Triple Tragedy And Thankful Theory
A guest post by Daniel Böttger
I accept guest posts from certain people, especially past Book Review Contest winners. Earlier this year, I published Daniel Böttger’s essay Consciousness As Recursive Reflections.
While we were working on editing it, Daniel had some dramatic experiences and revelations, culminating in him developing a theory which he says “will contribute to saving the world”, which he asked me to publish.
Although I can’t speak for its world-historical importance, and although he admits his mental state is fragile, after some discussion I decided to publish because - if nothing else - he’s a great writer with a fascinating story and some really interesting thoughts.
Content warning for medical horror; you can skip to the section “Thankful Theory” to avoid this.
I. Triple Tragedy
2nd of July, this year. I awaken to a strange man telling me that he just pulled me from the car that I have crashed into a tree at 100 kilometers an hour. I must be missing several minutes of short-term memory, because dozens of first responders are standing around me, reassuring each other and myself that my three kids, who were in child safety seats in the back, are fine. On the way to the trauma response unit, everyone agrees that I had a massive epileptic seizure. Utterly implausible; I have never had one as far as I know. But the absence of skid marks and the description of events my front seat passenger gave, plus the presence of the kids as evidence against suicidal intent, leaves this as the only explanation.
The hospital neurologist tells me I have a tumor in my right temporal lobe, which would have produced seizures. They found it in the CAT scan when they were looking for fractures. I say this must be a mistake, I have a [small] focal cortical dysplasia in my right temporal lobe. Very memorably, he says it is much bigger than that.
They release me from the hospital saying that cancer treatment needs to proceed much nearer to where I live. They give me a DVD with the scans. I am not a neurologist but I can look at images. The thing is bigger than a chicken egg! From this point on to the end of this narrative, I am constantly feeling intense fear for my life. It only varies in intensity enough to prevent desensitization.
Despair amplifies my responses to all negative stimuli. The urgency of any task rises to terrifying levels. Any pain becomes so excruciating that it wipes my mind blank and destroys whatever solutions I have been attempting to build. My mind becomes prone to collapse given any excuse. Survival becomes a matter of avoiding such collapses. Mercifully, at least the basic rules such as honesty and kindness continue to apply, and the methods of rationality continue to work. This limits the confusion, any trace of which has become excruciating as well.
An EEG confirms heightened epileptic potentials in my right temporal brain. Thinking back, I did have strange alterations of consciousness, but they were diagnosed as post-traumatic dissociations. I had previously described these alterations to competent friends and they had said my descriptions sounded epileptic. I had told my psychiatrist this; she had disregarded this idea. Understandable, I did have intense trauma from when my ex-wife had attacked me. I learn that seizures intensify with hyperventilation. I have never had an intense seizure, but in retrospect, these alterations may indeed have been mild seizures, because hyperventilation during the car ride is plausible, because I am not a proficient driver and the kids were being loud. The car I destroyed was not even my own, it belonged to my (officially ex-)parents-in-law.
My (officially ex-)father-in-law was the front seat passenger. Same story with him: put into the CAT scanner to look for fractures, and although he never had seizures they found a brain tumor in him as well. Utterly implausible, a literal one-in-a-million situation: that’s two out of two CAT scans finding undetected brain tumors, each with roughly a 0.1% prior probability from the epidemiology of undetected tumors in Germany. They've taken it out of him since, he is better now.
Scott publishes my first guest post. Analgesic joy provides my first respite in weeks. The early comments amount to “Hooray physicalism!” and “Boo physicalism!” Analgesic joy fades, I am returned into the existential horror of my life.
10 years earlier I had an MRI scan that showed something strange in my right temporal lobe. They said it was a focal cortical dysplasia and I should not worry about it. In retrospect, it is located in the exact middle of where my big tumor is now. So it seems it was that tumor at the size it had back then. In the present, both scans are studied by two separate professionals, who both confirm my amateur fears. The anomaly in the old scan is large enough to have been growing for 2 years by that point. I seem to have lived with a brain tumor for 12 years. Median survival time for the least aggressive brain tumors is 5 to 10 years - 12 years and alive, and my not even noticing the thing until days ago, is utterly implausible. I did do a lot of ketogenic and carnivore dieting; there is no evidence in the literature of this slowing the growth of glioma in particular, but there is such evidence for many other types of tumor and it is reasonable to assume it would generalize. Still, how could something literally this big have been happening without me realizing? My world is falling apart, I need an explanation to make sense of these utterly implausible observations.
Pattern match! A catastrophic realization comes crashing in. The necessary expertise has been available all along, from my academic degree in the psychology of religion. I did have seizures for years, I just didn’t realize it because of the misdiagnosis, and they came from my (right-side) temporal lobe. Temporal lobe epilepsy is strongly associated with mystical experiences, which I did have despite my atheist convictions, and with what is misleadingly termed “hyperreligiosity” but really means heightened interest in metaphysical questions: a temporal lobe epileptic can be an atheist who wants to talk to you for hours about why atheism is correct.
I have been doing way more than that by writing the Seven Secular Sermons - an atheist competitor to the Divina Comedia, the Bodhicaryāvatāra and the Mahābhāratam - for exactly the 12 years that I have evidently been hosting this tumor. And I did have a lot of headaches, which should have been my other clue, but which I had heroically/stupidly tolerated for years.
My official life’s work is a manifestation of cancer. I do not want to understand this, because it denigrates what is to me the closest thing to holy. But it is less implausible than all of this being a giant coincidence. I cannot pretend I don't believe it; the writings of Eliezer Yudkowsky have changed me.
At least thinking about this gives me analgesic joy, or it focuses energy consumption of my brain into particular areas, slowing the tumor growth and brain swelling, reducing the pain.
Neurosurgery is scheduled for the 16th of August. Survival is doubtful, because the thing had so much time to grow huge. Lasting neurological damage is likely. Although each of the first Sermons took months or years to write because of their highly constrained form, in a mad dash of activity the final Sermon (half-finished at the time of the car crash) and the video are complete in just a few weeks, so my life's work is done before I might die. With our joint task complete, the tumor should more easily depart. The idea is utterly implausible, that is not how causality works. How very right it feels is strong evidence of insanity.
I am flooded with assertions of assistance so massive that none of you would have managed not to cry, either. I am now primarily a patient, everyone agrees and I dutifully learn this role. I need to endure, maintain passivity, and thank a lot. As the relevance of my wishes fades, thanking becomes the only directional thing I can do, my only activity. I find myself thanking the fear and the pain for the lessons in humility, especially when it all gets so much worse at the hospital.
From the first few dozens of secretaries, doctors and nurses bouncing me between their areas of responsibility, I notice persistent puzzling problems in communication. The brain damage does not explain this, because I am right-handed: my speech comprehension should be mostly in the left hemisphere, where there is no tumor. (...yet. Tumor growth mindset.) And there is an obvious pattern. The trouble does not arise in all of communication, but specifically in collaborative problem-solving: the back and forth of problem-relevant pieces of information, and its unspoken but important implications about competence and priority. Miscommunications cause errors, all amplified by the extreme urgency of my problem. None of us have trouble solving problems by ourselves. Whenever these people think there is a minute for idle chat, that proceeds flawlessly. But the more urgent collaboration is, the more frequently it appears to fail, and only between them and me, not when they talk the same way amongst themselves. Several times these problems impede my treatment; every time, my unscheduled attempts to help turn out to aggravate the trouble instead.
Waking up from the surgery, I am told it went well. (Later I’ll learn this was untrue: much of the tumor remains, and there is lasting neurological damage.) The pain is so debilitating I can hardly think.
At the intensive care unit, I very nearly die. So many sensors and things are connected to me, they curl themselves into knots with every small movement, but I’m compelled to move to avoid bed sores. Eventually a cannula rips out of my forearm. The ICU nurses accuse me of being a suicidal idiot who must have done this intentionally. They save my life, while loudly proclaiming themselves more interested in saving their bed from the huge pulsing fountain of my blood. Their accusation of suicidality is an assumption of hostility. Although I used to work at this hospital (in research not care), I am tempted to assume hostility as well, but this is utterly implausible; their incentives are aligned with mine, my death would inconvenience us all. Trauma must be coloring my perception of events, so I keep assuming good intentions. They continue to assume hostility whatever I do. Still I need the cooperation of these women. Through intense pain and terror, I attempt to optimize my communication with the intensity of a fight for my life. Explicit thanks continue to help, especially because I keep failing to signal cooperativeness by collaborating the way I am used to.
I discover I need to talk to them the way that I need to talk to my mother, who is a highly experienced nurse, when she is managing another crisis. This is very counterintuitive, because my mother is the best mother I ever even heard of, while these women are the opposite of kind. But it is clearly working. I do not understand how, but it seems I will need to.
I am moved from intensive care to normal care, still struggling through intense pain and terror. The workflows I am dragged through are clearly, blatantly, unapologetically inefficient, producing abject horror in my efficiency-worshipping mind. This makes the people supposed to save me seem insane to me, but it is much more plausible that the insanity is mine. Collaboration keeps failing, and it is failing more often the more vital it is. As usual when that happens, each side is most likely to blame the other. But I can't believe these people are crazy or evil. Doctors and nurses can’t be dumb either; surely they don’t talk so impatiently and seemingly unthinkingly in their private lives. I am not evil either, but I might be going crazy with pain and maybe I’m dumb now because my brain is damaged? But I can still talk as correctly as ever, and I can still discover patterns. Of course I point the pattern recognition at why collaboration is so hard. They don’t like it when I volunteer relevant information! I habitually volunteer relevant information in order to cooperate and emphasize my cooperativeness, but here now this habit is counterproductive. They appreciate brevity; they say only what needs to be done, not why. They are assuming all communication to be about needs, not wants, and are expecting me to assume the same about their communication. I experiment with limiting myself to one word responses. This is Germany and these people are educated, the hierarchy of “Verstanden.” (“I have understood.”), ”Einverstanden.” (“I have agreed.”) “Versprochen.” (“I promise I will do that.”), “Erledigt.” (“I have done the thing you told me to do.”) may safely be presumed common knowledge. This is a breakthrough, the care I receive improves dramatically. But collaboration continues to fail occasionally, so I have not identified the root of the problem. And still they want me to participate in processes that I fail to comprehend, and still collaboration gets more difficult the more urgent it is. I think as hard as I can, with the intensity of a fight for my life (while very possibly going insane from the pain, and surely partially for the analgesic joy of the more focused energy consumption in my brain) about how we can collaborate better. I do eventually survive the hospital. Very good friends and my awe-striking nurse mother pick me up.
Under her guidance, using an embarrassing amount of painkiller prescriptions and much time and money, as well as lots of cellphones, we attempt to maintain a clinic-like level of care, without the multiresistant germs or the cost to my health insurer, in a home environment. Of course we cannot do radiotherapy or chemotherapy on the large remainder of the tumor, but they aren't doing that at the hospital anyway, because they're waiting on some biopsy process that's being inefficient. In what continues to resemble my home, we can at least do some basic care and prevention. I must participate. I can learn, and recognize patterns: do not set down steps heavily or drink any coffee, because any bit of pressure on the brain causes mind-wiping pain. I am proud to manage this, and embarrassed to continually fail to maintain an overview of non-obvious information such as how long ago I was in the bathroom. Collaboration is easier amongst ourselves, everyone’s goodwill and rank are obvious. Still there are millions of tiny collaborative interactions where minds need to meet, in order for pieces of information to meet, in order for urgent tasks to be completed. Lots of little ways these collaborations go right or wrong stream into me, all important, many highly informative.
Hundreds of kind messages on my phone. I continue to be in severe danger and intense pain, on the phone with doctors and nurses, and with many friends trying to help. I find myself communicating the exact way the nurses did: impatient with information that seems inessential, clearly inappropriately unkind. It makes me ashamed of myself. What is happening, what am I doing, what am I doing wrong?
Most confusingly, I am not doing it wrong. Failure or insanity on my part would be the easier explanation, but it does not fit the data. I cannot flee into the comforting excuse of excusable insanity and pretend not to believe what I believe. I can see the group chat messages of my friends and family: the nurses way of communication works much better in (this) crisis, the way of communication I consider normal and appropriate and efficient has clearly repeatedly caused damaging confusion, not just when I was the one who tried it. I get a lot of practice attempting to optimize my communication and actions accordingly, as I go through various subsequent crises. I begin writing this document, in an attempt to maintain the pursuit of knowledge that requires maintenance of my sanity, or at least documentation of its failure.
My health catastrophically fails again; They did not let us have a powerful immunosuppressant to stop my brain from swelling, which was available at the hospital but not allowed to be taken home to our makeshift clinic. Brain pressure spikes, there is talk of the tumor growing again or some new bleeding. I go back to the hospital, to the terror and to the attempt to think my way out. The situation there is similar to before, but now it is all just more relevant evidence, being provided for me to improve my understanding, until finally the most parsimonious explanation for all this traumatic data is to posit a new category: collapse-preventing systems, such as this hospital, are fundamentally different from efficiency-maximizing systems. Nurse talk is optimized for assuring survival, not thriving, and so is the previously incomprehensible logic of this hospital. I derive my communications and actions from my new theory, with spectacular success, compared both to before and to how my fellow patients are faring. Personal positive bias is the default explanation, but I compulsively continue to invest thought into this, for the analgesic joy if nothing else. Maybe all of this descent into darkness can be good for something.
Pattern match! Another catastrophic realization comes crashing in. The necessary expertise has been available all along, from my academic degree in computer science. 20 years ago, Algorithms 101. I receive what feels like a download of the following theory. I initially do not want to believe it; believing I could have stumbled into something new and so fundamental is the most obvious evidence of insanity yet. I am crying so hard it is impossible to tell whether the tears are of joyful relief or of my mind finally cracking. But whether this means I finally solved The Puzzle or whether it means I have finally gone insane, I cannot pretend I don't believe it; the writings of Eliezer Yudkowsky have changed me. Holy fucking shit, this is it. This explains everything. Thanks to the cancer!
New bleeding or new tumor growth is ruled out by another CAT scan, and we do finally get the immunosuppressant. Back at my mother's home clinic, I continue writing this. Writing carries me past levels of pain where I’m confident 90% of the population would have attempted suicide. I can’t do that to the kids, and my mother would see right through the pretend stupidity of an “accident”. But mostly I choose to live because I have Something To Return From This. And here it comes.
II. Thankful Theory
[An old Slate Star Codex post describes] systems trying to thrive [versus] systems trying to survive. Just as with the political left-right dichotomy that Scott described in these terms, this distinction between these types of systems is most obvious when both are in conflict with each other, such as when Silicon Valley startups complain that the regulations of the systems trying to ensure survival are stopping them from thriving, or when opponents of nuclear power proclaim themselves uninterested in the low price of this energy.
Such conflicts are terrible, because these two groups of systems should be natural allies:
We cannot thrive if we don’t survive.
Insufficient thriving causes slowdown and scarcities that might threaten survival.
Both types of systems are trying to be good, and they benefit from each other's function, yet they are so fundamentally different from each other that they have trouble collaborating. Both of them are fighting malignancy, another reason they should be allies, but therefore when they come into conflict and find each other to be so foreign, they are prone to assume malignancy. This is a tragic impediment to collaboration; we should all want a solution to this. But a solution requires locating the problem correctly.
The humble thing to do would be to
extend thrive/survive from its political domain only slightly, to a sociological or economic level, where we find the most publicized/obvious examples.
assume a “normal” problem of human incentives.
theorize some type of resilience/efficiency tradeoff in how we build our institutions (such as hospitals).
But this would not explain e.g. the near death at the ICU or the miscommunications in the home clinic text messages - individual humans with well-aligned incentives catastrophically failing to collaborate.
I believe that:
the problem is more fundamental, more fractal, more Moloch-like.
it occurs in conflicts between the immune system attempting to ensure survival through e.g. fever and the central nervous system attempting to thrive in ways that involve taking health risks.
it occurs in the conflict between defenders against artificial superintelligence (ASI) risk, trying to help humanity survive by stopping ASI from being built, and accelerationists trying to help humanity thrive by building ASI to efficiently solve climate change etc.
We would all be very lucky to find the following theory is right, because it implies solutions.
The most important thing to understand about trying to assure survival is that it is torturous. Even success necessarily involves constant exposure to catastrophe, even if only counterfactually in the service of prevention. Every emotion, including friendly unassuming gratitude, is heightened to painful intensity.
Human minds cannot deliberate under torture. This alone would explain why people doing torturous tasks should optimize against deliberation, for binary or multiple choice questions rather than open questions, for legibility rather than optimality. However, it is more generally true that the closer you get to collapse, the more urgently a survival-oriented system needs to act, the less capacity it has available to connect multiple pieces of information into a detailed model of what to aim for. Under the pressure of needing to perform their task unfailingly, both types of systems have converged into two different architectures, which are dictated by information theory.
This is the tense moment where most of you are impatient with how slowly I am getting to the point, while a minority has already matched the pattern and does not need me to spell it out. Here is a hint: the minority is among the people who understand Algorithms 101.
This is the theory: Survival-oriented systems are trying to be space-efficient rather than time-efficent, because they are at constant risk of mental memory (especially the phonological loop) getting overwritten by catastrophically urgent new data, and because this risk is strongly positively correlated to how urgently they need to act. The rest is commentary.
For those outside that minority: there are different ways you can implement any algorithm. For example, imagine you have to sort very many books into a large library, alphabetically. It is a private library, so you can't just make a public servant do it, but need to instead do it cleverly, methodically, and make a method you’ll use. This method is an algorithm.
Two of your options for how to go about this would be the following. Both will work.
Take a book from the unsorted pile, compare it to a sorted/shelved book. If the new book is closer to A than the sorted one, go left on the shelf, otherwise go right. Repeat until done.This method (algorithm) obviously does not require extra space for the temporary shelves of the 2nd, so we call this kind of method (algorithm) “space-efficient”.
Sort each book into an individual sorted shelf of a few books. Then merge the shelves. Repeat until done. This needs more space, but it needs much less time to complete, because it requires much fewer comparisons between individual books, so it’s more “time-efficient”. This is a mathematical fact that any computer scientist will, and any amateur coder can, confirm.
Apply this to communication that is required for collaboration. Space-efficient communication can’t cache a long message to be communicated, or a long message received to be understood. Therefore space-efficient communication has to rely on short, atomic messages, which in order to be informative have to be pre-agreed. This must necessarily take more time (be less time-efficient) because the information that needs to be transported must laboriously be translated into, and from, the pre-agreed alphabet of simple, atomic messages.
Both the survival-oriented systems and the thriving-oriented systems are trying to solve problems. For this purpose, the intelligent ones are attempting to improve themselves. As they do this, over time, they each apply their methods on themselves and their understanding of their situation, so their different methods produce even more different concepts of themselves and their environment. This creates misunderstandings and confusions between the two types of systems, and that’s how the inevitable, logical difference between time- and space-efficient algorithms leads to stubborn conflicts that impede collaboration.
IIb. How to improve collaboration between survival-oriented and thriving-oriented systems
There are existing computer science results about how to integrate space-efficient and time-efficient algorithms. Whatever the specific practical details of the malfunctioning collaboration that you want to improve, these results will be applicable. This is the principled solution!
In the context of this general essay, and assuming each system includes humans, I will now provide seven heuristics. If these don’t help, that is experimental evidence against this theory. But if you give non-negligible credence to the theory, the expected value of trying them is positive.
Thank each other a lot; explicitly affirm that you understand each other's good intentions. Within your home system, adherence to your communication norms sufficiently establishes your willingness to collaborate, but this does not work across competing communication norms, so explicit assertions of willingness to collaborate must serve as a substitute.
If you are trying to assure thriving, and frustrated by a party trying to assure survival:
Understand and affirm that preventing collapse is desirable in order to avoid the efficiency loss of collapse.
The brevity of their communication will feel hostile; interpret it charitably as an expression of urgency of concern.
Your problem-solving will be misunderstood and underappreciated; keep forgiving this.
If you are trying to assure survival, and frustrated by a party trying to assure thriving:
Understand and affirm that thriving is desirable in order to prevent scarcities, which cause collapses, which threaten survival.
The voluminousness of their communication will feel distracted and unfocused; interpret it charitably as an attempt to provide especially much information that might become more obviously relevant during the future of your alliance.
Your discipline will be misinterpreted as hostility; keep forgiving this.
All of this will be least intuitive, but most needed, when the matter under concern is most urgent.
Share this post