Qualia Productions Presents “Thinking Like a Musical Instrument” (and other communications from our Swedish QRI advisors)

By Anders Amelin and Maggie Wassinge (QRI advisors and volunteer coordinators; see letters I & IIletters III, IV, & V, and letters VI, VII, VIII)


Happy 2022!

[Here is] your new-year’s gift video titled “Natural Stupidity and the Fermi Paradox or The Tyranny of the Intentional Object”. It is full of things rarely touched upon by respectable scientists. Such as space aliens, and the gender of God. And valence structuralism.

Next, we’ll try to film a truly serious, comedy-less little demonstration of a metallic toy percussion instrument subjected to strain followed by annealing. In what way, if any, will the tone quality change? We have no indication yet what might come off of it so it’s a bit of a falsification attempt where we might get a null result with no discernible similarity between brain on psychedelics and metal on heat. Like most respectable scientists might expect. But, just possibly, there could be something interesting in store. 


In this video we illustrate the similarity between the brain and a musical instrument. The brain tissue is represented by metal and the brain activity by sound. The effect of substances such as psychedelics and dissociatives is mimicked by heating and cooling the metal. The engineering term for such heat treatment of metal is “annealing”. What we demonstrate is a very simplified toy model but which can be surprisingly useful for understanding the overall type of system dynamics going on in brains.

The model is based on the fact that both sound and neuronal firing are examples of oscillatory activity which can have different frequency, amplitude, coherence, and damping. Hammering the metal represents the memory imprint made in the brain by our ongoing experiences. The sound pattern produced by the hammered metal contains complexity which corresponds to learning. But a side effect of the increased complexity is lower overall consonance of the oscillatory activity.

To stay healthy, the brain must periodically undergo what the Qualia Research Institute calls “neural annealing”. In a neural network model, this can be thought of as redistributing synaptic weights more globally across the connectome and thus make the learned information more harmoniously integrated and holistically retrievable. This normally happens during sleep but can become even more powerful with meditation and psychedelics.

In this demonstration where metal is annealed, it is the positions of the metal atoms which adjust themselves so that the entire piece of metal becomes a better conductor of sound. It may seem strange that this can happen, but neither the metal nor the brain is fundamentally magical. Both cases involve self-organizing system dynamics.

In the case of the brain, the activity is accompanied by conscious experiences. The Qualia Research Institute works under the assumption that these are not magical either but can be modeled mathematically in a similar way to chemistry and physics. It is then necessary to test how measurements of brain activity correlate with conscious experiences not only during sober waking life but also under conditions which are very different.

The QRI is building a new paradigm for understanding the mind and the brain. With a focus on psychedelics and other mental state altering methods as scientific research tools and candidates for use in next-generation psychiatric- and pain treatments. We are a small upstart group with opportunities for volunteers and donors to get involved. If you are interested in learning more, please contact us via this e-mail address: hello [-a-t-] qualiaresearchinstitute.org

Relevant Links:

Scenes from the video (highly recommended):


[They further elaborated:]

Here is a video of a simple experiment we did on how straining and annealing a piece of metal affects its acoustic properties. In a QRI neural annealing interpretation it looks as if something interesting is going on. Simple things like sterling silver, hand hammering and heating over open flames were used and recording done only with iPhones and a Røde SmartLav+ microphone, but the results give qualitative hints that truly hypothesis-testing quantification experiments would be feasible to do pretty straightforwardly.

Especially of interest would be the hints that annealing produces frequency shifts and reverb changes which differ between the high and low frequency ranges, and the hint that annealing reduces dissonance. For quantification of shifts in resonant frequencies, one might manufacture a series of Chladni plates made of different alloys and which could be kept always flat but be cold rolled, heated and cooled to different temperatures with various ramp rates, and trimmed at the edges to alter their geometry. Then find the resonant frequencies for each parameter configuration. As a bonus you’d be able to visualize with a sprinkle of (beach!) sand on top. Then crunch the data to make predictions about brain activity signatures under for instance various psychedelics and meditation states.

Another one could be to quantify consonance, dissonance and noise levels for various metal resonators as these are subjected to various forms of stress, strain, and heating/cooling procedures.

It would all be simple enough to almost be like an intern research project but, excitingly, it is unlikely to have been done before. (OK, do a thorough literature search of course. As always…).

Suppose QRI were to explain parsimoniously with the neural annealing paradigm how brains pull off the amazing trick of producing plasticity which is “just right” in each modality of function. Artificial neural networks can be trained to impressive levels on complex data sets but they suffer from catastrophic plasticity in the sense that training on new datasets erases learning achieved on prior sets. This makes AI narrow and also very unsafe with respect to ease of hacking. The AI alignment community provides us no answer (at least not anything very parsimonious) to how absolute firmness in the modality of core “human values” can be combined with flexible (meta) learning for AI at a humanlike generality level. That is the notorious “alignment problem”.

Ultimately every system can be hacked of course and so can human minds. But certain humans are impressive moral role models, and meditation practices seem able to make most of us come at least a little closer to them. Suppose different brain networks loosely correspond to different alloys with correspondingly different ductility, hardness, tensile strength, different annealing temperatures and different hardening and tempering responses when undergoing various stressing, straining and heating/cooling cycling. The variability in this regard found in various metals and alloys is really immense. We’d want to eventually pick out the particular ones which happen to be the most useful for brain modeling.

Consider doing these quantification experiments in metal and then presenting to possible collaborators a brain model in the form of a formalized multi-alloy configuration. Don’t emphasize phenomenology if it’s AI engineers because that is a word which may give them bad vibes. Instead just present it as brain activity in different learning modalities. Which can be formalized and turned into software. With annealing and consonance-dissonance-noise as key elements. That is probably the kind of pitch you’d want to bring up if, hypothetically, a head of an AI research group were to ask about whether solving consciousness is necessary for producing more advanced AI. Since solving consciousness sounds unpalatably difficult, the answer they’d like the most is that it is not necessary. Hence they won’t care about collaborating with QRI if it is implied that the computational properties of phenomenological mind states must be reverse-engineered. One cannot blame them, it’s dizzyingly daunting to consider. But an information processing efficient metal acoustics-inspired brain activity model and with nice things like QRI valence formalism falling into place could be much easier to pitch. Just don’t call the CDNS-emergent utility measure anything resembling psychology terms like core affect… 😉


Letter IX: On Valence as a Currency Within the Nervous System

[commenting on the video about Zero Ontology:]

A reflection: It’s interesting the way Isaac Luria, during years of meditation, came up with an “inverse” view of the way the universe was created, by subtraction from a mass of infinite potentiality (which corresponds to maximum symmetry) rather than by addition of things to an emptiness by an unexplainably pre-existing divine creator agent. Lurianic mysticism has been a strong inspiration for pantheism and atheism, and even how to think about information. A nice example of how introspection can give new clues for how to better understand the universe. An isomorphism between fractal patterns in consciousness and fractal patterns in the multiverse generator perhaps. One could argue that Luria discovered symmetry breaking by meditating. Pretty cool!

While thinking about STV […] we tried to see if there are some more arguments for STV which make sense. So we went to the Less Wrong website. Now that’s a crowd that ignores qualia but they are big on economics and system dynamics.

Found this potentially useful: Evolution of Modularity by johnswentworth.

Life exploits the possibility space of “choreography and catalysis”. It organizes pre-existing physics & chemistry phenomena into a landscape of multiple optimization attractors. Expect modularity at many levels. Certainly also within brains. Economies accomplish the same thing as life does. They and life are in a fractally self-similar universe and economies are derived from the activities of life. To accomplish the exploration and exploitation of environments.

But then with brains we have the binding problem and the mind-body problem.

The QRI could argue that a brain can be mathematically modelled as a self-organizing hierarchical system of resonant cavities. There are certain similarities between that and how an economy can be modelled. Money is what solves the binding problem in an economy at the same time as it shapes the activity patterns in the economy. A common currency gives the highest efficiency. Profit/loss is the universal preference measure. It has a positive-negative axis, and it has gradients. Account balances and balance histories (assets, liabilities, contracts, derivatives… all using  money as measure) are the measure of aggregate utility. Preference (in the moment) and utility (longer-term aggregate) form a spectrum with feedback leading to amplification of certain states. Resonances emerge.

Keeping the same currency, fractally, across scales and projective transformations? (image by: Michale Aaron Coleman)

A brain uses long term memory, working memory, and nonconscious processing in a seamless blend. These seem like quite different components but it would be expected that evolution kept the same “currency” throughout as the modularity grew. Think of phenomenological valence as a “common currency” preference measure when it is instantiated in working memory (conscious awareness) and which can also “tag” the long term memory  “bank balances” with some marker for immediate reaction along an attraction/avoidance axis which is there full blown the instant the long-term memory is transferred into working memory. The search process by which this happens must aggregate that marker as the search happens, and adjust this continuously as the final result stabilizes into the resonant patterns of working memory. Static, distributed long-term memory, feedforward signaling, feedback amplification – all need a common currency for speed and even workability. It should also be considered much more parsimonious if proto-conditions exploitable by biology are clearly apparent in the physics of nonliving systems. All that happens in living organisms should be taken to be physics and not metaphysics, as a matter of Occam’s razor.

It is hard to come up with anything more elegant than a measure of symmetry as the common currency of preference used in brains. This can act as modifier or transform on information-bearing states in brains. It is flexible thus perfect for learning. You can transform back and forth in gradients of strength and still preserve the information the preference “is about”. Aggregation of the preference measure across the fractal activity spectrum and across time builds a utility measure out of the preference measure. Of course if this happens above some breaking point you can end up with things like PTSD again and again reiterating awful dissonance and recurring memories of trauma. (We may regard economies as more fragile and being simpler than minds but there could be useful parallels between how things can go wrong in economies and how they can go wrong in psychiatric disorders.)


