For what it’s worth, my personal judgement is that the filter lies before the creation of a central nervous system.
Why? Having dabbled a bit in evolutionary simulations, I find that, once you have unicellular organisms, the emergence of cooperation between them is only a matter of time, and from there multicellulars form and cell specialization based on division of labor begins. Once you have a dedicated organism-wide communication subsystem, why would it be unlikely for a centralized command structure to evolve?
My personal guess would be that the great filter isn’t a filter at all, but a great scatterer, where different types of optimizers do not recognize each other as such, because their goals and appearances are so widely different, and they are sparse in the vast space of possibilities.
Why? Having dabbled a bit in evolutionary simulations, I find that, once you have unicellular organisms, the emergence of cooperation between them is only a matter of time, and from there multicellulars form and cell specialization based on division of labor begins. Once you have a dedicated organism-wide communication subsystem, why would it be unlikely for a centralized command structure to evolve?
On Earth multicellularity arose independently several dozen times but AFAIK only animals have anything like a central nervous system.
If animal-complexity CNS is your criteria, then humans + octopuses would be a counterexample, as urbilaterals wouldn’t be expected to have such a system, and the octopus intelligence has formed separately.
The last common ancestor of humans and octopuses probably didn’t have a very complicated nervous system, but it probably did have a nervous system: most likely a simple lateral cord with ganglia, like some modern wormlike animals. That seems to meet the criteria for shminux’s “dedicated organism-wide communication subsystem”.
My personal guess would be that the great filter isn’t a filter at all, but a great scatterer, where different types of optimizers do not recognize each other as such, because their goals and appearances are so widely different, and they are sparse in the vast space of possibilities.
See James Miller here. Sure, the space of possible value systems is vaste, but I doubt that much less than (say) 0.1% of them would lead agents to try and take over the future light cone, so this could at most explain a small fraction (logarithmically) of the filter.
Having dabbled a bit in evolutionary simulations, I find that, once you have unicellular organisms, the emergence of cooperation between them is only a matter of time, and from there multicellulars form and cell specialization based on division of labor begins.
I’m very curious: in what evolutionary simulations have you seen these phenomena evolve?
If I recall, you start with a single cell like an amoeba, which has to be smart enough to not accidentally eat its own pseudopods, so the relevant mutation sticks, and results in it also not eating its clones and other cells of the same type. This only sticks if there is enough food around so that there is no competition between them. This is how you get cooperation with the same kind. At this point the mutation disappears if you reduce the food supply, as defection (evolving cannibalism) becomes the dominant adaptation. However, if you provide the right conditions for the collections of cells (colonies) to win over single cells (because feeding in packs gives you better odds of eating vs being eaten), then the simple defections do not stick, as single defectors lose to colonies of cooperators. The most fit organisms are those which create colonies right away, with each division, not waiting for a chance to cooperate.
Once you have cell colonies competing, the division of labor is next. A relatively simple mutation which lets a cell to become either a hunter, if it is outside-facing, or a food processor, if it is surrounded by the same kind during the first part of its life, is a simple model of how cell specialization might appear. Colonies with two kinds of dedicated cells are more efficient and win out. And so on. The immune system also appears naturally, as hunter cells already perform this role.
The models above are, naturally, a gross oversimplification, but they show how the multicellulars could evolve. The simulation code itself is almost trivially simple, I can probably dig it out at some point. I don’t recall doing much more than what I’ve described, but presumably a communication subsystem would increase genetic fitness, eventually resulting in the appearance of the nervous system. I kind of lost interest when it got overly complicated to code. I bet there are people out there who do this for a living.
I bet there are people out there who do this for a living.
There’s this Ph.D. thesis and video all about multicellular coordination in slime moulds. This is an organism which switches between unicellular and multicellular habits, with cell specialisation, and competition among cells to be among those that make spores and get their genes into the next generation.
Why? Having dabbled a bit in evolutionary simulations, I find that, once you have unicellular organisms, the emergence of cooperation between them is only a matter of time, and from there multicellulars form and cell specialization based on division of labor begins. Once you have a dedicated organism-wide communication subsystem, why would it be unlikely for a centralized command structure to evolve?
My personal guess would be that the great filter isn’t a filter at all, but a great scatterer, where different types of optimizers do not recognize each other as such, because their goals and appearances are so widely different, and they are sparse in the vast space of possibilities.
On Earth multicellularity arose independently several dozen times but AFAIK only animals have anything like a central nervous system.
If animal-complexity CNS is your criteria, then humans + octopuses would be a counterexample, as urbilaterals wouldn’t be expected to have such a system, and the octopus intelligence has formed separately.
The last common ancestor of humans and octopuses probably didn’t have a very complicated nervous system, but it probably did have a nervous system: most likely a simple lateral cord with ganglia, like some modern wormlike animals. That seems to meet the criteria for shminux’s “dedicated organism-wide communication subsystem”.
See James Miller here. Sure, the space of possible value systems is vaste, but I doubt that much less than (say) 0.1% of them would lead agents to try and take over the future light cone, so this could at most explain a small fraction (logarithmically) of the filter.
That quote doesn’t rule out it being before the earliest thing you mentioned.
Right, of course. My mind inserted “right before” without a good reason.
And I implicitly did mean “right before”, so your reading was correct.
I’m very curious: in what evolutionary simulations have you seen these phenomena evolve?
If I recall, you start with a single cell like an amoeba, which has to be smart enough to not accidentally eat its own pseudopods, so the relevant mutation sticks, and results in it also not eating its clones and other cells of the same type. This only sticks if there is enough food around so that there is no competition between them. This is how you get cooperation with the same kind. At this point the mutation disappears if you reduce the food supply, as defection (evolving cannibalism) becomes the dominant adaptation. However, if you provide the right conditions for the collections of cells (colonies) to win over single cells (because feeding in packs gives you better odds of eating vs being eaten), then the simple defections do not stick, as single defectors lose to colonies of cooperators. The most fit organisms are those which create colonies right away, with each division, not waiting for a chance to cooperate.
Once you have cell colonies competing, the division of labor is next. A relatively simple mutation which lets a cell to become either a hunter, if it is outside-facing, or a food processor, if it is surrounded by the same kind during the first part of its life, is a simple model of how cell specialization might appear. Colonies with two kinds of dedicated cells are more efficient and win out. And so on. The immune system also appears naturally, as hunter cells already perform this role.
The models above are, naturally, a gross oversimplification, but they show how the multicellulars could evolve. The simulation code itself is almost trivially simple, I can probably dig it out at some point. I don’t recall doing much more than what I’ve described, but presumably a communication subsystem would increase genetic fitness, eventually resulting in the appearance of the nervous system. I kind of lost interest when it got overly complicated to code. I bet there are people out there who do this for a living.
There’s this Ph.D. thesis and video all about multicellular coordination in slime moulds. This is an organism which switches between unicellular and multicellular habits, with cell specialisation, and competition among cells to be among those that make spores and get their genes into the next generation.