Qualia Computing at Consciousness Hacking (June 7th 2017)

I am delighted to announce that I will be presenting at Consciousness Hacking in San Francisco on 2017/6/7 (YMD notation).

Consciousness Hacking (CoHack) is an extremely awesome community that blends a genuine interest in benevolence, scientific rationality, experiential spirituality, self-experimentation, and holistic wellbeing together with an unceasing focus on consciousness. Truth be told, CohHack is one of the reasons why I love living in the Bay Area.

Here are the relevant event links: Eventbrite, FacebookMeetup.

And the event description:

What would happen if a bliss technology capable of inducing a constant MDMA-like state of consciousness with no negative side effects were available? What makes an experience good or bad? Is happiness a spiritual trick, or is spirituality a happiness trick?

At this month’s speaker presentation, Consciousness Hacking invites Data Science Engineer, Andrés Gómez Emilsson to discuss current research, including his own, concerning the measurement of bliss, how blissful brain states can be induced, and what implications this may have on quality of life and our relationship with the world around us.

Emilsson’s research aims to create a mathematical theory of the pleasure-pain axis that can take information about a person’s brain at a given point in time and return the approximate (or even true) level of happiness and suffering for that person. Emilsson will explore two dimensions that have been studied in affective neuroscience for decades:

  • Arousal: how much energy and activation a given emotion has
  • Valence: the “feel good or feel bad” dimension of emotion

If the purpose of life is to feel happy and to make others happy, then figuring out how valence is implemented in the brain may take us a long way in that direction. Current approaches to valence, while helpful, usually don’t address the core of the problem (ie. usually just measuring the symptoms of pleasure such as the neurotransmitters that trigger it, brain regions, positive reinforcement, etc. rather than getting at the experience of pleasure itself).

A real science of valence would not only be able to integrate and explain why the things people report as pleasurable are pleasant, it would also make a precise, empirically falsifiable hypothesis about whether arbitrary brain states will feel good or bad. This is what Emilsson aims to do.

You will take away:

  • An understanding about the current scientific consensus on the nature of happiness in the brain, and why it is incomplete
  • A philosophical case for both the feasibility and desirability of a world devoid of intense suffering
  • A new candidate mathematical formula that can be used to predict the psychological wellbeing of a brain at a given point in time
  • An argument for why bliss technology that puts us in a constant MDMA-like state of consciousness with no negative side effects is likely to become available within the next two to five decades
  • The opportunity to network with other people who are serious about figuring out the meaning of life through introspection and neuroscience

About our speaker:

Andrés Gómez Emilsson was born in México City in 1990. From an early age, he developed an interest in philosophy, mathematics, and science, leading him to compete nationally and internationally in Math and Science Olympiads. At 16, his main interest was mathematics, but after an unexpected “mystical experience”, he turned his attention to consciousness and the philosophical problems that it poses. He studied Symbolic Systems with an Artificial Intelligence concentration at Stanford, and later finished a masters in Computational Psychology at the same university. During his time at Stanford he co-founded the Stanford Transhumanist Association and became good friends with transhumanist philosopher David Pearce, taking on the flag of the Hedonistic Imperative (HI). In order to pursue the long-term goals of HI, his current primary intellectual interest is to reverse-engineer the functional, biochemical and/or quantum signatures of pure bliss.

He is currently working at a Natural Language Processing company in San Francisco, creating quantitative measures of employee happiness, productivity, and ethics at companies, with the long-term intent of creating a consciousness research institute that’s also a great place to work for (i.e. one in which employees are happy, productive, and ethical). In his free time he develops psychophysical tools to study the computational properties of consciousness.


6:30: Check in, snacks

6:45: Structured schmoozing

6:55: Event intro and meditation

7:00: Andrés Gómez Emilsson

7:50: Break

8:00: Break-out Sessions (small group discussion)

9:00: Break-out Recap

9:15: Closing meditation

About our venue:

ECO-SYSTM is a dynamic community of creative professionals, startups, and freelancers, founded on the idea that entertainment, creativity and business can come together to offer a truly unique work experience for Bay Area professionals. Check out membership plans here: http://eco-systm.com/


Principia Qualia: Part II – Valence

Extract from Principia Qualia (2016) by my colleague Michael E. Johnson (from Qualia Research Institute). This is intended to summarize the core ideas of chapter 2, which proposes a precise, testable, simple, and so far science-compatible theory of the fundamental nature of valence (also called hedonic tone or the pleasure-pain axis; what makes experiences feel good or bad).


