Epistemic status: interesting theory, not sure if true
TL;DR: Natural selection would favor the capability to change subjective probabilities in MWI as it significantly increases observed survival odds. Such manipulation could be implemented through a "Universal Lottery Winning Machine" (ULWM) that changes the number of observer-copies experiencing different outcomes. Brain can implement ULWM via delayed observation of the desired outcomes as delayed measurement will have higher measure in MWI. This can be tested via simple computer program, results inconclusive.
This post explores the following argument:
- If probability manipulation in MWI via the change of the number (and measure) of copies observing different outcomes is possible, evolution would favor minds capable of it, as they would have higher survival rates.
- Even small probabilistic advantages compound across generations, leading to the dominance of such minds and such worlds where it is possible.
- “Quantum suicide”, "The Anthropic Trilemma" and similar thought experiments describe a theoretical mechanism for changing subjectively observed probabilities through varying the number of observer’s copies based on the measurement outcomes.
- A biological brain might implement this mechanism as follows:
- A conscious mind enters a brief pause between observer-moments (~0.1 sec), corresponding to pause between normal brain waves.
- The unconscious part of mind evaluates incoming information about, say, a coin toss.
- For favorable outcomes, it slightly delays regaining of consciousness.
- Due to constant MWI branching, delayed observation produces more exact copies which don’t know yet the result of the coin toss.
- Applying Self-Sampling Assumption (SSA) to all exact copies, we get that I am more likely to be in the wining branch.
- This implementation suggests to count copies numerically, use SSA and ignore copies’ declining measure, which is a significant weakness. Other implementations of biological ULWM are possible under different assumptions. For example, if the total measure of the Universe is declining because of Youngness paradox, when the measurement delay results in less chances of that outcome.
- We are more likely to be in the world where biological ULWM is possible if it is based on the increase of the number of copies which observe desired outcomes as such worlds accumulate higher measure. Instead, Quantum Suicide’s mechanism assumes the decrease of the number of copies who observe undesired outcomes – to zero. This will decrease the total share of the worlds where QS works.
- Biological ULWM which helps to survive in the situations of danger or in search of the best mate will have the biggest impact on fitness. The observational consequences should be most prominent in my ancestors’ survival and the speed of their evolution.
This theory offers several explanatory advantages:
- Explains the function for consciousness (it is needed to fix the exactness of copies that perform the observation).
- Requires no novel brain mechanisms, as observer-moments are naturally discrete and the speed of their appearance is stress-dependent.
- Explains why controlled objective parapsychological experiments fail despite participants' subjective credence that they work: only subjective probabilities change.
- Explains the speed and seemingly impossible jumps in evolution without rejecting natural selection as its main mechanism.
The capability likely evolved gradually, suggesting humans may be not only the most intelligent but also the most "magical" organisms.
Future technological exploitation of this capability could yield vast improvements but also novel risks and weapons.
This theory should not be confused with other similar theories: magic by forgetting, quantum immortality, doomsday argument, future anthropic shadow, transcendental advantage and anthropic principle as retrocausality (not published). While these theories all suggest that subjective probabilities may differ from objective ones due to observer-selection effects and this can be exploited, their mutual compatibility remains unclear.
1. Introduction
This analysis assumes the classical many-worlds interpretation (MWI) of quantum mechanics as described by Deutsch (2002), with real, continuously branching worlds. While this represents a simplification of Everett's original relative states interpretation, we will follow Deutsch's branching-worlds model for clarity.
Another (probably not necessary) assumption is that the total measure of the universe is exponentially declining. This is the idea behind Youngness paradox by Guth and such decline is explained by exponential growth of the number of young universes because of cosmic inflation.
2. Manipulation of observed probabilities through copy number variation
The Universal Lottery Winning Machine based on quantum suicide
The concept of quantum suicide can be extended to create a variant of Universal Lottery Winning Machine (ULWM), consisting of a conscious observer and an unconscious winning-measurement system with a bomb:
1. The unconscious system checks the lottery result first.
2. If there's no win, it triggers the bomb, eliminating the conscious observer.
3. Thus, the observer only experiences winning outcomes in MWI and subjectively always win (but objectively – form the third point of view – loses as often as dies).
ULWM by boosting measure via copies
Yudkowsky in the The Anthropic Trilemma proposed a non-quantum (and no suicide) version of ULWM using digital consciousness. His thought experiment:
1. Upload a mind to a computer
2. Create 1000 copies on the mind in the computer.
3. If loosing is observed by some external controller, preserve one mind copy; for wins, maintain all copies.
4. Continuously merge all copies into one mind.
Assuming equal probability of being any copy, this creates a 1000:1 shift in subjective winning probability. Rejecting this conclusion leads to logical inconsistencies in probability theory or personal identity.
