in the sense that they always have some goal. (17:21)
And then again, they’re heritable; they’re not learned, they are innate. So, they’re presumably in the genome in some way. There is some epigenetics involved, in some that we found but, for the most part, we think they have to be in the genome — although in many cases we really don’t know how they’re actually coded in the genome. (17:56)
Casey Luskin: This is one of the most incredible things you talked about in your book, how these tiny brains of an ant, or the worm C. elegans, which has only, I think you said, 302 neurons. How can a mind that small be programmed, and not just programmed with behaviors, but you talk about how these small brains are programmed to learn new behaviors? And (the worm can) also then have a memory so it can remember what it learned. (18:22)
We are told at Open Worm that despite C. elegans’s only 302 neurons, it can display complex behaviors.
Eric Cassell: Well, that’s kind of the big question. And it doesn’t just apply to C. elegans but many other animals as well. In this case, they’ve been able to map the entire brain, because it’s such a small number of neurons and the neurons are relatively large. But we have no idea how the programming of the behavior actually goes on.
Just to expand that analogy, the, other kinds of insects that are discussed in the book — honeybees in particular and ants — they also have relatively small brains. Now, in their case, they’re approximately a million neurons, which is still a really, really small brain compared to mammals for example. But they also have really sophisticated behaviors. (18:54)
And so somehow or another, these behaviors are programmed into the neural networks that comprise their brains. And again, we really have no idea how that’s done. There has been, there has been some research showing how, some of the most simple behaviors might be, encoded in a simple neural network, but that’s kind of the extent of it. We really don’t have any idea how these more sophisticated behaviors are programmed. (20:02)
Next: Can behavior simply be captured and transferred into the genome?
You may also wish to read: A navigator asks animals: How do you find your way? The results are amazing. Many life forms do math they know nothing about. The question Eric Cassell: asks is, how, exactly, is so much information packed into simple brain with so few neurons?