All At Sea?
Max Tegmark and the Phenomenal Binding Problem.
1. The Binding / Combination ProblemThe hypothesis that organic minds are quantum computers running at anything up to sextillions of quantum-coherent frames-per-second might have been called Tegmark's Conjecture in honour of the physicist Max Tegmark on whose mathematical spadework for approximate CNS decoherence times this paper relies. Alas the title probably won't work, not merely because Tegmark treats quantum-coherent neuronal superpositions at these time-frames as a reductio ad absurdum of quantum mind, but also because Tegmark claims that there is no binding problem in the first instance. In "Why the brain is probably not a quantum computer"(1), Tegmark dismisses the binding problem in a couple of paragraphs. In Our Mathematical Universe (2014), written for a popular audience, he drops the "probably". Tegmark's brisk purported dissolution of the problem is worth quoting verbatim:In common with Eddington and other perceptual direct realists, Tegmark simply helps himself to "the ocean" without explaining how a pack of allegedly discrete, decohered classical neurons - our Jamesian "mind-dust" - generates a phenomenally bound expanse of wave-tossed sea in front of one's phenomenal nose - a misappropriation that offers all the advantages of theft over honest computational toil. There is no binding problem for the perceptual direct realist only because a mysterious entity called the "observer" somehow directly accesses a phenomenally bound extra-cranial ocean - an expanse of phenomenal sea lying not just beyond the familiar empirical skull internal to one's world-simulation, but beyond the transcendental skull that encloses one's mind-brain and the world-simulation it runs. By contrast, the world-simulationist recognises that the expanse of phenomenal ocean in front of one's phenomenal body-image is an aspect of a neuronal simulation that the mind-brain is running. This world-simulation is sculpted, but not generated, during waking life by cross-modally matched inputs from the optic auditory and other peripheral nerves. When one is dreaming, the phenomenal ocean is likewise generated by the CNS but runs autonomously - its contents unselected and unconstrained by the daily bombardment of input from the peripheral nervous system. The challenge for the classical computationalist is to explain how, if individual nerve cells / "psychons" are essentially discrete, decohered, classical objects separated by synaptic clefts, the individual nerve cells that mediate the micro-qualia of colour, motion, edges and other features of the mind-dependent macroscopic world combine in classically forbidden ways without invoking neuronal quantum coherence of otherwise spatially distributed feature-processors."4.4.3. The binding problem
One of the motivations for models with quantum coherence in the brain was the so-called binding problem. In the words of James [77,78], ``the only realities are the separate molecules, or at most cells. Their aggregation into a `brain' is a fiction of popular speech''. James' concern, shared by many after him, was that consciousness did not seem to be spatially localized to any one small part of the brain, yet subjectively feels like a coherent entity. Because of this, Stapp  and many others have appealed to quantum coherence, arguing that this could make consciousness a holistic effect involving the brain as a whole.
However, non-local degrees of freedom can be important even in classical physics, For instance, oscillations in a guitar string are local in Fourier space, not in real space, so in this case the ``binding problem'' can be solved by a simple change of variables. As Eddington remarked , when observing the ocean we perceive the moving waves as objects in their own right because they display a certain permanence, even though the water itself is only bobbing up and down. Similarly, thoughts are presumably highly non-local excitation patterns in the neural network of our brain, except of a nonlinear and much more complex nature. In short, this author feels that there is no binding problem."
The hopelessness of the Eddington / Tegmark ocean-wave analogy can again be highlighted using the example of skull-bound American mind-brains who agree to participate in an experiment. Once again, the inter-communicating but skull-bound minds in this thought-experiment do duty for membrane bound-neuronal "pixels" of experience. Using this analogy, it's tempting, but disastrously misconceived, to imagine an airborne observer discerning collective wave-like patterns displayed on the ground below. The airborne observer can't make out individual Americans bobbing up and down. Yet (s)he can see their experimentally choreographed wavelike behaviour - semi-permanent waves that are "objects in their own right" as Tegmark puts it - generated autonomously (cf. dreaming consciousness) or in response to external input (cf. waking consciousness). Perhaps, runs this disastrously misconceived analogy, we may imagine the Americans performing some sort of glorified Mexican wave. Individual "pixels" of stationary Americans bobbing up and down can't be discerned, merely dynamical patterns of semi-permanent waves.
Not in the slightest, alas.
