My blog about philosophy, coding, and anything else that interests me.
This article was first published with Conjecture Magazine.
The age-old question of how the mind works is a fascinating one. Yet researchers have been ignoring it strategically in the hopes of achieving AGI (artificial general intelligence) through an ever larger repertoire of tricks (Deutsch 2012). But there is no way around explaining intelligence if we want to program it.
Here, I take the challenge head-on with what I call the neo-Darwinian theory of the mind. It simply states that minds create new knowledge through the neo-Darwinian evolution of self-replicating ideas. It builds on philosopher Karl Popper’s guess that knowledge evolves through alternating conjecture and criticism (1983), but considers that a rather high-level explanation of a richer, underlying phenomenon. It solves several open problems in Popperian epistemology in addition to some others, such as how people evolved, how memory works, what it means to believe one thing rather than another, and why we are conscious of some phenomena and not others. Last but not least, it upholds and strengthens the Popperian notion that what’s central to the mind — the thing intelligence researchers have been bending over backward to ignore for the past 70 years — is knowledge creation.
Neo-Darwinism in general starts with a remarkably simple core concept — replication — yet it is rich in implications.
The first implication is that if you start with a single self-replicating something—anything, it need not be biological—pretty soon, you will have a population of them. I will call that something an instance of a replicator. A single instance creates another, each of which repeats this process, and so on. One turns into two, then four, eight, etc.
The second implication is that, because the world is messy and mistakes are inevitable, an instance of even the best replicator will eventually make a mistake during replication. That mistake changes the information the replica contains compared to the original. This change — a mutation — is inherited by all of its offspring (if it manages to replicate further). Mutations introduce variations. A mutated instance is called a variant. Usually, a mutation impedes any replicator’s ability to spread. Why? Because the replicator is already adapted to spreading, meaning any slight change would lessen its ability to do so (compare Paley’s remarks on adaptation, 1813). Also, keep in mind that mutations are accidental. A good replicator prevents them from happening.
Sometimes, however, a mistake can be beneficial. Beneficial for what? For the replicator’s ability to spread. That is its primary concern. Remember this crucial point, as many people tend to think of biological evolution in particular as optimizing “the good of the individual organism” or even “the good of the species.” Neither is the case (compare Dawkins 2016 and Deutsch 2012). Evolution does not optimize anything — it merely favors good replicators. That’s it. In biological evolution, that replicator is the gene, not an organism, nor any group of organisms (Dawkins 2016).
The third implication is that better replicators will spread through the population. This fact may sound tautologically true, and neo-Darwinism has repeatedly gotten a bad rap for it. It doesn’t matter. This fact — that better replicators will spread through the population — is the most critical thing you can realize not just about biological evolution, but also about meme evolution and the evolution of ideas in a mind. I will get to those other types of evolution a bit later. For now, just know that it’s that simple: better replicators will spread through the population at the expense of their rivals. Worse replicators will fall behind. We call these differences in the rate of replication selection (Dawkins 2016). Selection can result from competition, predation, mutations, changes in the environment, etc.
The fourth implication is that because of this repeating interplay of variation and selection, complex adaptations can emerge, and replicators can gain abilities beyond mere replication. Each adaptation will always be in the service of the primary purpose: replication (otherwise, it would not have been selected). Still, sometimes adaptations are complex enough to have sub-purposes that distract from this primary purpose. (Remember this below when I discuss the question of why we believe some things and not others.)
An age-old example of a curious adaptation that doesn’t serve any apparent purpose is the peacock’s tail. Why do peacocks have such beautiful and conspicuous tails? We could conjecture that it’s because their feathers are pigmented (WebExhibits n.d.), or because they “impress the opposite sex” (Peacock 2014 — yes, if you can believe it, the author’s name is Peacock and she wrote about peacocks), or whatever. These reasons are true, but they do not get at the heart of the matter. The primary reason is the same not just for peacocks’ tails, but also for why baboons have such prominent butts, why birds can fly, why bats hang upside down, and any other arbitrary trait: it happened to help the organism’s genes spread through the population at the expense of their rivals. I say “happened” because adaptations are not planned. Peacock genes that happened to code for slightly more colorful tails spread better through the population (i.e., gene pool) than those that didn’t. Baboon genes that happened to code for slightly more prominent butts spread better through the population than those that didn’t. Etc. Generally speaking: the reason any arbitrary trait x exists in some organism y is that genes that happened to code for slightly more of x in y spread better through the population than those that didn’t.
