Questions about Epistolution
These are some of the questions that came up from my last article, and my best attempts to answer them.
Atrophy is defined as the shrinkage of cells due to loss of organelles and other materials. Is that really what you mean to present as the opposite of sleep?
No, it is not. As far as I know, medical science does not have a word for the gradual dissolution of a functional whole system into a non-functional whole system for the purpose of introducing creative novelty into its structures. What I am talking about could be called “phenotypic mutations,” or even “stochasticity.” It presents as a combination of inflammation, forgetfulness, atrophy, and generalized poor coordination between the structures and functions of the body. The important feature is that it is the opposite of cognition. Because there is no word for this, I made up the word “blundersplat” to describe it, but this seemed to confuse everyone even more, so now I am calling it atrophy, since that conjures an image people can relate to a little better.
Am I right in thinking that your theory predicts cave dwelling and deep sea creatures should still sleep about as much as other creatures, and if this turns out not to be true then your theory would be refuted?
I think a better test might be to take a species like this cave fish Astyanax mexicanus, which has evolved to suppress its previously evolved mechanism to sleep by producing more of this peptide HCRT, and place a population of them in a very challenging cognitive environment (perhaps still underground in a cave), while inhibiting HCRT in a selected sample from the group (without damaging them in other ways). This should give those selected fish the ability to sleep normally again. If the sleeping group can outperform the sleepless group in learning, my theory would be supported, while the reverse or no effect would refute it. Of course the reverse experiment, preventing sleep in sighted, sleep-dependent organisms like ourselves in complex environments, has been done over and over and it does interfere with learning.
The removal of light, for an organism that has long evolved to conduct its cognition with the use of light, seems to me to be a significant diminishment of the possibility for pattern-processing from its niche. Its niche not containing light, for its body plan, makes that niche very pattern-poor. To get this it might help to use the metaphor from the Santiago theory of cognition. In The Tree of Knowledge, Maturana and Varela argue that a nervous system is like a submarine operator who never knows he is underwater. He simply looks at his charts and turns the dials of his submarine controls, adjusting this and that to maintain the most important readings within certain critical ranges. For him, there is no “inside” and “outside.” The operator, the submarine, and the parts of the sea that appear on his readings are one seamless interconnected set of interactions. Anything in the sea that doesn’t appear on his panel doesn’t exist for him. They would say that the “distinction between the organism and its environment exists in the domain of the observer.” The way I would say this is that the niche is the outer boundary of the living system, and there really is no inner boundary. Making a grand distinction between the organism and its niche is like trying to make a big distinction between the circulatory and endocrine systems, and study each one in isolation.
You might object that blind people have relatively normal levels of intelligence. That is a good point, but I’m not entirely sure this is as good a test case as the blind fish, since so much of our intelligence (our niche) is acquired from other people. Also the human brain in blind people has been shown to transfer other inputs to the visual cortex, like hearing and smell, so that cognition proceeds relatively normally. A less cognitively adaptable species would actually pose a better refutation of my theory.
Humans are clearly capable of intelligent learning over periods that don’t involve any sleep, e.g. someone can learn about something over the course of a day and apply that knowledge the same day.
I think I disagree with this. I think that the knowledge (set of connections) that exists in the human body gives us the ability to apply ourselves to problems that appear in front of us by focusing on them. This does not necessarily require making any new connections immediately. We may learn a new idea, in the sense that we can go back over something we just heard and apply it to something we see, or something we already knew. This process could be just recursively tracing over again threads of trauma in our brains that just occurred, or pre-existed, and then drifting from them to other patterns that pre-existed. But over the long term, these recursive actions would tend to degrade our neurons if we didn’t shut down and support the permanence of the connections somehow. Maybe this is why it appears that we are learning while awake, while I’m insisting it must be when we are asleep that we really cement learning. That seems to be how most neuroscientists describe it, that sleep “consolidates” learning.
Why do “trauma” and “repair” necessarily have to happen in distinct stages, rather than happening in parallel?
This is really the gist of my argument. This is the fundamental question: why on earth do we have a consolidated shutdown of our whole bodies to repair? Why do we sleep? Why can’t we repair on the go? The answer must be that whatever sleep does, it requires the whole body to change. This leads to the insight that cognition is just a way of connecting the various processes of the body as a whole. Learning is just a way of linking all the actions of the body together in a certain way, including the actions of neurons. This is the type of insight that could easily evade modern science, since it often ignores wholes and looks at parts in detail. So we must repair the body first (NREM), and then repair the connections (REM). If we did the reverse, then we would end up with connections that did not quite match the structures they were connected to, and the whole system would function in a less coordinated way. And that is in fact the order that our sleep cycles (and all sleep cycles in animals, I think) follow.