For the wave analogy to work, there can be no airborne spectator discerning semi-permanent waves indiscernible to their constituent skull-bound American minds below. Reverting to the real-life mind-brain, there is no external spectator notionally "observing" semi-permanent patterns of firing neurons / psychons from afar. All that exists, supposedly, are individual classical membrane-bound neurons or neuronal "pixels" of consciousness separated by the synaptic clefts of the connectome. Insofar as neurons are classical physical objects, all that any individual decohered neuron can "know", supposedly, is the effects of ligand binding to receptors on its plasma membrane, or if the synapse is electrical rather than chemical, the voltage changes induced via the gap junctions on its post-synaptic membranes. If the neuron always functioned as an effectively classical object rather than the component of a quantum-coherent neuronal superposition, then the neuron would no more have access to the state of distant neurons mediating colour, motion, edges (etc) implicated in bound phenomenal objects than skull-bound American minds have access to the inner states of their distant skull-bound counterparts. Indeed, just as philosophically-inclined skull-bound Americans may worry about the Problem of Other Minds, a notional classical neuron in a notionally classical neural network confronts the Problem of Other "Psychons" / "Mind-Dust". How does the supposedly classical neuron "know" that it's not a neuron in a test-tube rather than a neuron in a brain? How does the neuron "know" that's it's not surrounded by prosthetic silicon surrogates? How does the neuron "know" that other neurons in the network support rudimentary experience; and insofar as they do support rudimentary experience, how is such experience computationally and/or phenomenally relevant to the properties of the conscious neuronal "pixel"?
The epistemological worries of a Cartesian neuron can be treated as fanciful - neurons aren't really sceptical epistemic agents - but the ontological reality of phenomenal binding is a manifest property of our minds. If neurons were discrete classical objects, then the state of distant neurons in the "pack" would be phenomenally irrelevant. In practice, the phenomenal signature of macroscopic neuronal quantum coherence - and the phenomenal signature that the mind-brain is a quantum computer - lies in front of one's virtual eyes in the form of the dynamical objects populating ones visual field ("local" binding) apprehended by a unitary phenomenal self ("global" binding) that they partially instantiate.
The binding problem is most commonly formulated using visual examples of distributed feature-processing. But it's endemic to our minds - if the naive "pack of neurons" story of the CNS is true. Again by way of analogy, let's consider millions of individual classical skull-bound Americans minds each experiencing a pinprick. Assuming reductive physicalism, the multiple pinpricks don't combine into a pan-continental experience of agony, even with an ultrafast network of reciprocal electromagnetic communications. Likewise, a classical decohered neuron undergoing a classical micro-pinprick of phenomenal pain couldn't contribute to a state of phenomenal agony in virtue of proximity to a multitude of other classical neuronal micro-pinpricks. In stark contrast, a sequence of individual quantum coherent neuronal macro-superpositions of what would otherwise be discrete neuronal pinpricks is excruciating - says the quantum-mind theorist. Or to use yet another example, a multitude of skull-bound American minds each experiencing an individual musical note does not generate a pan-continental musical symphony enjoyed by a unitary subject of experience. By contrast, a sequence of individual, quantum coherent neuronal macro-superpositions of what would otherwise be discrete neuronal musical notes is Beethoven's Fifth. Critically - and here is where mere philosophising becomes a testable scientific hypothesis - the physical signature of these quantum-coherent macro-superpositions will be experimentally detectable in the CNS when tools of interferometry sensitive enough indirectly to probe their nature are developed: our probes are currently too coarse-grained. Thermally-induced decoherence, collisional decoherence, and dephasing due to inertial forces and vibrations all conspire to make experimental detection and controlled manipulation of the process of neuronal decoherence - let alone pan-cerebral hemispheric decoherence - a challenge to human ingenuity. Yet the mind-brain is no more classical at fine-grained temporal resolutions than it's classical at fine-grained spatial resolutions. There is no evidence the superposition principle ever breaks down. For sure, it sounds bizarre to claim that quantum superposition, not classical synchrony, is the essence of mind. However, the proposal is not merely a "philosophical" opinion but also an experimentally falsifiable conjecture.