Most variants don’t have what it takes to spread through the pool of replicators. But sometimes, a mutation causes a variant to spread — like this genetic mutation resulting in white peacocks. Photo by Gio Bartlett on Unsplash
What does all of this have to do with the mind? Popper discovered that human knowledge evolves (1983). In other words, evolution as a phenomenon is not limited to the biosphere; it also explains how people create new knowledge. (Dawkins later added that evolution also explains how ideas change and spread between people (2016). He calls such ideas memes.) Popper offered a strict analogy between genetic mutations and conjectures on one hand and natural selection and criticism on the other. His explanation was a huge epistemological breakthrough. That’s why I have called (2020b) Popper the foremost AGI researcher of his time, even though he wouldn’t have called himself that.
Popper’s epistemology is problem-driven. He argued that we create knowledge by conjecturing solutions to problems, then criticizing them, and finally adopting them tentatively if we fail to criticize them. He considered the quest for certainty and the notion that we should try to confirm our theories justificationist mistakes.
I think he was correct about all of these things. Nonetheless, there are open problems with Popperian epistemology.
First, where do conjectures come from? They have the appearance of design (compare Paley 1813 and Hackethal 2020a), and so we need to explain their emergence as the result of evolution. Therefore, saying that human knowledge evolves using conjecture and criticism just kicks the can down the road, for, conjectures are themselves a product of the evolution that happens in a mind. (The same is true for criticism, as it is conjectural as well and also has the appearance of design.)
Second, Popper’s evolutionary model lacks the neo-Darwinian notion of a replicator. There may be a historical reason for this: he started developing his theories before the modern synthesis of neo-Darwinism was completed and before the discovery of DNA. That discovery was the first of a physical instance of a replicator (although it was known from theory beforehand). It wasn’t until the modern synthesis that the importance of replication was widely recognized. I have not read all of Popper’s books, but as far as I am aware, he did not introduce replication into his theory of human knowledge. (There has been an ongoing, friendly point of contention between the intelligence researcher Ella Hoeppner and me about the role of replication in any theory of evolution. Citing Donald Campbell’s evolutionary epistemology, she argues that replication can drive evolution, but is not necessary. All that’s needed, she argues, is variation and selective retention. I am agnostic on the question of necessity and think that if we introduce replication in a mind, we can explain a great many things that we couldn’t (at least not as well and as easily) without it. Hoeppner and I discussed this topic in more detail on my podcast (Hoeppner 2020).)
Third, some other problems aren’t directly related to Popperian epistemology but are nonetheless fascinating and in dire need of solving. Some of them relate to memory. For example: why do we remember some things but not others? Why do some people remember some things better than others? Why do some memories last longer than others? Why are they so notoriously unreliable? Further open problems relate to the question of beliefs: why do some people believe some things over others? Why is it easier to convince one person of idea x than it is another? (Popperian readers may recognize that the issue of “beliefs,” and especially “reasons for belief,” is foreign to Popperian epistemology — for good reasons, as they pave the way to justificationism. I do not intend to reinstate the traditionally important but undeserved and justificationist role of beliefs — instead, I want to understand why some people hold some ideas dearer than others. That’s all I mean by “belief.”)
To solve these problems, I propose the following conjecture: the mind contains an arena of self-replicating ideas. This conjecture is at the heart of my theory.
When we hear “replicating ideas,” we may immediately think of memes (Dawkins 2016), but they are not what I mean. Memes are ideas that spread between people. I am concerned here only with the dynamics within a single mind. The neo-Darwinian theory of the mind is not about meme evolution (though, like any theory of creativity, it does lay the groundwork for explaining how complex memes can spread between people at all: it takes creative minds to do that).