Why can’t you just test your theory with a simulated agent in a simulated environment?
This is the hardest part of my theory to “see.” This is the key insight I am really trying to get across, and it has basically not worked for almost anyone, and I have been trying with over a hundred different people for over a year, so believe me if you are asking this you are in good company. What I am arguing is that intelligence is a function of the niche, the outside of the envelope the organism is in, so to speak. Of course it will be shown that intelligence is supported by the structures and processes of the body, including DNA, but the development of intelligence is a synchrony with the niche as a whole, and can’t be reduced to singular parts or processes. If this is the case, it will require that we sync the machine to our own niche, not to a simulated niche. I don’t think the relevant features of our niches can be simulated, because we don’t understand them at a quantum level yet. But this question still could be right. I simply have no frame of reference for how complex these simulated computer-generated niches can be. I know that the real inputs to the human body, and the body’s real components, and their real interactions, in our real niche are so unimaginably vast and hard to compute that one millisecond of a human life should surely require more processing power than we have created in all the world to “simulate” it. From this perspective it seems incomparably easier to just approximate the body with a robot rather than approximating the niche. Think about it. Using the submarine example…is it easier to just build a simple submarine and put it in the sea that already exists, or to model the behavior of every atom of a submarine and a sea? I’m not sure.
When an organism is being repaired by sleep, what comprises the set of configurations that it is aiming for? What’s being repaired to what?
I’m definitely not arguing that the body has a pre-determined shape that it should be in, and trauma is any deviation from that shape. This question is related to what is a problem, and what is problem-solving. I’m arguing that instead of there being any predetermined shape, there is simply a process by which a living system tries to maintain homeostasis. Any problem is a problem for homeostasis. Anything that threatens to undermine the process of homeostasis is a problem. So then I’m making the step to say that the explanation for the structure and process of an organism is that it anticipates certain problems coming from the niche and heads them off. In this sense the body of an organism is an expectation. The ability to change behavior in the light of new experience (intelligence) is the ability to create new expectations when some of the old ones are refuted. How are they refuted? Something unexpected happens. So physiology and epistemology are the same project.
There is a really fascinating take on all this in the Santiago theory of cognition. See the submarine metaphor above.
I think that the Santiago theory is really much closer to the truth about living systems than anyone has realized. The problem is it’s very hard to see the world the way they are seeing it, so once the authors went quiet, everyone just stopped trying. Since Plato’s Allegory of the Cave, we are accustomed to distinguishing between objective and subjective reality, and trying very hard to “live” in objective reality. But this is contrary to the empirical facts about physiology. The photons that strike my eye are not your photons, and the sound waves that strike my eardrums are different from yours. Everyone lives in a sealed envelope of physical interactions that comprise the outer boundary of their organismic system, and anything outside that system does not exist for them. This is very hard to visualize. The skin and shape of an organism’s body is easy to see, while the shape of its niche, for the observer, will always remain imagined. That’s why many neuroscientists are on a futile search for “engrams,” and representations, and things like that in the brain.
The Turing principle describes the universality of computation (which generally wipes out hardware concerns.) Why does this not apply or miss the point?
This explanation has been hard for me to grapple with in these discussions because I don’t really understand why computation applies to explaining life or to explaining intelligence. As close as I can get is to understanding that they think of an organism as a computation of some sort, that is itself doing computations. But David Deutsch himself argues that reductionist explanations, like this one, are not always necessary. My favorite is his example of the copper atom in Churchill’s nose, which he put in both his books. I actually have a totally different way of putting it together.
I would make a distinction between information, which I would call “alphabetical knowledge” and knowledge in general, which I would call “processible patterns” or something like that. Claude Shannon wrote about information that “the significant aspect is that the message be selected from a set of possible messages.” If we distinguish very sharply between messages like this, which are composed in alphabets, and all the other messages in the world, which are simply patterns we can detect, then some things become clear. Put all the alphabetical messages on one side, everything that can be precisely error-corrected in translation: the letters, symbols, words, bits, pixels, and, most significantly, the base pairs of the DNA alphabet, the genome. Then put on the other side all the significant patterns in our niche that are non-alphabetical: facial expressions, waves on a beach, the image of a tree on your retina, the pressure of the earth on the soles of your feet, the firmness of a handshake. Humans are the only species that really use external alphabets, but all species use internal genetic alphabets. Think about the way those alphabetical pieces interact with the living system, as opposed to the other forms of patterns.