2. Dynamical TimescalesOn the face of it, Max Tegmark is on firmer ground with his criticism that quantum mind conjectures get the "dynamical timescale" of our minds spectacularly wrong. The classically parallel connectionist systems ("neural networks") that we use to model cognition and perception capture the timescale of human information processing: "there is nothing fundamentally wrong with the current classical approach to neural network simulations."(1) According to Tegmark, whether we're investigating either the transitions between our serial logico-linguistic thought-processes, or the phenomenal binding of neuronal feature-processors into perceptual objects, makes no difference. The sub-femtosecond lifetime of neuronal superpositions leaves inadequate time to do any computational work. Action potentials, and signalling across chemical and electrical synapses, take place over milliseconds. Femtoseconds - let alone attoseconds, zeptoseconds or less - are not long enough to generate a unitary subject of experience, a cognitive agent."The calculations above enable us to address the question of whether cognitive processes in the brain constitute a classical or quantum system in the sense of Fig. 1. If we take the characteristic dynamical timescale for such processes to be τdyn ~ 10-2–100s (the apparent timescale of e.g., speech, thought and motor response), then a comparison of τdyn with τdec from Table 1 shows that processes associated with either conventional neuron firing or with polarization excitations in microtubules fall squarely in the classical category, by a margin exceeding 10 orders of magnitude. Neuron firing itself is also highly classical, since it occurs on a timescale τdyn ~ 10-3–10-4s. Even a kink-like microtubule excitation is classical by many orders of magnitude, since it traverses a short tubule on a timescale τdyn 5 × 10-7s."
If non-materialist physicalism is false, then this argument is decisive. Any "emergentist" theory of consciousness that invokes quantum mechanics to explain how nonexperiential fields of insentience can generate - somehow - conscious percepts is constrained by sub-femtosecond decoherence timescales, just as Tegmark argues in "Why the brain is probably not a quantum computer".
No such time constraint holds if (1) quantum mechanics is complete, and (2) experience discloses the intrinsic nature of the physical. Recall from textbook quantum mechanics how a superposition is an individual physical state, not a classical ensemble. According to a Schrödinger's neurons conjecture, the components of an individual superposition of neuronal feature-processors don't interact with each other to "create conscious experience". In quantum mechanics as distinct from everyday speech, interference and interaction are distinct. The linearity of the wave equation entails that Everett "branches" don't interact with each other: they interfere, as countless variants of the double-split experiment demonstrate. So if non-materialist physicalism is true, then a coherent neuronal superposition must be an individual macroexperience - regardless how short-lived that individual macroexperience must be, and irrespective of any computational or functional role that experience may or may not play in the CNS. What philosophers like to call "mereological nihilism" is vindicated - but not the sort of mereological nihilism that would entail the CNS consisted of individual neuronal pixels of Jamesian mind-dust. This flawed classical background assumption gives rise to the insoluble binding problem. The difference between a superposition of neuronal feature-processors and, for instance, a superconducting quantum interference device (SQUID) isn't just the immense disparity in their comparative duration. Rather, it's that the SQUID, consisting of a superposition merely of clockwise and anticlockwise magnetic-flux states, must be internally phenomenally simple, whereas the individual macro-experience of a neuronal superposition of diverse feature-processors is internally phenomenally differentiated: for example, the experience of a classical-looking pointer reading, or a live and classical-looking black cat. Or rather, this is the claim, incredible as it sounds. Note that although the interference signature of neuronal superpositions hasn't yet been experimentally detected, what is not being proposed here is a new theory of physics, or even the modification of an existing theory (cf. the Penrose–Hameroff Orch-OR model), but rather the significance for conscious mind of existing quantum physics.
Counterfactually, if the phase coherence of neuronal superpositions endured for milliseconds, then the possibility of such a perfect structural match to solve the binding problem would cry out for independent investigation via molecular matter-wave interferometry as an alternative to Chalmersian dualism. Phase coherence of neuronal superpositions doesn't remotely last so long, as Tegmark's calculations of sources and credible timescales of decoherence convincingly demonstrate. Unfortunately, Tegmark treats these sub-femtosecond neuronal decoherence times as a reductio of quantum mind rather than an experimental challenge. If pressed, Tegmark would presumably argue that all we'll discover in the CNS via tomorrow's molecular matter-wave interferometry is functionally meaningless "noise", not a perfect structural match between the phenomenology of our minds and physics. After all, the superposition principle doesn't break down in the CPU of a PC either; and your PC is not a quantum computer. Quite so; but unlike your PC, the computational architecture of our minds is an open question.
3. What Is An Observation?Tegmark has another argument against quantum mind. The processes of perception unfold over scores of milliseconds or more. Therefore,
Psychotic Binding versus Functional Binding"...we are never even close to being able to perceive superpositions of different perceptions. Reducing object decoherence (from Hoe) during measurement would make no difference, since decoherence would take place in the brain long before the transmission of the appropriate sensory input through sensory nerves had been completed." (4.2.5)Such an account presupposes an uncomplicated realism about perception. According to perceptual realism, subjects of experience ("observers") apprehend - somehow - mind-independent experimental apparatus, measuring devices, cats, and so forth in their local surroundings. However, subjects don't "perceive superpositions of different perceptions" - any more than they commune with determinate mind-independent classical objects. The external world is an inference, not an observation. Such "epistemological idealism" holds true independently of the far more speculative conjecture advanced here that experience discloses the intrinsic nature of the physical. On a world-simulation model of perceptual experience, the extracranial environment does not create phenomenal content, it selects phenomenal content internal to the CNS - again, via a complicated causal chain.
Traditional world-simulation models of perceptual experience are classical by default: they must explain why a pack of membrane-bound neurons aren't, at most, micro-experiential zombies, i.e. the phenomenal binding / combination problem as first clearly formulated by William James. By contrast, in our version of the world-simulation model, subjects of experience are instantiated by phase-coherent neuronal superpositions. These neuronal superpositions function as the raw material for our phenomenally bound world-simulations.
Tegmark would dismiss any such conjecture:"This means that we should strictly speaking not think of macrosuperpositions such as Eq. (28) as first forming and then decohering... - rather, subject decoherence is so fast that such superpositions decohere already during their process of formation."However, what's at issue here is not the fanciful existence of an individual phase-coherent superposition of your CNS and a perceived e.g. extracranial cat beyond your transcendental skull in some quantum analogue of a Vulcan mind-meld. We don't literally "observe" our local surroundings, or "perceive" mind-independent cats and shared public pointer-readings. Such an out-of-body feat would be physically impossible. Rather, inputs from peripheral sensory transducers in the extracranial environment pass up the optic (etc) nerve and go on to select the internal contents of one's phenomenally bound world-simulation - with a time-lag of anything from tens to hundreds of milliseconds. Thus Tegmark is right to say that awake, naturally evolved biological minds track only gross, fitness-relevant classical-like patterns in the local environment, not e.g. the heavily scrambled complex phase relationships as demanded by Everettian QM. Yet only a quantum mind could phenomenally bind distributed neuronal feature-processors into a world-simulation - a simulation whose phenomenal features may be described "from the inside" by an approximation of classical Newtonian physics. Thanks to quantum coherence, one's internal phenomenal world seems robustly classical - and robustly mind-independent.
But how? Assuming non-materialist physicalism is true, and the unmodified and unsupplemented unitary Schrödinger dynamics is correct, quantum coherence explains the brute existence of phenomenal binding. Yet if the superposition principle doesn't break down in the CNS, then why isn't "psychotic binding", i.e. psychotic minds and psychotic thought, all but ubiquitous? How can we theoretically derive from sub-femtosecond superpositions the multi-millisecond state-transitions of familiar neural network modelling, the slow processes of more-or-less orderly human serial thought, and the temporally coarse-grained updates to our world-simulations described by textbook neuroscience accounts of perception?
Here we must explore the implications of Wojciech Zurek’s “Quantum Darwinism"(2) for the CNS. The decoherence program of post-Everett QM pioneered by Zeh, Zurek et al. explains the emergence of a seemingly robust mind-independent classical world from the quantum world via a mechanism strongly analogous to Darwinian natural selection. This is no casual metaphor(3). So what happens when we apply quantum Darwinism to the CNS? The emergence of quasi-classical neurons - just as the neuroscience textbooks and light microscopy suggest. But note the "quasi". If we start by simply assuming - rather than rigorously deriving from QFT - the existence of decohered classical neurons, then we face the insoluble phenomenal binding / combination problem that drives David Chalmers to dualism. (cf. 7) In short, Schrödinger's neurons plus Zurek's Quantum Darwinism can “save the phenomena" and preserve monistic physicalism. Or rather, this is the proposal: experiment will decide: 6.
(September 2014, last updated December 2016)
Social Media Postings
The Abolitionist Project
The Biointelligence Explosion
MDMA: Utopian Pharmacology
Contact : Webmaster