We start with the same, simple concept at the core of neo-Darwinism: replication. Then we apply its implications to the mind.
First, one self-replicating idea will generate a pool of ideas in a mind: an “idea pool.” (Since a single such idea is sufficient to do so, of all the inborn ideas we have, perhaps only one of them self-replicates.)
Second, because the mind is messy and mistakes are inevitable, sooner or later, an idea will make a mistake during replication: a mutation occurs. This mutation results in a variant idea. Usually, the variant will have a harder time spreading than the original. But now and then, the variant is better at spreading.
Third, ideas that are better at spreading will spread through the mind’s idea pool. There will be selection because some are better at spreading than others.
Fourth, because of the repeating interplay of variation and selection, complex adaptations can emerge, and ideas can gain abilities beyond mere replication. Such adaptations do not necessarily benefit the mind: evolution, as always, merely favors better replicators.
I believe it’s truly as simple as that. We can now solve the problems mentioned above:
Where do conjectures come from? They are newly-evolved variants of ideas.
Why do we remember some things but not others? Because some ideas are better replicators. Memories are not some distinct type of idea. They are simply self-replicating ideas of high longevity and relatively high copying fidelity. In other words, they manage to stick around for longer than most other ideas in the mind, and they manage to stay more or less the same for longer periods. Ideas that have these characteristics are the ones we call “memories.”
Why do some people remember some things better than others? Let’s say a husband and wife both remember their last vacation generally, but the wife remembers many more details. Why? Because her mind happens to be a suitable environment for the ideas encoding those details — meaning they can spread through her mind relatively smoothly — whereas her husband’s mind is not.
Why do some memories last longer than others? Because some replicators — and this is true of any replicator, not just ideas — are longer-lived than others. (However, I suspect that any particular instance of that replicator is usually pretty young. For example, if you are 30 and remembering something that happened when you were three, that particular instance of the idea encoding the event is not 27 years old — it may only be a few seconds old!)
Why are memories so notoriously unreliable? Because even the best and most consistent replicator won’t stay consistent forever. Mistakes are made eventually during replication. The same holds for ideas in a mind.
Why do some people believe some things over others? Because some ideas spread through their minds better than others.
Why is it easier to convince one person of idea x than it is another? Because one person’s mind makes it easier for idea x to spread through it than another person’s.
You may have noticed that questions of belief/conviction have very similar answers as those related to memory. That’s because memories are just long-held beliefs: long-lived self-replicating ideas. While there are significant differences between biological and mental evolution, the neo-Darwinian theory of the mind has a strongly unifying character generally.
As I said before, one answer to the question “why do peacocks have such colorful tails?” is “because their feathers are pigmented.” That answer is correct but misses the point and the central role of replication: again, peacocks have colorful tails because those genes that happened to code for slightly more colorful tails were better able to spread through the gene pool. The primary explanation involves replication, even if we may be tempted to forget that. Observe the same phenomenon when you ask people why they think some idea is true. If you ask someone “how come you think you were born in 1960,” they will almost certainly answer along the lines of “because I really was born in 1960,” or “because it says so on my birth certificate,” or “because my parents told me so.” All of these things may be true, but the primary reason he thinks he was born in 1960 is that the idea encoding this information managed to spread through his mind. Yet he will never give that as the reason.
The same holds for opinions as well. Let’s say Charlotte believes that people should pay their taxes. Why? If we ask her, her answer will invariably relate to why people should, in her opinion, pay taxes, not why she believes they should. To most, there is no difference between these two questions, yet this difference is crucial if we want to begin to understand the mind. She may respond that taxes are a great way to help the poor. Or that “transferring” money from the rich to the poor is a moral imperative. Or that we need to finance the government somehow. Or that everybody has to pitch in and do their part. Whether these statements are morally or factually correct is not at issue here — just note that she will never mention the primary reason: that those ideas have spread through her mind at the expense of their rivals. It is the primary reason anyone holds any arbitrary belief in mental evolution, just like it is the primary reason any organism has any arbitrary trait in biological evolution. Yet it will never occur to her. That is why I said earlier that the phenomenon of complex sub-purposes of adaptations distracting from the primary purpose plays an important role in the mind as well.