A living system, when it finds a pattern, tries to adjust to it, to maintain homeostasis. In fact, a living system takes its niche as a whole in every moment, and works out how to survive based on the totality of these patterns. If you are walking across dunes, you modulate your ankle postures so that you remain upright. If you pass an overhanging limb, you duck your head. No alphabets were involved. The living system absorbs or reacts to the patterns, in other words, to keep itself whole. When a living system encounters alphabetical information, it likewise uses it as a tool to maintain homeostasis, but there is a difference. The alphabetical information, since it can be perfectly error-corrected in transmission, can travel from one living system to another and be re-translated into the physical “language” of the new system. If you write STOP on a sign, it can enter the niche of every human who comes along in much the same way, unlike, for example, a painting of haystacks. This means that these alphabetical characters (memes, or genes) can replicate. And the use of these alphabetical tools have upward consequences, that affect the development of the ontogeny of the organism. Genes for legs allow you to walk. If you learn to stop at stop signs, we can trust you with the car keys.
For me, it takes both of these forces in combination to create competition. The homeostatic, top-down force is primary, because it can exist without the informational bits, though sometimes not very well. For example, the niche can exist without the organism, but not vice versa. An enucleated cell could live, and even divide, but naked DNA cannot act at all. But top-down is always looking for tools, to help it maintain homeostasis. It wants to eat these tools like food, to incorporate them into its system. Since alphabetical tools can be replicated and passed along, this creates a competitive dynamic. In order to be sustained, competition requires both that the winners prevail, and that the winning information be shared. If no one wins, competition will come to an end. But likewise, if the successful information isn’t shared, competition will come to an end.
So computation, to me, is just organisms using these tools to construct various structures inside our niches. We can all participate in the benefits of these problem-solving structures, because all alphabets have universality, as in they can be translated into sets of reactions in our bodies that produce similar effects. But without the homeostatic force, there is really no knowledge creation. In order to create knowledge, a system (living or not) must be trying to maintain homeostasis in a niche. And in order for us to perceive and share in that knowledge, it must be describing something that helps us in our own niche, otherwise it won’t look like knowledge at all to us. Anything outside our own niche doesn’t exist for us at all. So right now the only way to create knowledge is either to create patterns that help solve a homeostatic problem for a living system (like a bird dog pointing to a bird while hunting) or to make alphabetical figures that do the same thing (like a mathematician writing equations that predict eclipses, or some such thing.) I don’t know if this jibes with Deutsch’s theory of what is computable or not.
I rather suspect it does not. It seems to me like everyone in this field is thinking they can build knowledge just by manipulating alphabets, without maintaining homeostasis. But I see a similarity between my description of these epistemological dynamics and the construction of every intelligent nervous system that has ever evolved. Iain McGilchrist’s amazing book The Master and His Emissary makes this clear. They all have an internal division between a structure that deals with the holistic flow of the whole niche, and another structure or structures that deals with manipulating important details. McGilchrist argues that the holistic structure is primary, and I am inclined to agree. The detail structure is the one that does computation, of course, because computation is a bottom-up process. The equivalent top-down process is what the Santiago theorists call “structural coupling,” which is basically adjusting one’s internal processes to minimize surprises on the whole. It is “getting in sync” with one’s niche. Of course this requires a homeostatic process, and a niche.
Why couldn’t one, non-bifurcated brain just devote part of its effort to holistic attention and another part of its effort to small-scale, detail-oriented attention?
Because if that were the case, then evolution would have produced one, and as far as I know, it has not. I feel like I am on really firm ground here. The octopus example was my own, but that, too, I think is firmly supported by the work that looks at cephalopod cognition. It would be much more efficient to process the world using only one brain, and computer AI research has assumed that this is possible. But Nature has never done that. This must be because there are two extremes in the way we can look at things…extreme reduction to one thing and extreme holism devoted to all things. We require a brain with access to some sort of full physical body to accomplish either extreme.
