Mind Matter Interaction

Written by d0ublezer0, inventor of the fatum project and the creator of the noveltism philosophy behind Randonautica

Attractor/Void points are the most popular function in Randonautica, but not everyone understands what exactly is behind them, and why their calculation is so resource-hungry. Many even express the opinion that these are just random points, which, of course, is fundamentally wrong.

This article will probably be the most difficult in the series. In it, we will try to figure out whether thought can influence matter and lift the veil of secrecy over technological problems and approaches that have not previously been publicly discussed in the randonaut community. Let's start with what MMI is.

"Mind-Matter Interaction" (MMI) is the idea that focused mental intent can influence physical events at a distance. MMI also refers to the study and development of theories and technologies based on the influence of the mind on material events.

In 1979, scientists from the Princeton PEAR laboratory (Princeton Engineering Anomalies Research) took up the study of this phenomenon. They used the Random Event Generator (REG) for their experiments. This is such a device that, using the processes of quantum tunneling, creates a random sequence of zeros and ones, that is like tossing a coin, only 200 times per second and at the quantum level.

The fact is that the Theory of Probability tells us that if you toss a coin 200 times, then each side will land upwards about 100 times, that is, 100 (heads/tails) is the average expected value. But, since the process is random, one of the sides can land upwards, for example, 101 times (deviation = 1), or 96 times (deviation = -4). If you repeat a series of trials again and again, the arithmetic mean of all deviations will tend to zero, and the deviation values themselves will have a normal distribution. When the device worked in an empty room, it was so, but when a person was watching the device, initially expecting to see more ones than zeros, then there actually were more deviations in favor of ones. You can read more about the research here: Correlations of Random Binary Sequences with Pre-Stated Operator Intention: A Review of a 12-Year Program

It also became known from the experiment that the effect does not depend on how far the person is from the generator and even on how much time has passed between measurement and observation. It is enough for this to work, that the observer has expectations and can see the result. Unfortunately, the effect itself was so small that even the data collected over 26 years was not enough to convince the academic community of its existence, which is why the study itself is mercilessly criticized.

However, the phenomenon turned out to be of great interest to the Fatum project, whose goal was to try to escape the patterns of fate using random number generators. The project was not an academic study, and could afford to create technologies based on phenomena not accepted by mainstream science. The main idea of the Fatum project was the existence of the "Stasis Field" - the totality of all methodological templates that keep us in a routine and a certain perception of reality. Visiting random coordinates made it possible to bypass those patterns and get to places where no decision chains lead, but this process was purely probabilistic and it was impossible to control it. The MMI phenomenon opened up the possibility of purposefully finding anomalous places, as if creating them with the power of intention. Thats why, in the terminology of the Fatum project, the MMI phenomenon was called the "Genesis Field".

The idea is: In the PEAR experiments, the user did not know which physical processes he was affecting, he simply expected the graph on the screen to go up and watched as it happened. And this means that as long as there is expectation and observation, intermediate abstractions can be anything. For example, instead of expecting the deviation graph to go up, the user might expect to get to an unusual place where he will find something new and interesting. It remains only to figure out how to entangle this expectation and observation through a system based on quantum random processes. As a result of solving this problem, the system that now underlies the Randonautica app was born.

However, unlike the PEAR experiment, which only allowed to determine the presence or absence of an effect, here it is required not only to isolate the signal created by the user's intention, but also to extract information from it about geographic location, which adds a couple of new dimensions to the problem. To solve it, a scheme was invented that includes the following steps: 1) The user's intention affects the quantum noise generator. 2) The noise is processed to amplify the signal created by the intent. 3) Quantum noise is converted into a coordinate field. 4) In the coordinate field, statistical distortions created by intention are searched, which give a point on the map. 4) Many users visit points and observe the result. 5) Results data is collected and analyzed to understand if the system works and what patterns it shows. All these stages, of course, have many nuances, which we will now try to consider in turn.

To create an MMI system, a Random Event Generator (REG) is required, which will be influenced by the user's intent. Such a generator produces a random sequence of zeros and ones. However, the current understanding of the issue suggests that the user's intention affects not so much the random events themselves as their probability, and only a small part of the bits leaving the REG will carry a trace of this influence. In the PEAR experiments, the signal was very weak and out of 200 "coin tosses", the result was "non-random" only in a couple. This brings us to the concept of generator’s responsiveness, that is, how easily the user's intention can change the value of the output bits. It is obvious that software pseudo-randomizers do not have such responsiveness, since the randomness in them is purely mathematical, which means it is predetermined, and so it is impossible to influence it from the outside in any way. The only changeable parameter for such a generator is the time when the number is received from the sequence. Every true random number generator uses a physical source of entropy to provide its inherent unpredictability.

Such entropy sources can be divided into the following types:

- Thermal noise, also called Johnson, Nyquist or Johnson-Nyquist noise, is electronic noise created by the thermal excitation of charge carriers - usually electrons - in an electrical conductor.

- Shot noise, sometimes called Poisson noise due to its statistical distribution, occurs because the electrical current is made up of a very large number of discrete charges, making its flow grainy.

- Avalanche noise occurs in semiconductor junctions such as Zener diodes when the breakdown voltage exceeds approximately 7 volts. The strong electric field at the junction accelerates the charges until collisions set off a chain reaction that generates additional charge carriers.

- Jitter "noise" is not a type of noise, but tiny deviations in the rise and fall times of waveforms.

- Pure quantum systems, such as a photon passing through a beam splitter, are inherently random as described by quantum mechanics. A single photon detector in each of the two possible outputs from the beam splitter detects the photon. Each detector output is assigned to represent a 1 or a 0, which is the random output of the detection.

- Nuclear decay noise, obtained by measurements of the decay of radioactive material. The time of nuclear decay is determined by quantum mechanics and is considered inherently random within its expected statistical distribution.

MMI researcher Scott Wilber claims that thermal noise and quantum systems show the best results, shot noise in Zener diodes is less responsive, and REG based on avalanche noise in Zener diodes has the lowest responsiveness. By the way, the PEAR experiments used the shot noise of Zener diodes, so their technique was not the most efficient. In the Fatum project, it was decided to use a pure quantum system, which was a random number generator from the Australian National University, which measures the quantum fluctuations of virtual particles in a vacuum. The Fatum server receives entropy from it through an open API.

However, such a system also has flaws. Since there is only one source of entropy, and there are many users, their intentions can interfere with each other if they affect the REG at the same time. Not to mention that if the source of entropy breaks down, the whole system will stop working, which happened in January 2019 during the fires in Australia. After that the developers of Randonautica began to consider alternative entropy sources that could be built directly into the user's smartphone.

As a result, two new experimental REGs were temporarily added to the application:CamRNG, which used thermal and shot noise from individual sensors in the smartphone's camera matrix, resulting from the fact that photons are emitted by a light source at random times, and Temporal, which gets physical entropy from the processor cycle counter. However, in addition to the insufficiently high rate of entropy production, a serious problem with these REGs is the dependence on the phone model and its operating system. To partially solve this problem, a server version of CamRNG was created.

Also worth mentioning is a line of experimental REG devices designed by Scott Wilber. These devices are called the Mind Enabled Devices (MED) and are often used in Fatum Project's internal research. The uniqueness of such devices lies in the fact that they are specially designed to work with MMI and are a chip with many ring oscillators that produce jitter from thermal and quantum shot noise. Further, the noise from different generators is combined by the XOR logical operation, which allows eliminating statistical defects without deterministic post-processing, which is important, since such post-processing is often used in hardware RNGs and reduces the MMI effect. In addition, the MED has a built-in bias amplification algorithm to amplify the signal, which we will discuss next. It is believed that such devices are the most advanced REGs and can be accessed by anyone involved in the development of Randonautica research initiatives.

You can read about Scott Wilber's research here and here.

As we said earlier, the MMI effect is very small, and out of hundreds or even thousands of random bits, only a few become flipped by intention. Bias amplification algorithms have been devised to enhance this effect and make the signal more intense.

Bias amplification is an algorithmic method of compressing a large sequence of binary bits into a shorter one without loss of information. That is, the sequence becomes shorter, and the bits changed by intent in it are denser. Thus, when intention affects quantum noise, it is distorted more, and when there is no influence, the noise remains statistically unchanged.

As a rule, such algorithms are created based on complex mathematical calculations, the task of which is to estimate what is the probability of the prevalence of bits that match the user's intention in the input and output sequence. The ratio of these probabilities is called the amplification factor. Currently, there are several methods of bias amplification:

“Random Walk” is the most theoretically efficient method. It consists in the fact that one and zero in a random sequence are replaced by +1 and -1, and then summed up sequentially. The result of the addition "wanders", and the longer the series, the further its value can go from zero. So the values are summed up until the result reaches a certain value, for example 142 or -142, and if the result is greater than zero, the output is 1, and if it is less, then output is 0. The output threshold can be adjusted on the fly depending on the statistics of the results. If each result needs to be delivered in a fixed time, then the threshold is not set, but simply a fixed number of bits are input and we look in which direction the result deviated the most during the walk.

“The Majority Voting” method is less efficient, but quite simple. For it, a series of bits of a fixed length is taken, and the bit with the bigger presence in this series goes to the output. It is worth noting that the number of bits in the series must be odd to prevent ties. Calculations also show that such algorithms cannot be applied sequentially one after another, otherwise their efficiency drops.

“Bayes Updating” - The method is based on Bayesian statistics, where the probability that a bit matches the user's expectations is calculated based on that probability for the previous bits. Due to its high complexity, we will not describe it in detail in this article.

In practice, the effectiveness of such algorithms is tested as follows: The user presses the button many times, and receives one output bit. In this case, the user focuses on the intention to get a One. Then, practical efficiency is calculated according to what percentage of such clicks was successful. However, it is noticed that the longer the user trains, the higher the result he shows. Even one week of daily practice can significantly increase the frequency of hits. The effectiveness of such training is also enhanced if the user can see how the results change in real time.

However, a series of such attempts should not be made too long, otherwise, due to the laws of statistics, the percentage of hits will inevitably tend to 0.5, since the intention does not work constantly. We can draw an analogy with lifting a kettlebell, the weight of the kettlebell here is the deviation value, it will be possible to raise it sequentially only a small number of times, after which rest is needed. In addition, there is an observation that after the end of the intention influence, a long series of hits can be abruptly followed by the same long series of misses, sharply leveling the statistics. This phenomenon is called the Rubberband effect. The nature of this effect is not yet clear, as well as whether it exists at all, or is just an illusion.

Since in experiments it is often important that the result comes with the same delay after each button press, preference is given to bias amplification algorithms that take a fixed number of bits as input. One of the problems of such algorithms is the need to make sure that the sequence of identical bits created by the user's intention gets into the input set of bits in its entirety and with a minimum of extraneous noise.

One way to solve this problem is the channel method. The entire series is shifted by 1 bit and measured how ordered it is. Sequences with many repetitions like 1111111 or 000000 or 101010101 are more ordered than random ones. The bit-shift is repeated until a maximum ordering is found, and thus it is determined that the bias is captured in the bit sequence in an optimal way.

(You can read about how bit order is measured here)

Further, in order to get a point on the map from quantum noise, we must first transform it into a coordinate field. Based on the perturbations of this field, we will look for the place indicated by the MMI effect. A coordinate field is a set of points evenly distributed throughout the map. In reality, these points are scattered on a simple plane surface, and a ready-made result of calculations is projected onto the geographic map (i.e., the system contains no actual information about the real world location, it is working only on a coordinate plane). Moreover, in order for the calculations to be statistically reliable, the field must be sufficiently dense, that is, several thousand points must be scattered over the map, which requires a lot of entropy, and, consequently, fast REG. But how exactly do we get coordinates from a random series of ones and zeros?

There are two methods for converting random bits to coordinates:

Binary Word (BW) - this method is currently used on the Randonautica server, as it is the fastest and does not require too much entropy. The method is that 32 bits are simply taken from the entropy stream and translated from the binary code directly into a decimal number. From two such numbers, you can get the coordinate (x, y). Thus, for 10,000 coordinates, approximately 80Kb of entropy is required.

We will not cover all the mathematical problems that arise with this number generation method, such as Modulo Bias, but we will consider a number of problems related directly to MMI.

The first of these is the bit significance problem. In a series of bits that form a number, the first bit will always affect the position of the coordinate more than the last. For example, let's take the number 214 (11010110) and change the first bit (01010110): we get 86. But if we change the last bit in it, (11010111), we get 215. That is, the first bit changed the number by almost 2.5 times, and the last one only by a single unit. Given that the intention can change any bit in the series, such an influence of its position on the result is highly undesirable, not to mention the fact that we will get a completely different set of coordinates if we start reading the entropy not from the first bit, but from the second, for example.

To get around this problem, a bit-rotation method was developed, where the entire bitstring from REG passes through a 32-bit window, shifting 1 bit to the side at a time. Thus, each shift gives a new number, and each bit goes through all positions, participating in the generation of 32 different numbers. This method in theory not only eliminates the problem, but also allows you to get more numbers from the same amount of entropy, however, the question of whether it is effective in practice and whether it reduces the MMI signal remains to be answered in the course of experiments.

The second important issue is the compatibility of the BW method with bias amplification algorithms. Basically, such algorithms compress a sequence of bits, storing information only about statistical deviations towards one or zero, but information about more complex patterns is lost. If for a one-dimensional up and down movement of the graph it is enough for us to make REG emit additional ones/zeros sometimes, then in order to create coordinates close to the desired place, we need to make REG produce a repeating series of bits that look like certain numbers. Moreover, the number 42 (101010), for example, will not contain any statistical deviation at all, which means that for a conventional amplifier it will be just noise. Such patterns require a completely different approach, related to the detection of repeatability in series of bits, or the amplification of the accumulated deviation of already generated numbers, and not the bit sequences that form them.

The problem is that decimal numbers cannot be effectively applied to algorithms like the Majority Vote Method, since it would require a huge number of inputs, several times the maximum number used. A compromise can be an algorithm of 32 classical random walks, the input of which is the corresponding bit from each position of the bit sequence of each number. Then the output bits from all walks will give the bit-sequence of the output number. However, a more suitable bias amplifier for BW can be created based on machine learning.

But in order to use neural networks to detect patterns, we need to create a certain standard of MMI-influence, that is, a source of intention that could influence REG constantly and without interruption. A person is not able to focus attention for a long time without getting tired or distracted, so the developers of the Fatum project placed their hopes for creating such a standard on plants. Some experiments show that plants are capable of MMI. To do this, installations are created where REG controls the lamps that illuminate the plants, and they have to influence it in order to get more light.

Random Walker (RW) is the main alternative to the Binary Word method. In this method, we initially create the required number of points with coordinates (0,0), and then, depending on which bit we get from REG, we shift the coordinates of each axis by one up or down / left or right. Thus, the coordinates "wander" and scatter from the middle of the plane in random directions. However, when they disperse, their distance from the center will have a normal distribution, that is, there will be more points in the middle, and fewer at the edges. To turn them into a uniform coordinate field, the coordinates of the points are converted to z-scores by dividing their coordinates by the square root of the number of steps, and then converted to uniform coordinates using a complex mathematical polynome.

The advantage of this method is that it does not have the problem of bit significance, it is maximally compatible with bias amplification algorithms, and by itself it performs such an amplification right in the process of creating the coordinate field. But the method also has disadvantages. The most important of these is that the accuracy of the z-scores into which normally distributed points are converted depends on the number of steps they take while random-walking. This accuracy, in turn, affects the resolution of the points after being converted to a uniform distribution. That is, if the points have taken a few steps, then the coordinate field will look like a square grid with large gaps between points. And to get a coordinate field with the right resolution, we need to take hundreds of thousands or even millions of steps for each point, which requires an insane amount of entropy and very long calculations.

You can, of course, initially generate points with a normal distribution for a given number of steps, and then add several steps from REG to the points with each user request, but even in this case, you will need to calculate exactly how many steps are needed so that the points on the map completely change their position. And, most likely, this number will also be quite large and will grow with each step.

After we have converted quantum noise into a coordinate field, we need to understand where this field is most distorted. This place will show us the final location on the map. In other words, if our intention distorts the random bits in such a way as to show us the right place on the map, then the points of the coordinate field will shift in the direction of this place.

Of course, since the MMI effect is very small, the points will not move much. The points will still be randomly scattered around the plane, but some of them will be a little closer to the right place than if they were scattered evenly. Since the initial position of each point is random, we cannot tell if a certain point is shifted in any direction or not, but when there are many points, such a shift will affect their density. Moreover, if several points are moved in the direction of one specific place, the density of points around it will increase.

Such deviations of the density of coordinate field in Randonautica are called Intention Driven Anomalies (IDA), and they can be of two types: Attractor Points and Void Points. Attractor points are areas where the density of the coordinate field is higher than average, and void points, respectively, are areas where the density is lower. That is, in the case of an attractor, the points are shifted towards the center of this attractor, and in the case of void, the points are shifted in the opposite direction, and kind of run away from this place. Thus, the names of these anomalies have nothing to do with what you find in these places, they only show in which direction the coordinate field was distorted when they were calculated.

Since the points are randomly scattered on the plane, their density will also be uneven, and small areas of attractors and voids will appear randomly throughout the map and without any MMI influence. Therefore, in order to determine which of them did not arise by chance, the size of the deviation from the average density is calculated for the anomalies, and if it is large, then most likely we are dealing with MMI. Such a value corresponds, for example, to the Power of an anomaly, which is calculated as the ratio of the density of points inside the anomaly to the density of points on the entire map. So if the Power of an anomaly is 3, it means that the density of points around this place is three times bigger than the average on the map. A more precise value is the z-score, which is used to measure the dispersion of values around the mean in mathematical statistics. For attractors, the z-score will always be positive, and for voids it will always be negative, but the farther it is from zero, the more likely we are dealing with MMI. If the z-score is greater than 5, there is a high chance that you have a valid intention driven anomaly.

It is not known whether there is a difference between how our intention affects the outcome of visiting attractor points and void points. Some users believe that void points are more often deserted, abandoned places, but it can be assumed that this is an illusion created by their name. From the point of view of mathematics, attractors and voids are equally anomalous and differ only in the direction of deviation of the points. It is also not entirely correct to assume that high power anomalies will contain stranger things. In fact, the power does not characterize the strangeness of what is in these coordinates, it only shows how likely this point was created by your intention, and not by pure chance.

But back to the algorithms. The task of searching for perturbations in the coordinate field itself also has many complexities and nuances. The first prototypes of the Fatum Bot used a fairly simple algorithm, named after the developer's nickname "Zhaba". This algorithm calculates the arithmetic average coordinate of all points inside a certain circular zone, then makes this average coordinate the center of the zone and reduces its radius by one. The process is then repeated over and over until the zone radius reaches a minimum. Thus, the center of the last zone will coincide with the center of the densest cluster of points on the map. With a slight modification of this method, one can also search for voids.

However, despite its simplicity, this method has many disadvantages: It can only find one attractor on the map, it cannot determine its size, moreover, if there is a void next to the attractor, this can make it invisible to the algorithm. And since the research needed to find all the anomalies with high power on the map, and not just one, this algorithm was soon abandoned.

In mathematics, to solve such problems, there is a more efficient method called Multivariate kernel density estimation. We will not delve into its essence, so as not to overload the reader with mathematics, of which there is already a lot. The developers of the Fatum-bot were faced with the task of creating an algorithm that would not only find all anomalies on the map and accurately determine their power, size, and z-score, but would also do it as quickly as possible. Many algorithms cope very slowly with such a task with a large number of points, and in our case there are tens of thousands of them.

As a result, the “libattract” library, also known as “NewtonLib”, was developed, which solved the problem using the clustering algorithm.

This library became the core of the Fatum-3 bot and is still used as a component of the Randonautica server. And although the library copes with the calculations quite quickly, the consumption of processor power remains quite high, which does not allow it to be built into an application on a smartphone. Thus, all calculations are performed on the server side and require significant resources if many users request points at the same time. It is for this reason that the number of points that a user can request per day for free has been limited in order to avoid overloads.

The Randonautica developer community conducted many experiments in order to improve the efficiency of algorithms and technologies. Usually, the effectiveness of the algorithm is determined as follows: a target is placed in a pseudo-random way on a certain plane. The operator looks at it and tries to focus his attention on it. Then a coordinate field is applied to the same plane and attractors are calculated. The process is repeated many times to measure how often the attractors hit the target. However, the problem is that each coordinate field is created at one particular moment and we do not yet know how the same attractors behave in continuous time.

The fact is that the user's intention affects the REG for some time, it means that the anomaly must persist throughout the exposure. And if we take into account the lifetime of the anomaly, it would be possible to determine its MMI origin with much greater accuracy than by the magnitude of its power. After all, high power deviations can occur by chance, but only anomalies created by the MMI can persist in time. So, if we learn how to determine the lifetime of attractors and voids, then the user's intentions in them will manifest with a much greater probability.

But on the way of such calculations, of course, there are many problems: The first of them is that the calculations will become more difficult, since the time axis will also be added to the plane on which the points are plotted, which means that you will have to look for attractors and voids in three-dimensional space. The second problem is that a huge number of points will be needed to fill the volume of that space, which means a very fast REG. And since we do not know what the minimum lifetime of the anomaly can be, we will have to generate points with a high temporal resolution, at least 10 layers per second. Such a choice is dictated by the average human reaction rate to a visual stimulus, 200 ms, and as already mentioned, feedback is key to the effectiveness of the MMI. At the moment, there are no entropy sources with such performance in Randonautica.

If these problems are resolved in the future, it will still be problematic to perform such calculations for each user, since the system will not be able to process two requests at the same time period. For this reason, it is planned to create a separate server called "Beholder" that will constantly scan the entire planet and, if anomalous deviations are detected, plot them on a public map so that randonauts living nearby can check what is there. Either a collective intention or people with a very high skill of MMI influences will be able to influence this server.

However, even now there are ways to explore the life dynamics of anomalies without having super-powerful REGs and 3D point arrays. One of these methods was the scanattractor function, introduced in later versions of Fatumbot and early versions of the Randonautica app. The principle of this function is that the entropy for generating points was divided into several parts and each part came from the entropy source with a difference of several seconds or minutes. Since the entropy is then converted into points, taking it in parts at different times is like adding together several layers of points taken at different times. If attractors in these layers arose by chance, then they will not exist on other layers, and their density will be blurred. But if there is an attractor on several layers, its density is summed up and becomes stronger.

In the laboratory, to study such dynamics, the plane is reduced to a one-dimensional line. Thus, each new line on the plane is filled with dots once a second and you can see how the density of dots changes in certain places over time. Such experiments have shown that the attractors found on the plane obtained from the sum of such lines are, on average, closer to the target than the attractors calculated for each row separately.

When an anomalous place is calculated, its center and radius are sent to the user. But in order for the causal loop of expectation and observation to close, the user must visit this place and observe what it contains. In theory, a trip to the place itself should lead to the fact that the user will observe the expected result. This means that the result does not even have to be at the point itself, just its location will be exactly such that on the way there, the user will meet what he expected. And from here it follows that if the user is not going to visit the point or expects failure, then this will directly affect the result.

Do not forget that the MMI effect is rather weak and works with some probability. For anomalies with high power, the probability to meet expectation is higher than for anomalies with low power, but this probability is never equal to 100%, and therefore you should not be disappointed if you did not find anything when you came to the attractor point. It should be understood that the probability of success of the experiment grows proportional to the number of attempts. That is, if a million people visit anomalies, or you visit a million places, the probability that the algorithm will work for some of them at least once increases a million times.

For example, let's imagine that you are looking for a UFO. In a normal situation, the probability of finding it is close to zero. Let's say you used MMI to find it, and that increased the success rate to 0.01%. With such a probability, you may never find a UFO even in 100 attempts. However, if a million people search for a UFO using the same MMI, then up to 100 people have a big chance of finding it. In this way, a large community increases the probability of finding something interesting, and in order to make these finds available to everyone, reports about them are collected into a public database, such as the Discover Feed in the Randonautica app. That is, even if you do not find anything yourself, with your journey you are contributing to the general search, the results of which you can get from the reports of other randonauts. Usually the most interesting reports can be selected by rating, so the community not only collects information about anomalies for you, but also sorts it. And as distance filters become available, you'll be able to track what randonauts have found nearby your location and join the exploration of their finds.

Also, some randonauts have noticed that the probability of success increases if you travel in "chains". To do this, after visiting the first point, you need to generate a new point with the same intention while being in the first point. Then, from the second point, generate the third one, and so on. Many reports indicate that such a search brings results already at the third point, even if the user was skeptical. However, there is no reliable mathematical model describing the cause of this phenomenon yet.

And although Randonautica receives a large number of reports from users, it has not yet been possible to extract scientifically reliable statistical information from them about how effective MMI technology is and how to improve it. Therefore, we are looking for big data experts who would like to help us. If you have experience in solving similar problems and want to join the community in solving them, feel free to offer your help.