General principle of ULWM which uses self-sampling assumption
The algorithm assumes state-based identity theory, where personal identity depends solely on current state rather than historical path. We will use it as a main model. (Path-based ULWM is also possible).
1. Preparation Phase
- Conscious mind places bet
- Enters sleep/unconscious state
2. Copy Generation Phase
- Unconscious processes create multiple copies of sleeping conscious mind
3. Hidden Observation Phase
- Unconscious system observes lottery results
- Conscious awareness remains suspended
4. Copy Manipulation Phase
Unconscious system either:
- Eliminates copies in losing branches, or
- Multiplies copies in winning branches
5. Consciousness Resume Phase
- All conscious minds reactivate in identical states
- Self-Sampling Assumption applies
- Minds distribute proportionally to copy numbers per branch
(Requires state-based rather than path-based identity)
6. Awakening Phase
- All conscious minds receive bet results
Delayed observation for probability shift in MWI
It looks like we can create something similar to ULWM from Anthropic trilemma by delaying observation in the winning branches. In MWI the multiverse is constantly branches and the delay in observation in the winning branch will result in more copies as it presented in fig 1.
The delayed-observation-ULWM works according to following algorithm:
1. An observer tosses a coin and while it is in the air, enters in the state of sleep. (The number of observer branches doubles each step).
2. Coin fails and non-conscious AI measures the results.
3. If heads, the AI wakes up the observer in observer branch, but if tails, the observer remains in sleep for one more step. The observer, after awaking, only knows that he is awake, but does not know the result of the toss.
4. In the next moment of time, the head group of observers learns the result of the toss, but the tail-group only learns that it is awake.
5. Therefore, after awakening, but before learning the result, the observer is in the set with the same epistemic state which includes 2 observers in the tails group and 4 observers in the heads branches. And if the observer is thirder in Sleeping beauty, that is, he is using SSA, he should conclude that he has 2/3 chances to be Tails branches.
Note that after learning the truth, this advantage will be not (immediately) erased: the first observer-moment when he learns the Tails-truth is still delayed relative to the first observer-moments in heads-truth.
Fig1. If total measure of the universe is not changing, time delay doesn’t give the increase of measure of Awake 1 states, but increase the number of such copies.
Measure decline problem and Youngness solution
The immediate counterargument to Fig 1 is that each tails’ observer-moment with the state “Awake 1” has only 0.5 measure of the head “Awake 1” observer, and if we take this into account, there is no change of subjective probability.
One way to overcome this is to assume that the total measure of the universe is slowly growing (or declining).
Youngness paradox states that earlier observers are extremely more likely.
The youngness paradox is caused by the fact that the volume of false vacuum is growing exponentially with time with an extraordinary time constant, in the vicinity of 10E−37 s. Since the rate at which pocket universes form is proportional to the volume of false vacuum, this rate is increasing exponentially with the same time constant. That means that in each second the number of pocket universes that exist is multiplied by a factor of exp {1037}. At any given time, therefore, almost all of the pocket universes that exist are universes that formed very very recently, within the last several time constants. The population of pocket universes is therefore an incredibly youth-dominated society, in which the mature universes are vastly outnumbered by universes that have just barely begun to evolve.
In our situation this means that younger observers are more likely, and thus measure in the universe effectively declines with every step.
To exploit youngness paradox, evolution has to awake the conscious observer earlier the desired outcomes (not later as in fig 1). 10E37 increase per second gives only one doubling around every 10 milliseconds which is timescale comparable with possible delays of neurons activity between brain cycles.
Note that Youngness paradox itself predicts many strange things: I will be more likely to be intelligent trilobite which unexpectedly gained consciousness than naturally evolved human being. So, Youngness paradox can’t be the real solution as well as any other theory in which the total measure of universe exponentially grows or decline, as it will quickly compound to astronomical numbers and make our position catastrophically unlikely.
I have an intuition that there is a way to “accumulate measure” by withholding observation even without Youngness paradox. For example, it could be a way to calculate only the number of copies, but not their measure – but that would destroy Born’s rule, as there is there are many improbable branches.
For example, magic-by-forgetting can be used: If tails, the unconscious part of mind erases one bit of information from the conscious part, and as a result it merges from SSA-perspective with another multiverse branch, and thus doubles its measure.
3. Magic in the biological evolution
We do not observe mass suicide of living beings which do not get desired outcomes, so direct implementation of quantum suicide in biology doesn't exist. The only potential exception is the theory that the whole biosphere "miraculously" survived eons – the Gaia theory (Doolittle, Darwinizing Gaia, 2014), which parallels Almond's idea of civilization-level quantum suicide. However, ULWM may still exist in biological evolution through unobservable changes in the measure of different MWI branches.
Evolution: probability increase through copies multiplication
The mechanism of increasing probability through copy creation and selective deletion mirrors evolutionary processes.
Evolution preserves and amplifies any trait that enhances survival probability and reproductive success. If quantum probability manipulation via copy generation and selection is possible, evolutionary processes would inevitably exploit this mechanism. Even minimal advantages per generation would compound over time through successive iterations.
The parallel between ULWM and evolution runs deeper than surface similarity:
- ULWM creates copies and lower the share of unfavorable outcomes
- Evolution generates offspring and eliminates less-fit individuals
- Both systems increase desired outcome probabilities through creating more copies
- Both accumulate advantageous outcomes over multiple iterations
Consequently, if quantum probability manipulation mechanisms exist, evolution would likely have incorporated them into biological systems, regardless of how subtle the initial advantage might be.
The importance of world-share increase
A potential objection to evolutionary ULWM might be that it affects only subjective probabilities while evolution operates objectively. However, this overlooks two key points:
First, quantum lottery winning mechanisms have already operated at the planetary scale through the anthropic principle, helping Earth survive numerous extinction-level events (asteroids atmosphere composition changes, temperature changes). Evidence of this appears in the anticorrelation between Solar luminosity and CO2 content, among other atmospheric phenomena (Armstrong).
Second, I as observer, observes only those timelines where the system works and created observers. This is because biological ULWM accelerates evolution significantly and it can create conscious minds before Sun destroys Earth.
Only ULWM variants based on measure increase, rather than measure destruction, can maintain dominance across the multiverse. Worlds implementing measure-increasing ULWM should grow in relative proportion compared to those using conventional evolution alone.
How a biological being can manipulate measure
Measure manipulation in the brain can occur through differential timing of conscious observations. Information is typically processed through unconscious neural pathways before reaching awareness. Visual data, for instance, travels through the visual cortex to both conscious perception area and to the (presumably quicker) limbic system, allowing measure manipulation via precise timing of conscious observation: subconscious system can quickly preprocess data and if it finds bad, it can accelerate its conscious perception.
It is rather common experience: I can be startled if I see an object flying in my direction.
In the human brain, sensory data typically processes unconsciously for approximately 0.1 seconds before reaching conscious awareness, with transfer speed potentially varying based on content type. A relaxed mental state, characterized by lower-frequency brain waves, performs environmental measurements less frequently.
Threatening stimuli propagate more rapidly through sensory channels – for example, sudden dangerous sound elicit faster reactions, resulting in a shorter period of perceptual inactivity.
The whole setup is very similar to time-delay ULWM in MWI. While I can’t definitely prove that this actually works or works in described ways, I show that all needed instruments to build time-delay-ULWM are in place in human brain.
Consciousness appears fundamental to this process, as benefiting from quantum-mechanical "lottery winning" requires a first-person subjective viewpoint fundamentally different from the third-person perspective. Qualia may play an even more important role, as qualia are definite states that cannot exist in superposition. This is discussed further in Chalmers' "Consciousness and the Collapse of the Wave Function". Such mechanisms could potentially enable certain quantum lottery implementations, though this requires additional research.
Adrian Kent also suggested that consciousness had some evolutionary advantage.
What kind of past evolutionary evidence can be?
Biological ULWM which helps to survive in the situations of danger or in search of the best mate will have the biggest impact on fitness. The observational consequences should be most prominent in my ancestors’ survival and the speed of their evolution.
By definition, it is impossible to observe the changes of the subjective probabilities of others.
But if ULWM works, it should worked in my past (here I can list several events where I survived some risks).
It also work for my parents: obviously, my parents must have survived until my conception, just based on conditional probability. If my father played Russian roulette in his young ages, I can observe only worlds where he won. Similarly, I “inherited” any ULWM results he used, and this is true for all my ancestors (here goes some stories how my parents and grandies survived WW2 and miraculously met each other.)