But why exactly will the primary reason not occur to Charlotte? I think it’s because she does not know how her mind works. None of us do. That’s the thing we are trying to explain here. And the vast majority of mental activities seem to happen subconsciously. We are unaware of the millions and millions of variants of self-replicating ideas in our minds, some collaborating, some competing, and we are unaware of the replication that occurs there. But we know of all of these things from theory.
Awareness brings us to the question of consciousness. It is perhaps the single most fascinating and mysterious phenomenon in the universe yet. Popper (1983) conjectured it has to do with disappointed expectations: we often do not even know we had a particular expectation until it is disappointed. For example, if you go up a flight of stairs and reach the end, thinking there is one more step, you will become acutely aware that there isn’t one and of your disappointed expectation. More broadly speaking, consciousness seems to have to do with error correction. (Error correction generally plays a central role in Popperian epistemology, not just when it comes to consciousness.) I have given the example (2020c) of learning to ride a bike as a child. Initially, you are acutely aware of every little movement you make: steering, keeping your balance, pedaling, etc. The process may be a bit overwhelming at first. Your mind evolves many new ideas for how to develop this new skill. Most of them won’t do the trick, but some variants happen to improve your abilities. Over many iterations, your ideas about how to ride a bike further improve. And as they do — i.e., as you correct errors — you become less aware of riding your bike. After riding thousands of miles, you won’t even know anymore how you do it. You just do it.
I believe disappointed expectations will play an essential role in any good upcoming explanation of consciousness, but they do not strike me as sufficient (or even necessary). Whenever you sit down, close your eyes, and observe the thoughts that pop into consciousness, not all of them have to do with disappointed expectations. Some of them are merely memories, for example. However, we do know that all of them are ideas — be they descriptive, argumentative, or explanatory — and they all have the appearance of design. We never seem to be aware of any “junk” ideas, although our minds must contain many of them at any given moment because of imperfect replication. Therefore, I conjecture that a necessary condition for something to pop into consciousness is that it be sufficiently — by some yet-to-be-determined criterion — adapted.
Maybe part of a mind is like a brewing soup of ideas, some of which manage to bubble up here and there. Those may be our conscious thoughts. When competition between ideas is especially fierce, and the soup goes through some turbulent times, it “boils,” and many ideas bubble up to the top, demanding urgent attention and occupying our thoughts. Maybe further development of this theory would explain anxiety.
Perhaps consciousness takes snapshots of the soup to examine ideas that have bubbled up far enough. Or perhaps it only ever looks at a single instance of an idea at a time. That could be why we are not aware of replication.
I hinted above that the neo-Darwinian theory of the mind sheds light on how people evolved. For more on that, read my book (2020c), which contains a much more detailed account of the theory generally. Here I will only mention briefly that the explanation involves something very similar to the RNA-World Hypothesis occurring in a single one of our ancestors’ minds. I conjecture that a genetic mutation resulted in a self-replicating idea in that ancestor’s mind. Once he invoked that idea, it replicated and created an idea pool, which kickstarted the first instance of mental evolution.
The theory is incomplete because it doesn’t explain consciousness. I suggest that as the next problem to work on. I conjecture that consciousness is software that can be written like any other software and that, due to computational universality, it can run on any existing universal computer. But whatever the case, we will need the neo-Darwinian theory of the mind — or something like it — to explain the mind and build AGI.
Only time (and criticism) will tell whether there is something to this theory. As a solution to the problems I laid out above, I think it works well. All I am doing is taking neo-Darwinism seriously. And all I ask is that you do the same — and help the theory spread.
For more content like this, follow me on Twitter.
There are 9 references to this post in: