Question about energy conservation in MMI noise signals

Hello to everyone!

This is going to be a technical question, and is directed to anybody who has done experiments and has analysed the affect of intent on noise signals and the power spectra of those noise signals.

First, a quick background on myself. I have a PhD in Physics, and my PhD project was centred on modelling and reducing the thermal noise generated by a kind of superconducting detector called a Transition Edge Sensor (these are used in sensitive space imaging telescopes, but also terrestrially in quantum computers, radiation detectors and other things).

I say this so that you can feel free to take some liberties when answering a technical question for my benefit: feel free to assume I have a moderate understanding of Applied Physics, Applied Mathematics, Thermodynamics and Statistics. If I don’t understand something (and I’m sure that will also likely happen), I will ask for further clarification.

But please do not feel overburdened by the idea that you might have to give extensive technical explanations of fundamentals if I am asking a complex question - I don’t want to make such a burden an obstacle to the person answering. If you don’t have the time to elaborate to a wider audience, then please do give the concise answer (for example, do not feel the need to answer the underlying question “what is thermal shot noise?” for my benefit alone - I already know).

My question (which is probably primarily directed to Scott Wilber, although anyone else is welcome to shed some light on this) is as follows:

Has anybody found that the expected noise power spectrum of a thermal signal (which is determined strictly by the temperature of the device and the speed, temperature and design of the measuring circuit) is affected in any way by the higher order statistical deviations caused by intention?

[For further background of my understanding, I am extremely new to all work regarding MMI - especially theory - so please do assume I know little about MMI and REG experiments.]

From looking through some of the threads on this forum, I did come across something Scott mentioned somewhere, in which he explained that a series of 1s in a binary random stream “borrowed” (I’m paraphrasing) from the 1s elsewhere, causing the 0s to increase in other runs.

To me, this would imply that intention cannot “add energy” in the average measurement sense; which would imply that energy conservation would be preserved in closed systems of thermal noise; even for systems highly sensitive to intent.

Ultimately, my question is: Has anyone, anywhere, directly tried to measure this power spectrum both with and without the presence of intent? If literature exists about this I would love to read it.

Hi dj3I4I,

Glad to see your question. Certainly people have looked at the power spectrum of random noise signals (both analog and binary forms) to see if they are altered by intention. Unfortunately any direct alteration is typically too small to rise above the noise floor, since the noise floor is virtually the entire baseline noise “signal.” That is, the non-stationary alterations caused by mental intention would only appear by averaging a very large number of trial periods or trials. If the size of mental influence is on the order of 200ppm (a typical level in PEAR-type measurements), it would take as many as 25,000,000 trials to increase the signal to noise to around 1. Enough to see the effect, but not with much confidence.

The main issue is a power spectrum is an insensitive measure of the type of statistical deviation we would look for in MMI. Most MMI measurements look at a very specific type of deviation from the baseline level. The simplest and most typically used is measuring the number of 1s in a binary sequence sampled from the analog noise source. If the fraction of 1s corresponds to the intended bias or direction – either more or less than average – then the result of that trial (measurement) is a Hit (success).

In such a measurement, timing is critical in two ways: 1) the beginning, and hence the ending of a fixed-length trial, is selected by the subject; and 2) the precise timing of each of a number of bits used to calculate bias is usually determined by an oscillator free-running in the background. To accomplish statistically significant results takes energy, since the entropy is shifted during a series of trials. I have derived equations to calculate the amount of energy depending on the design of the entropy source, but I have not provided that information yet.

In any case, either there is some balancing or compensating production of bits to make the overall average of bits a stationary variable, or, during a long series of trials with a fixed intention of either increasing or decreasing the number of 1s measured, the baseline measurement of 1s will eventually show a statistically significant deviation, as if the generator had drifted. A more direct measurement is to take two measurements for each trial, one corresponding to the usual initiated trial, and another one of the same length, taken approximately midway between initiated trials. One would expect the midway periods to exhibit a bias opposite the intention of the actual trials if there is a “balancing” effect. Or, if there is not a balancing, the midway periods should show little or no correlation to the subject’s intention.

Generally, the maximum effect size is achieved when the subject receives real-time feedback of success or failure in the intended outcome. Since only the actual trial period corresponds to that measurement, the midway periods show a much smaller effect, one way or the other. The implication is there is not a type of balancing effect as hypothesized. On the other hand, I and others have observed what I call a “rubber band” effect. That is, after an intense period of intending trials in one direction, when the concerted effort is lost or stopped, there seems to be a short period when the results snap back in the opposite direction in a significant or non-random way. This further implies the dynamics, or perhaps the human component of MMI, is more complex than any simple answer suggests. I believe this latter idea is more likely to be true, based on several other observations over the years.

This is by no means a thorough answer to your question, but at least an introduction.

Thank you for your answer, I very much appreciate you taking the time to help me get to grips with some of the fundamentals of the MMI phenomenon.

The following is a summary of my understanding from what you have said (if anything I say below is incorrect, please do correct me):

  1. The MMI effect is small, and thus the power spectra are only measurably effected over many samples (on the order of 25 million samples, one would expect to see just 5000 samples deviate from the statistical expectation).

  2. One can enhance the MMI effect by allowing the person with intent to choose the start and end time of their intention, and to measure only the fluctuations in the specified time window as the “target” samples. If the person can monitor the result in real-time, the effect is enhanced further.

  3. One should expect some external noise affecting the timing of noise signals, which derive from the accuracy of the reference oscillator used as the clock in the time/frequency analysis (presumably more of an issue for high-frequency noise oscillations?).

  4. It takes energy, by definition, to accomplish statistically significant results. The source of such energy is unknown, and if known would illuminate hypotheses regarding the energy conservation question.

  5. It is unclear whether that energy is compensated by the surrounding signals (in time). There may be some evidence to suggest the “intention energy” is not compensated (insignificant deviations from baseline in between windows of intention), but there is also some evidence to suggest this energy is compensated (the “rubber band” effect, unreliably seen by some).

If I’ve said nothing incorrect in the above, I have a couple of further questions regarding these points, which I will ask below, and will number with the relevant previous point:

(1). My understanding from reading around (mostly from your writings) is that thermal shot noise is the component which is affected by intention. Thermal shot noise follows a different frequency distribution than the classical noise component in a noise frequency spectrum. It is a mid-to-low frequency “hump”, centred with greatest influence on a very narrow range of oscillation frequencies that are determined by the temperature of the noise source. One would then presumably expect the “intention effect” to influence those same frequencies of noise oscillations. Has any processing ever been done in MMI thermal noise devices to filter out AC currents with frequencies outside the central shot noise band?

(2). Is there any understanding as to what the user is “doing” when they are monitoring the results and improving it?

(3). Has there been any experimentation with regards to the use of different (differing accuracy) reference clocks, and how this might change the strength of the intention signal? There exist some extremely accurate GPS clocks for example, that would achieve the best results for measuring high frequency signals across different devices. Have these ever been used?

(5). With regards to the possibility of a “bounce back” effect, my intuition (from what I understand to be the physical sources of thermal shot noise in circuits) is that a prolonged bias of a noise signal in some direction would cause anomalous heating in the substrate of the device (I am presuming that the bias is coming from the phonons in the substrate, and is therefore actually heating it). After the period of heating caused by intention, I would conjecture that the heated parts of the substrate would have to cool back down to equilibrium temperature - causing a “rubber band” effect in the measured current. If a thermal source of noise did not have a rubber band effect, then would this not imply that the temperature of the device was heated permanently to some higher-than-equilibrium temperature?

Further from (5), perhaps there are sources of quantum noise that do not derive from temperature-like resevoir-equlibrium effects; such that biasing the fluctuation of those signals would not lead to any thermal consequences in the measurement system. However, one would expect to be always bound somewhere by Landauer’s principle: although what exactly constitutes a “state” in such a case would be highly counter-intuitive (I’ve never thought about this until writing this sentence, and at first glance, this is very confusing).

However, if the energy debt is at the Landauer level then I presume it would be unmeasurably small compared to the deviation one desires. This might relate in some part to what the answer to the energy conservation question might be.

I have been quite busy lately, but I will get to your last question in a separate message.

  1. To be more precise, the 200ppm effect size is at the level of single bits. Historically, much data was taken in 200 bit blocks per trial. Therefore, the probability of having a majority of bits in the intended direction (excess 1s or 0s) in a trial is .501128 – about 1 in 443 trials. The higher figure I mentioned was meant for raw analog noise, but that is a measure that is rarely considered in MMI. I am not as certain about the exact statistical properties of MMI effects on analog noise.

  2. There are some ideas, but they are not proven. I hypothesize in a simple MMI trial the possible outcome exists in a superposition of a hit and a miss (correlated versus non-correlated outcome) until the user observes (becomes aware of) them both simultaneously. At that point the user’s intention shifts the probability of a hit occurring. The trial doesn’t actually exist until the intended outcome is compared with the “physical” measurement of the entropy source, which would also be considered in a superposition of a 1 and a 0.

  3. Noise in the timing signal can be considered a component of the nondeterminism in the primary entropy source, so it does not add or detract from the overall effect size of the MMI measurement.

  4. It seem likely that MMI effects arise from some type of quantum mechanical principle, though exactly how they arise is still an important question. It goes along with the question of, what is consciousness and what is its function with respect to creating physical reality? There are certainly potential sources of energy (zero-point energy – quantum vacuum) that affect real physical measurements. This is clearly evident when measuring pure quantum states, where the “random” outcome may be considered to be “caused” by the interference pattern between a wavefunction and a virtual wavefunction in the quantum vacuum. Anyway, our actual understanding of quantum mechanics is very far from complete.

  5. Yes. Though related questions could be subject to experimental exploration.

(1) Let me disambiguate the various entropy sources that typically exist in MMI generators and REGs. Thermal noise and shot noise are usually considered as deriving from two different physical mechanisms, though they are invariably both present in every real entropy source. Both these sources are normally distributed with respect to instantaneous amplitude. This is a consequence of the enormous number of charge carriers involved in conjunction with the Central Limit Theorem. Even though there is typically a lot of correlation in the electric fields in motion of charge carriers resulting in thermal noise, there is still such a large number of independent vibrations that deviations from a normal distribution are too small to measure. Both thermal and shot noise are considered to be white, that is, their power spectra are effectively flat from very low frequency to as high as we can effectively measure. Of course, quantum effects limit the ultimate high-frequency possible, but that is in the terahertz range. For practical purposes, the real range and flatness of power spectra are limited by the transfer function of measurement electronics. The electronics usually include analog amplifiers that can provide (with proper design) analog noise signals with flatness better than ±0.5db from about 0.1Hz up to about 1GHz.

(2) Partly answered, or at least commented on in 2. and 4. above. MMI skills can be improved by practice in ways similar to those used in biofeedback. That is, given real-time feedback, the user learns how to produce the desired (intended) result. There is always a question as to how the user is actually accomplishing that result. The answer is, we don’t actually know – is an ability or porperty of consciousness (which is not even acknowledged to exist by most neuroscientists in the USA). We can observe neuronal firings and see there is a change, but we cannot observe what causes those changes.

(3) See 3. above.

Thank you so much @ScottWilber for this thorough reply, it clarifies many things on my end!

There is just one point that I think we may be misunderstanding one another on, and that is what we mean when we each mention shot noise. I think I have been unclear in expressing what shot noise I am referring to.

There are two sources of shot noise in the measured noise currents. 1) The shot noise of the fluctuating number of electrons flowing through a cross-section of the circuit at any point in time, and 2) the shot noise of a fluctuating number of phonons (quantised packets of heat) transferred in the substrate (for example, in the silicon onto which the circuit is etched).

The 2nd (thermal) form of shot noise is coupled to the fluctuations in the flow of the current, and that coupling strength (and frequency-current response) is determined by the thermal conductance between the circuit and the substrate.

I personally find it interesting to consider that, in this case, there are at least two sources of shot noise affecting the measurement signal (one thermal and one electron based - there could also be a 3rd from incident radiation absorption).

If MMI effects are more dominant in shot noise, it seems to me that an electric circuit etched onto a substrate might possibly have two different underlying dynamics contributing to the overall noise outcome. Perhaps this is too academic a consideration to matter, but there could also be a chance to enhance the MMI effect by understanding this a little more.

In fact, it is fascinating to hear of the mostly flat noise response over such a large range of frequencies. In my mind I am still wondering if some of that flat noise response is actually more “quantum” than previously thought, when considering the possibility of a significant thermal coupling between the electric current and thermal shot noise in the substrate. Although I honestly don’t have any intuition for whether this would be significant or not.

Thanks again for your enlightening answers, and I look forward to your last answer!

You may already know, or that’s what you’re saying, that what I call thermal noise (Johnson-Nyquist noise) and phonon noise (thermal fluctuation noise) are fundamentally different. If we used a phonon noise detector as the entropy source for MMI, I would expect very similar results to current REG designs. From my modeling of charge-carrier shot noise or photon shot noise, the amount of energy required to “flip” a single measured bit from its baseline state to its opposite mentally-intended state is about the same as for Johnson-Nyquist thermal noise.

Note, in most MMI detectors or REGs, the primary noise component is usually thermal noise in a continuous or analog signal. Even when the circuit is designed to look at mostly shot noise, the amount of energy is nearly the same if the measured signal is continuous. This is only different when the entropy source is a pure quantum type, like single-photon detectors at the output ports of a polarization beam splitter observing single photons. Then the energy difference (between the baseline state and the “flipped” state) can be substantially lower.

Regardless of design, the energy difference is nearly always sub-Landauer. There is a strangeness in random number generators versus the standard Landauer limit, where a bit must be erased before it is changed to the opposite state. The measurement of a random bit does not require the erasure, so the dissipated heat is zero, or can even be negative, at least theoretically. What I don’t know, is there an analogous limit for changing a measured bit by mental intention while it is being measured? Even more complex, what is the energy difference, or is there a different theoretical lowest energy limit if the MMI effect is a type of quantum entanglement between a “hit” and a “miss.” Quantum theory – or at least cultural belief systems – is far from the point of explaining mentally induced entanglement or collapse of quantum states.

In any case, given the expected sub-Landauer energies involved, it is unlikely any simple experiment would see temperature-related effects in the substrate, even if one were looking specifically at phonons.

Hi @ScottWilber,

I just want to thank you for this answer, I believe it clears up every question I have. Apologies for the later response, I have been quite busy myself lately!

I’m looking forward to participating and being a more active member of the work done in this community very soon.