One more general prediction is that the evolution of species which are needed for human existence should have been quicker. For example, insects stagnated for hundred millions years (dragon fly is 250 million years old). However, flowering plants seems to be necessary condition for our existence as they can provide a lot of food and they appeared relatively recently.
Future technology could exploit this capability but there are risks
Future technology could potentially exploit and amplify these probabilistic effects by orders of magnitude, creating both new capabilities and risks. Human civilization has consistently demonstrated the ability to scale minor phenomena into powerful applications – from fireworks to space rockets, from trace radioactive decay to nuclear weapons.
If subjective probability manipulation proves viable, it could be developed into systematic technology. While the subjective nature of probabilities here presents challenges, these might be overcome through the "delegation" mechanisms or civilization-scale Universal Lottery Winning Machines (ULWMs). It could work as "civilization-level quantum observation" as a universal problem-solving system similar to Almond’s civilization-level quantum suicide.
For civilization-level ULWMs, one potential approach involves coordinated observation: multiple scientists simultaneously receive binary experimental results (preventing inference from others' reactions), with neural observation processes in the brain of each scientist is optimized to maximize probability-shifting effects. This creates a collective subjective observer. The observer group could theoretically scale to global population levels for significant world-state modifications.
Like all transformative technologies, probability manipulation carries catastrophic risks when pushed to extremes (comparable to nuclear warfare or unaligned AI). Probabilistic weapons could potentially increase observed frequencies of natural disasters like asteroid impacts or supervolcanic eruptions. Even Stanisław Lem's conceptual "probabilistic dragons" become theoretically possible (Lem's story partly inspired this analysis). Additional unknown risks may emerge, such as "disordered collapse of branches" or other currently indescribable quantum-level effects from excessive probability manipulation.
Relation to AI
Measure manipulation presents computational implications. In effect, Universal Lottery Winning Machines could theoretically resolve P=NP by generating all possible solutions and selecting only those timelines where the correct solution manifests. This mechanism effectively functions as an Oracle AI system.
Experimental testing
In general, the theory can be tested if a conscious observer will record a series of observations of binary quantum coin, but each time it is heads, the observation’s result will be delayed for a few milliseconds before it will be given to the observer. The observer thus more likely to observe later Tails, if the theory is true. Can be done as a computer program.
Experimental results: Sonnet generated a program which takes a random value from quantum public generator and demonstrate me Heads of Tails, with Tails were delayed for 10 milliseconds.
I tested it only twice with 10 days between tests as I have intuition that probability manipulating capability can be somehow be exhausted. Initially, in the first attempt Tails were more probable – 60 per cent up to first 40-50 tests, but after 66 tests the results become 50-50 (regression to the mean). I didn’t screenshot the first experiment. You can run this program here: https://poe.com/preview/8s4sTcXTHIVs8EPDfbRr
Results of the second attempt is 21 head for 50 trails, which has p= 0.1611 that head were 21 or less. Tails were always leading during test. I screenshotted the results. On subjective level, the test feels exhaustive somehow.
The p = 0.16 result is insignificant, but also the theory predicts only small probability shift. It can’t be countered by repeating the experiment as repeating will cause exhaustion. Though of we combined the two experiments, in both of which Tails had small lead, we get p-value 0.11 for hypothesis testing according to Grok (can’t just multiply them, need to use Fisher's combined test).
However, it is only subjective result, so each observer has to perform the experiment on herself. The publications by other observers about such experiment will report negative result.
Some additional even more speculative considerations (inspired by conversations with J. Miller and includes some direct quotes from him)
Generally speaking, if you have quantum immortality or various types of anthropic reasoning, probabilities have weirdness that you cannot eliminate, so it is reasonable to think if QI true, evolution has found a way to exploit.
This approach also helps explain why if we are in a papercliper simulation it would bother to make us conscious because that is a more critical part of everything you couldn't easily simulate.
More speculatively, there can be a form of natural mass suicide of beings who do get desired outcomes. One is spontaneous abortions "If I don't get the right gene mix, I abort myself."
Another is “normal” human suicides in the case bad life situations. By precommiting to perform suicide if I, say, don’t have luck in life, I somehow trigger observation selection in the direction of more luck. Note I do not endorse such precommitments as they eventually will not work in MWI with quantum immortality for me personally. But they can be easily naturally selected as only lucky ones will survive; here however we suggest that such selection is empowered by small probabilistic shifts via biological ULWM. This explains suicide ideation during depression.
In Hansonian Ems world, probability manipulation similar to Anthropic trilemma – multiplication of digital copies who get the desired outcome – can become a norm.