VII. Three principles for a mathematical derivation of valence

We’ve covered a lot of ground with the above literature reviews, and synthesizing a new framework for understanding consciousness research. But we haven’t yet fulfilled the promise about valence made in Section II- to offer a rigorous, crisp, and relatively simple hypothesis about valence. This is the goal of Part II.

Drawing from the framework in Section VI, I offer three principles to frame this problem: ​

1. Qualia Formalism: for any given conscious experience, there exists- in principle- a mathematical object isomorphic to its phenomenology. This is a formal way of saying that consciousness is in principle quantifiable- much as electromagnetism, or the square root of nine is quantifiable. I.e. IIT’s goal, to generate such a mathematical object, is a valid one.

2. Qualia Structuralism: this mathematical object has a rich set of formal structures. Based on the regularities & invariances in phenomenology, it seems safe to say that qualia has a non-trivial amount of structure. It likely exhibits connectedness (i.e., it’s a unified whole, not the union of multiple disjoint sets), and compactness, and so we can speak of qualia as having a topology.

More speculatively, based on the following:

(a) IIT’s output format is data in a vector space,

(b) Modern physics models reality as a wave function within Hilbert Space, which has substantial structure,

(c) Components of phenomenology such as color behave as vectors (Feynman 1965), and

(d) Spatial awareness is explicitly geometric,

…I propose that Qualia space also likely satisfies the requirements of being a metric space, and we can speak of qualia as having a geometry.

Mathematical structures are important, since the more formal structures a mathematical object has, the more elegantly we can speak about patterns within it, and the closer our words can get to “carving reality at the joints”. ​

3. Valence Realism: valence is a crisp phenomenon of conscious states upon which we can apply a measure.

–> I.e. some experiences do feel holistically better than others, and (in principle) we can associate a value to this. Furthermore, to combine (2) and (3), this pleasantness could be encoded into the mathematical object isomorphic to the experience in an efficient way (we should look for a concise equation, not an infinitely-large lookup table for valence). […]


I believe my three principles are all necessary for a satisfying solution to valence (and the first two are necessary for any satisfying solution to consciousness):

Considering the inverses:

If Qualia Formalism is false, then consciousness is not quantifiable, and there exists no formal knowledge about consciousness to discover. But if the history of science is any guide, we don’t live in a universe where phenomena are intrinsically unquantifiable- rather, we just haven’t been able to crisply quantify consciousness yet.

If Qualia Structuralism is false and Qualia space has no meaningful structure to discover and generalize from, then most sorts of knowledge about qualia (such as which experiences feel better than others) will likely be forever beyond our empirical grasp. I.e., if Qualia space lacks structure, there will exist no elegant heuristics or principles for interpreting what a mathematical object isomorphic to a conscious experience means. But this doesn’t seem to match the story from affective neuroscience, nor from our everyday experience: we have plenty of evidence for patterns, regularities, and invariances in phenomenological experiences. Moreover, our informal, intuitive models for predicting our future qualia are generally very good. This implies our brains have figured out some simple rules-of-thumb for how qualia is structured, and so qualia does have substantial mathematical structure, even if our formal models lag behind.

If Valence Realism is false, then we really can’t say very much about ethics, normativity, or valence with any confidence, ever. But this seems to violate the revealed preferences of the vast majority of people: we sure behave as if some experiences are objectively superior to others, at arbitrarily-fine levels of distinction. It may be very difficult to put an objective valence on a given experience, but in practice we don’t behave as if this valence doesn’t exist.


VIII. Distinctions in qualia: charting the explanation space for valence

Sections II-III made the claim that we need a bottom-up quantitative theory like IIT in order to successfully reverse-engineer valence, Section VI suggested some core problems & issues theories like IIT will need to address, and Section VII proposed three principles for interpreting IIT-style output:

  1. We should think of qualia as having a mathematical representation,
  2. This mathematical representation has a topology and probably a geometry, and perhaps more structure, and
  3. Valence is real; some things do feel better than others, and we should try to explain why in terms of qualia’s mathematical representation.

But what does this get us? Specifically, how does assuming these three things get us any closer to solving valence if we don’t have an actual, validated dataset (“data structure isomorphic to the phenomenology”) from *any* system, much less a real brain?

It actually helps a surprising amount, since an isomorphism between a structured (e.g., topological, geometric) space and qualia implies that any clean or useful distinction we can make in one realm automatically applies in the other realm as well. And if we can explore what kinds of distinctions in qualia we can make, we can start to chart the explanation space for valence (what ‘kind’ of answer it will be).

I propose the following four distinctions which depend on only a very small amount of mathematical structure inherent in qualia space, which should apply equally to qualia and to qualia’s mathematical representation:

  1. Global vs local
  2. Simple vs complex
  3. Atomic vs composite
  4. Intuitively important vs intuitively trivial


Takeaways: this section has suggested that we can get surprising mileage out of the hypothesis that there will exist a geometric data structure isomorphic to the phenomenology of a system, since if we can make a distinction in one domain (math or qualia), it will carry over into the other domain ‘for free’. Given this, I put forth the hypothesis that valence may plausibly be a simple, global, atomic, and intuitively important property of both qualia and its mathematical representation.

IX. Summary of heuristics for reverse-engineering the pattern for valence

Reverse-engineering the precise mathematical property that corresponds to valence may seem like finding a needle in a haystack, but I propose that it may be easier than it appears. Broadly speaking, I see six heuristics for zeroing in on valence:

A. Structural distinctions in Qualia space (Section VIII);

B. Empirical hints from affective neuroscience (Section I);

C. A priori hints from phenomenology;

D. Empirical hints from neurocomputational syntax;

E. The Non-adaptedness Principle;

F. Common patterns across physical formalisms (lessons from physics). None of these heuristics determine the answer, but in aggregate they dramatically reduce the search space.

IX.A: Structural distinctions in Qualia space (Section VIII):

In the previous section, we noted that the following distinctions about qualia can be made: Global vs local; Simple vs complex; Atomic vs composite; Intuitively important vs intuitively trivial. Valence plausibly corresponds to a global, simple, atomic, and intuitively important mathematical property.


Music is surprisingly pleasurable; auditory dissonance is surprisingly unpleasant. Clearly, music has many adaptive signaling & social bonding aspects (Storr 1992; Mcdermott and Hauser 2005)- yet if we subtract everything that could be considered signaling or social bonding (e.g., lyrics, performative aspects, social bonding & enjoyment), we’re still left with something very emotionally powerful. However, this pleasantness can vanish abruptly- and even reverse– if dissonance is added.

Much more could be said here, but a few of the more interesting data points are:

  1. Pleasurable music tends to involve elegant structure when represented geometrically (Tymoczko 2006);
  2. Non-human animals don’t seem to find human music pleasant (with some exceptions), but with knowledge of what pitch range and tempo their auditory systems are optimized to pay attention to, we’ve been able to adapt human music to get animals to prefer it over silence (Snowdon and Teie 2010).
  3. Results suggest that consonance is a primary factor in which sounds are pleasant vs unpleasant in 2- and 4-month-old infants (Trainor, Tsang, and Cheung 2002).
  4. Hearing two of our favorite songs at once doesn’t feel better than just one; instead, it feels significantly worse.

More generally, it feels like music is a particularly interesting case study by which to pick apart the information-theoretic aspects of valence, and it seems plausible that evolution may have piggybacked on some fundamental law of qualia to produce the human preference for music. This should be most obscured with genres of music which focus on lyrics, social proof & social cohesion (e.g., pop music), and performative aspects, and clearest with genres of music which avoid these things (e.g., certain genres of classical music).


X. A simple hypothesis about valence

To recap, the general heuristic from Section VIII was that valence may plausibly correspond to a simple, atomic, global, and intuitively important geometric property of a data structure isomorphic to phenomenology. The specific heuristics from Section IX surveyed hints from a priori phenomenology, hints from what we know of the brain’s computational syntax, introduced the Non-adaptedness Principle, and noted the unreasonable effectiveness of beautiful mathematics in physics to suggest that the specific geometric property corresponding to pleasure should be something that involves some sort of mathematically-interesting patterning, regularity, efficiency, elegance, and/or harmony.

We don’t have enough information to formally deduce which mathematical property these constraints indicate, yet in aggregate these constraints hugely reduce the search space, and also substantially point toward the following:

Given a mathematical object isomorphic to the qualia of a system, the mathematical property which corresponds to how pleasant it is to be that system is that object’s symmetry.


XI. Testing this hypothesis today

In a perfect world, we could plug many peoples’ real-world IIT-style datasets into a symmetry detection algorithm and see if this “Symmetry in the Topology of Phenomenology” (SiToP) theory of valence successfully predicted their self-reported valences.

Unfortunately, we’re a long way from having the theory and data to do that.

But if we make two fairly modest assumptions, I think we should be able to perform some reasonable, simple, and elegant tests on this hypothesis now. The two assumptions are:

  1. We can probably assume that symmetry/pleasure is a more-or-less fractal property: i.e., it’ll be evident on basically all locations and scales of our data structure, and so it should be obvious even with imperfect measurements. Likewise, symmetry in one part of the brain will imply symmetry elsewhere, so we may only need to measure it in a small section that need not be directly contributing to consciousness.
  2. We can probably assume that symmetry in connectome-level brain networks/activity will roughly imply symmetry in the mathematical-object-isomorphic-to-phenomenology (the symmetry that ‘matters’ for valence), and vice-versa. I.e., we need not worry too much about the exact ‘flavor’ of symmetry we’re measuring.

So- given these assumptions, I see three ways to test our hypothesis:

1. More pleasurable brain states should be more compressible (all else being equal).

Symmetry implies compressibility, and so if we can measure the compressibility of a brain state in some sort of broad-stroke fashion while controlling for degree of consciousness, this should be a fairly good proxy for how pleasant that brain state is.


2. Highly consonant/harmonious/symmetric patterns injected directly into the brain should feel dramatically better than similar but dissonant patterns.

Consonance in audio signals generally produces positive valence; dissonance (e.g., nails-on-a-chalkboard) reliably produces negative valence. This obviously follows from our hypothesis, but it’s also obviously true, so we can’t use it as a novel prediction. But if we take the general idea and apply it to unusual ways of ‘injecting’ a signal into the brain, we should be able to make predictions that are (1) novel, and (2) practically useful.

TMS is generally used to disrupt brain functions by oscillating a strong magnetic field over a specific region to make those neurons fire chaotically. But if we used it on a lower-powered, rhythmic setting to ‘inject’ a symmetric/consonant pattern directly into parts of the brain involved directly with consciousness, the result should produce good feeling- or at least, much better valence than a similar dissonant pattern.

Our specific prediction: direct, low-power, rhythmic stimulation (via TMS) of the thalamus at harmonic frequencies (e.g., @1hz+2hz+4hz+6hz+8hz+12hz+16hz+24hz+36hz+48hz+72hz+96hz+148hz) should feel significantly more pleasant than similar stimulation at dissonant frequencies (e.g., @1.01hz+2.01hz+3.98hz+6.02hz+7.99hz+12.03hz+16.01hz+24.02hz+35.97hz+48.05hz+72.04hz+95.94hz+ 147.93hz).


3. More consonant vagus nerve stimulation (VNS) should feel better than dissonant VNS.

The above harmonics-based TMS method would be a ‘pure’ test of the ‘Symmetry in the Topology of Phenomenology’ (SiToP) hypothesis. It may rely on developing custom hardware and is also well outside of my research budget.

However, a promising alternative method to test this is with consumer-grade vagus nerve stimulation (VNS) technology. Nervana Systems has an in-ear device which stimulates the Vagus nerve with rhythmic electrical pulses as it winds its way past the left ear canal. The stimulation is synchronized with either user-supplied music or ambient sound. This synchronization is done, according to the company, in order to mask any discomfort associated with the electrical stimulation. The company says their system works by “electronically signal[ing] the Vagus nerve which in turn stimulates the release of neurotransmitters in the brain that enhance mood.”

This explanation isn’t very satisfying, since it merely punts the question of why these neurotransmitters enhance mood, but their approach seems to work– and based on the symmetry/harmony hypothesis we can say at least something about why: effectively, they’ve somewhat accidentally built a synchronized bimodal approach (coordinated combination of music+VNS) for inducing harmony/symmetry in the brain. This is certainly not the only component of how this VNS system functions, since the parasympathetic nervous system is both complex and powerful by itself, but it could be an important component.

Based on our assumptions about what valence is, we can make a hierarchy of predictions:

  1. Harmonious music + synchronized VNS should feel the best;
  2. Harmonious music + placebo VNS (unsynchronized, simple pattern of stimulation) should feel less pleasant than (1);
  3. Harmonious music + non-synchronized VNS (stimulation that is synchronized to a different kind of music) should feel less pleasant than (1);
  4. Harmonious music + dissonant VNS (stimulation with a pattern which scores low on consonance measures such as (Chon 2008) should feel worse than (2) and (3));
  5. Dissonant auditory noise + non-synchronized, dissonant VNS should feel pretty awful.

We can also predict that if a bimodal approach for inducing harmony/symmetry in the brain is better than a single modality, a trimodal or quadrimodal approach may be even more effective. E.g., we should consider testing the addition of synchronized rhythmic tactile stimulation and symmetry-centric music visualizations. A key question here is whether adding stimulation modalities would lead to diminishing or synergistic/accelerating returns.

The Most Important Philosophical Question

Albert Camus famously claimed that the most important philosophical question in existence was whether to commit suicide. I would disagree.

For one, if Open Individualism is true (i.e. that deep down we are all one and the same consciousness) then ending one’s life will not accomplish much. The vast majority of “who you are” will remain intact, and if there are further problems to be solved, and questions to be answered, doing this will simply delay your own progress. So at least from a certain point of view one could argue that the most important question is, instead, the question of personal identity. I.e. Are you, deep down, an individual being who starts existing when you are born and stops existing when you die (Closed Individualism), something that exists only for a single time-slice (Empty Individualism), or maybe something that is one and the same with the rest of the universe (Open Individualism)?

I think that is a very important question. But probably not the most important one. Instead, I’d posit that the most important question is: “What is good, and is there a ground truth about it?”

In the case that we are all one consciousness maybe what’s truly good is whatever one actually truly values from a first-person point of view (being mindful, of course, of the deceptive potential that comes from the Tyranny of the Intentional Object). And in so far as this has been asked, I think that there are two remaining possibilities: Does ultimate value come down to the pleasure-pain axis, or does it come down to spiritual wisdom?

Thus, in this day and age, I’d argue that the most important philosophical (and hence most important, period) question is: “Is happiness a spiritual trick, or is spirituality a happiness trick?”

What would it mean for happiness to be a spiritual trick? Think, for example, of the possibility that the reason why we exist is because we are all God, and God would be awfully bored if It knew that It was all that ever existed. In such a case, maybe bliss and happiness comes down to something akin to “Does this particular set of life experiences make God feel less lonely”? Alternatively, maybe God is “divinely self-sufficient”, as some mystics claim, and all of creation is “merely a plus on top of God”. In this case one could think that God is the ultimate source of all that is good, and thus bliss may be synonymous with “being closer to God”. In turn, as mystics have claimed over the ages, the whole point of life is to “get closer to God”.

Spirituality, though, goes beyond God: Within (atheistic) Buddhism the view that “bliss is a spiritual trick” might take another form: Bliss is either “dirty and a sign of ignorance” (as in the case of karma-generating pleasure) or it is “the results of virtuous merit conducive to true unconditioned enlightenment“. Thus, the whole point of life would be to become free from ignorance and reap the benefits of knowing the ultimate truth.

And what would it mean for spirituality to be a happiness trick? In this case one could imagine that our valence (i.e. our pleasure-pain axis) is a sort of qualia variety that evolution recruited in order to infuse the phenomenal representation of situations that predict either higher or lower chances of making copies of oneself (or spreading one’s genes, in the more general case of “inclusive fitness”). If this is so, it might be tempting to think that bliss is, ultimately, not something that “truly matters”. But this would be to think that bliss is “nothing other than the function that bliss plays in animal behavior”, which couldn’t be further from the truth. After all, the same behavior could be enacted by many methods. Instead, the raw phenomenal character of bliss reveals that “something matters in this universe”. Only people who are anhedonic (or are depressed) will miss the fact that “bliss matters”. This is self-evident and self-intimating to anyone currently experiencing ecstatic rapture. In light of these experiences we can conclude that if anything at all does matter, it has to do with the qualia varieties involved in the experiences that feel like the world has meaning. The pleasure-pain axis makes our existence significant.

Now, why do I think this is the most important question? IF we discover that happiness is a spiritual trick and that God is its source then we really ought to follow “the spiritual path” and figure out with science “what is it that God truly wants”. And under an atheistic brand of spirituality, what we ought to figure out is the laws of valence-charged spiritual energy. For example, if reincarnation and karma are involved in the expected amount of future bliss and suffering, so be it. Let’s all become Bodhisattvas and help as many sentient beings as possible throughout the eons to come.

On the other hand, IF we discover (and can prove with a good empirical argument) that spirituality is just the result of changes in valence/happiness, then settling on this with a high certainty would change the world. For starters, any compassionate (and at least mildly rational) Buddhist would then come along and help us out in the pursuit of creating a pan-species welfare state free of suffering with the use of biotechnology. I.e. The 500 odd million Buddhists world-wide would be key allies for the Hedonistic Imperative (a movement that aims to eliminate suffering with biotechnology).

Recall Dalai Lama’s quote: “If it was possible to become free of negative emotions by a riskless implementation of an electrode – without impairing intelligence and the critical mind – I would be the first patient.” [Dalai Lama (Society for Neuroscience Congress, Nov. 2005)].

If Buddhist doctrine concerning the very nature of suffering and its causes is wrong from a scientific point of view and we can prove it with an empirically verified physicalist paradigm, then the very Buddhist ethic of “focusing on minimizing suffering” ought to compel Buddhists throughout the world to join us in the battle against suffering by any means necessary. And most likely, given the physicalist premise, this would take the form of creating a technology that puts us all in a perpetual pro-social clear-headed non-addictive MDMA-like state of consciousness (or, in a more sophisticated vein, a well-balanced version of rational wire-heading).