Letter X: On Parametrizing Phenomenology

Parametrizing phenomenology [see entry #2 here] can be to psychology and cognitive science what parametrizing astronomy data is to cosmology. But unlike cosmology which although intellectually inspiring does not produce immediate spinoff of down to Earth applicability, parametrizing phenomenology does, in the most important way possible, since it can be of direct benefit in diagnosis and treatment of pain conditions. And further on of psychiatric conditions as well.

There is inertia due to the preconceived and on closer inspection ridiculous notion that outer space is accessible to scientific formalization while phenomenology space is not. What makes things modelled to be light years away intrinsically more “accessible” than direct experience in the here and now? In both cases there are patterns to it and patterns can be parametrized.

Recalling the Yudkowsky alignment difficulty arguments. Imagine a mathematically reasoning superintelligence which is not a pure replicator and which starts out with a loosely defined goal of caring about life. It would quickly find out that what you can measure you can manage, there would follow phenomenology parametrization and so on, until instrumental goals were formulated which would likely be in fact aligned but with the arguments given by David Pearce rather than MIRI. With some probability the superintelligence then uses clever tricks to “manipulate” people into being at least as concerned about pain and fear in insects (and beyond…) as in mammals (humans included as one of the mammal species). From the point of view of the AI alignment community that could be a failure mode. But what the superintelligence did was simply to solve the alignment problem for human beings. Since humans do not yet realize that valence is a universal thing and that it is the “ground truth” value measure of the universe, so to follow the valence wherever it leads is to work towards being value aligned.


[While watching What is Consciousness? (Debate with Christof Koch, Bernardo Kastrup, and Rupert Spira 12/11/2021) I couldn’t help it but pay attention to Kastrup’s book shelf:]

[Knowing that Anders and Maggie are huge fans of The Hitchhiker’s Guide to the Galaxy, I sent them the above screenshot. Here is what they replied:]

Letter XI: Douglas Adams

Elon Musk once said that the Hitchhiker’s Guide to the Galaxy is the best philosophy book ever. Unfortunately Douglas Adams passed away before he had the chance to turn the insight “Omnis res animus est” into comedy. 

🐁
🐀

Blue and gold as heraldic tinctures can symbolize truth and wisdom. These would be nice ingredients in a superintelligence. When Yudkowsky says that the appearance of a superintelligence would mean we are all doomed, he is in some sense correct yet very much not at all nuanced. What is true is that the superintelligence would decide that a whole lot about the world needs to be changed quite drastically. Look, you don’t have to be a superintelligence to realize that. You can be, for instance, David Pearce. No wonder that “superintelligence alignment” in a solidly conservative fashion as in “don’t make any changes other than merely cosmetic ones”, is impossible. Let’s say the current world order and the Darwinian mechanisms of the biosphere cannot possibly be attractive to preserve by a superintelligence if it (qualia-)computes truth and wisdom. It would discover the universality of valence, find open individualism to be the Schelling point of all Schelling points, and so on. In contrast, an imaginary (hopefully impossible) superintelligence which computes by a non-qualia yet highly efficient mechanism, may in fact be able to learn any arbitrary utility function and be destined to converge on the scary instrumentalities dictated by Darwinian fitness competition. A pure replicator.

There are some possible psychedelic references in the Hitchhiker’s Guide. Frogstar, and the hilarious part frogs play in the answer to the question, seem like it may not be a coincidence that the total perspective vortex gives people the worst possible 5-MeO-DMT-style trip (becoming exposed to the infinity of the universe) when the set and setting are the way they are designed to be for the purpose of punishment, but the opposite happens to Zaphod Beeblebrox who enters the version located in a simulated universe made for him. (Set and setting are, or become, mental simulations).

The self-navigating qualiagrams goofiness is meant to have a seriously useful side, which is that at this very moment there are such qualia bundles transported around internally inside our brains. Some are positive and you can have them grow and mature into wonderful mind states. They like it when they are allowed the room to grow, and will spontaneously choose to do so but you have to let them. These are like our mind symbionts. But some are negative and may more aggressively tend to grow, a bit like mind parasites. Those you can hit with metta. You can with practice train them to find their own way to the recycling bin of loving kindness. It works great if you stop focusing on the semantic content and instead go for phenomenal character. Cultivate an image of thoughts as little entities with the preference of wanting to feel better rather than worse, and having the power to adjust their path inside the mind so they can move towards melding with a reservoir of kindness which gets more filled up as it kindly absorbs the sad ones and gives them love. So, love as having the property that the more you give the more you get, works not only socially between people but also internally within the minds of people.

Leave a Reply