In a recent article I discussed Robert Wright’s new book. His general thesis is that we ought to frame artificial intelligence cosmically, as a kind of evolutionary phase shift that will end up turning our civilization into some kind of global brain. We just have to decide on the kind of global brain we want to have.
I like Wright’s framing of the issue and I even sympathize with his invocation of Teilhard de Chardin, the maverick priest who started predicting the evolution of a “brain of brains” at the beginning of the 20th century. It seems evident to me that AI is just the pinnacle of a kind of process that has been going on for a long time, a process of ever greater complexification, cooperation and integration of living things and societies. If you think this is in any way plausible it throws up the question of the operand mechanism. Why does biological and cultural evolution show this tendency?
Teilhard de Chardin’s vitalistic answer to this question seems to be a poor fit with modern science. We naturalist Teilhardians need to come up with a better one! And here is where I start disagreeing with Wright: In his view, non-zero-sum dynamics, in conjunction with technological evolution, generates the vector of complexification. In my mentioned article I discuss in somewhat more detail why I think this does not work, but the general idea is that while non-zero-sum dynamics indeed occur, this by itself is not an explanation. It is just a formalization of conditions under which cooperation is helpful; it doesn’t explain why such conditions are becoming more frequent in the course of evolution in the first place.
My alternative explanation for why complexification occurs is metaevolution — the evolution of evolution. In the course of evolution, the medium of evolution itself evolves. The most obvious example of this is the emergence of memetic evolution next to genetic evolution. Memes are around because they allow us to adapt more efficiently as they are faster and more versatile replicators.
However, and now we come to the point of this post, I was kind of disappointed with the resulting argument because I could not give a nice and precise model for the metaevolutionary process going on. And while I still think that metaevolution is still waiting for proper general formalization, I think I can now give a first version of a simple computational model of the interaction between biological and cultural evolution in the emergence of ever more networked societies.
The general idea is that there are two kinds of replicator dynamics going on. Here are the assumptions (details can be gleaned from the repo):
There are a certain number of creatures. We hold that number fixed for simplicity’s sake. At each iteration of the simulation, each creature interacts with a certain number of other creatures.
Each creature has a preferred number of interactions per round. These are determined biologically (think prosocial instincts) and culturally (think social norms of cooperation like, of course, Zeus’ law). Each creature carries 100 binary genes and 100 binary memes. The preferred number of interactions of a creature is the sum over its genes and memes. So if you have 10 prosocial genes and 5 prosocial memes switched on, your preferred number of interactions is 15. All replicators are set to 0 at the outset.
Each round we pair creatures such that each one, as far as possible, has the number of interactions that it wants to have.
We assume that having 20 interactions is biologically optimal and generates the highest fitness. Imagine for instance, that less interactions means insufficient social support and more interactions tends to make social relations hard to manage.
Before introducing memetic dynamics, let’s have a brief look at the dynamics. As expected natural selection up close to the optimum of 20.
Now we stipulate dynamics for cultural evolution:
Memes have a very low mutation rate and can randomly flip. But they can also spread by contagion: Whenever two creatures interact each gets one meme from the other, chosen at random.
Memes spread faster than genes. For every 10 memetic update cycles, there is one genetic one.
We can now rerun the experiment:
Once an equilibrium is reached we see that the number of interactions happening is almost double of what it was before. Why? Because the memes that are prosocial are conducive to their own proliferation!
The simulation captures an instance of metaevolution because we are dealing with a conflict of two different orders of evolution. Genes and memes are in conflict here. While memes can never drive genes to extinction, because without biology there is no culture, it is still the case that memetic evolution leads to far more social interactions than would be biologically desirable. Of course, cultural evolution must still give some fitness payoff in biological terms, otherwise natural selection would just tend to favor creatures that are incapable of any kind of culture. Still, the model nicely illustrates that different orders of evolution can be in conflict, or at least tension, with each other.
A slightly unintuitive side effect of this is that cultural evolution is likely to make us less genetically disposed towards prosocial behavior. In the image on the right you see that the genetic preference is close to zero now because biological evolution tries to mitigate the fitness-decreasing tendencies of cultural evolution.
I suspect this kind of model is a far better intuition pump for grasping civilizational integration than mere non-zero-sum dynamics.
Teilhard de Chardin
This is a crosspost from my subtack.
In a recent article I discussed Robert Wright’s new book. His general thesis is that we ought to frame artificial intelligence cosmically, as a kind of evolutionary phase shift that will end up turning our civilization into some kind of global brain. We just have to decide on the kind of global brain we want to have.
I like Wright’s framing of the issue and I even sympathize with his invocation of Teilhard de Chardin, the maverick priest who started predicting the evolution of a “brain of brains” at the beginning of the 20th century. It seems evident to me that AI is just the pinnacle of a kind of process that has been going on for a long time, a process of ever greater complexification, cooperation and integration of living things and societies. If you think this is in any way plausible it throws up the question of the operand mechanism. Why does biological and cultural evolution show this tendency?
Teilhard de Chardin’s vitalistic answer to this question seems to be a poor fit with modern science. We naturalist Teilhardians need to come up with a better one! And here is where I start disagreeing with Wright: In his view, non-zero-sum dynamics, in conjunction with technological evolution, generates the vector of complexification. In my mentioned article I discuss in somewhat more detail why I think this does not work, but the general idea is that while non-zero-sum dynamics indeed occur, this by itself is not an explanation. It is just a formalization of conditions under which cooperation is helpful; it doesn’t explain why such conditions are becoming more frequent in the course of evolution in the first place.
My alternative explanation for why complexification occurs is metaevolution — the evolution of evolution. In the course of evolution, the medium of evolution itself evolves. The most obvious example of this is the emergence of memetic evolution next to genetic evolution. Memes are around because they allow us to adapt more efficiently as they are faster and more versatile replicators.
However, and now we come to the point of this post, I was kind of disappointed with the resulting argument because I could not give a nice and precise model for the metaevolutionary process going on. And while I still think that metaevolution is still waiting for proper general formalization, I think I can now give a first version of a simple computational model of the interaction between biological and cultural evolution in the emergence of ever more networked societies.
The general idea is that there are two kinds of replicator dynamics going on. Here are the assumptions (details can be gleaned from the repo):
Before introducing memetic dynamics, let’s have a brief look at the dynamics. As expected natural selection up close to the optimum of 20.
Now we stipulate dynamics for cultural evolution:
We can now rerun the experiment:
Once an equilibrium is reached we see that the number of interactions happening is almost double of what it was before. Why? Because the memes that are prosocial are conducive to their own proliferation!
The simulation captures an instance of metaevolution because we are dealing with a conflict of two different orders of evolution. Genes and memes are in conflict here. While memes can never drive genes to extinction, because without biology there is no culture, it is still the case that memetic evolution leads to far more social interactions than would be biologically desirable. Of course, cultural evolution must still give some fitness payoff in biological terms, otherwise natural selection would just tend to favor creatures that are incapable of any kind of culture. Still, the model nicely illustrates that different orders of evolution can be in conflict, or at least tension, with each other.
A slightly unintuitive side effect of this is that cultural evolution is likely to make us less genetically disposed towards prosocial behavior. In the image on the right you see that the genetic preference is close to zero now because biological evolution tries to mitigate the fitness-decreasing tendencies of cultural evolution.
I suspect this kind of model is a far better intuition pump for grasping civilizational integration than mere non-zero-sum dynamics.
Subscribe to my subtack