Panpsychism is sometimes dismissed as a crazy view, but this reaction on its own is not a serious objection. While the view is counterintuitive to some, there is good reason to think that any view of consciousness must embrace some counterintuitive conclusion.
– Panpsychism and Panprotopanpsychism, David Chalmers (2011)
As Chalmers points out in this 2011 paper, any theory of consciousness will probably have counterintuitive conclusions. It should thus not come as surprise that almost every single consciousness scholar will be ridiculed as crazy by at least a minority of commentators. However, aside from omnipresent cognitive and affective biases, the vast majority of consciousness researchers are using their brains to their full capacity. Their search for understanding is sincere. It simply happens that the problem is, indeed, very hard.
Thus, when someone who is otherwise rational and intelligent has weird views about consciousness, one of several things could be going on. Instead of dismissing the view outright, ask the following four questions:
- What conception of consciousness does this person have?
- What criteria does he or she believe that a theory of consciousness must satisfy?
- What information does this person know about, and how deeply is it being incorporated into the theory?
- What are the relevant implicit background assumptions that color one’s reasoning?
Asking these questions will help you sort out the root causes behind the differences in beliefs you and the theorist may have. It will, in turn, help you see how, in a sense, uncrazy the person may be.
I recently had the opportunity to practice asking these questions over and over again in the 2016 “Science of Consciousness” conference in Tucson, Arizona. Every single presenter, panelist and poster-er could be framed in such a way that he or she would look outright crazy. In reality, the reasons behind their views are, for the most part, tractable.
Instead of focusing on the individual craziness of the participants, it is more sensible, and indeed more accurate, to simply realize that the crazy step is the very first: to dare attempt to understand, as a human, what consciousness is.
Ok, so let us just agree that all participants, including me, are crazy for simply trying to make a contribution to this field. After all, our conscious mind evolved to maximize inclusive fitness in complex, Machiavellian social structures, so when we repurpose this machinery to investigate the intrinsic nature of consciousness we are bound to have serious challenges. Starting from this understanding will make it easier to have an open mind when evaluating the merits of different theories of consciousness proposed in this conference. Do not get too fixated on how counterintuitive the theories sound to you; focus on whether they are capable of satisfying at least some minimal requirements we would want from such theories.
Conversely, it could be argued that what is truly crazy is to stand idly at the center of this monstrous philosophical conundrum.
Physicalism.com
My friend and colleague David Pearce persuaded me to accompany him to this year’s instance of this conference. He submitted an abstract of his paper on consciousness and physicalism. If you have been to Qualia Computing before, you may recognize that I heavily draw from Pearce’s philosophy. Not only do we share the belief that the problem of suffering is an ethical emergency best addressed with biotechnology, but we also have substantially similar views about consciousness.
David Pearce (left) and Andrés Gómez Emilsson (right) at the 2016 Science of Consciousness conference
Our Conception of Consciousness
Consciousness is very hard to define. But we agree on something: every single experience is an instance of consciousness. The possible components of a conscious experience come from a wide variety of qualia spaces (e.g. the state-space of phenomenal color). Importantly, we do not restrict our conception of consciousness to high-level thought, reasoning or social cognition. In all likelihood, consciousness is extremely ancient (possibly preceding the Cambrian explosion), and it is present in every animal with a thalamus (if not every animal with a nervous system).
More poignantly, the true state-space of possible conscious experiences is unfathomably large. Not only does it include the mental states of every possible animal doing any conceivable activity, but it also includes the ineffable weirdness of LSD, DMT and ketamine, not to mention the countless varieties of consciousness that are yet to be discovered.
Theoretical Requirements
If it weren’t for David, I would probably still be a neuron-doctrine functionalist who believes that we may be just a few decades away from programming a full Artificial General Intelligence in silicon computers.
How did Pearce help change my mind? Well, it comes down to the second question: I used to have an impoverished set of constraints that a theory of consciousness would have to satisfy. The main addition is that I now take extremely seriously the phenomenal binding problem (also called the combination problem).
For the sake of clarity and intellectual honesty, here is the answer that David and I give to the second question:
Back when I was in high school, before meeting David in person, I used to believe that the phenomenal binding problem could be dissolved with a computational theory of consciousness. In brief, I perceived binding to be a straightforward consequence of implicit information processing.
In retrospect I cannot help but think: “Oh, how psychotic I must have been back then!” However, I am reminded that one’s ignorance is not explicitly represented in one’s conceptual framework.
Background Assumptions
In order to make sense both of physicalism.com and Qualia Computing, it makes sense to be explicit about the background assumptions that we hold. Without explaining them in depth, here are some key assumptions that color the way we think about consciousness:
- Events of conscious experience are ontologically unitary: The left and right side of your visual field are part of an integrated whole that stands as a natural unit.
- Physicalism: Physics is causally closed and it fully describes the behavior of the observable universe.
- Wavefunction realism: The decoherence program is the most parsimonious, scientific, and promising approach for interpreting quantum mechanics.
- Mereological Nihilism (also called Compositional Nihilism): Simply putting two objects A and B side by side will not make a new object “AB” appear ex nihilo.
- Qualia Realism: The various textures of qualia (phenomenal color, sounds, feelings of cold and heat, etc.) are not mere representations. On the contrary, our mind uses them to instantiate representations (this is an important difference).
- Causal efficacy: Consciousness is not standing idly by. It has definite causal effects in animals. In particular, there must be a causal pathway that allows us to discuss its existence.
- Qualia computing: The reason consciousness was recruited by natural selection is computational. In spite of its expensive caloric cost, consciousness improves the performance of fitness-relevant information processing tasks.
A Battle of Wits
A Broken Political Analogy
Naïvely, people may get the impression that there are only a few well-defined camps when it comes to scientific theories of consciousness. The layman’s conception of the explanation state-space tends to be profoundly impoverished: “Are you a scientific materialist, or one of those religious dualists?” In this sense, people may picture the discussions that go on in places like The Science of Consciousness conference as something akin to what happens in political debates. There may be a few fringe camps, but the bulk of the people are rooting for one (often very popular) party.
Magic: The Gathering analogy
Instead of imagining a political rally, I would ask you to imagine a Magic: The Gathering tournament. For those unfamiliar with this game: Magic is a card game with two competitive components. First, one selects a set of cards from a pool of allowed cards (which depends on the format one is playing). With these cards one constructs a deck. The cards within a deck tend to have synergistic interactions, and ultimately define a range of possible strategies that the player will be able to use.
And second: one can be better or worse at playing one’s deck. The skills required to play a deck properly often involve being good at estimating odds and probabilities, bluffing, and mind-reading. In terms of knowledge, one needs to be familiar with the sorts of decks that are common out there and the typical strategies that they are built around. This leads us to the concept of deck archetypes.
Types (Clusters)
Often referred to as the flavors of the month, tournament decks tend to cluster rather neatly into deck types. In brief, certain clusters of cards tend to work very well with each other, which means that they will appear together in decks with a frequency that is much higher than chance. Arguably the process of block design is in part responsible for the emergence of these clusters. But even if, I would argue, you were to select at random a pool of 500 Magic cards from its entire history, we would still see clusters emerge: strategizing, trial and error, memetics and the natural synergy between some cards would lead to this outcome.
Intuitively, the game should then be entirely dominated by the deck types that are the most powerful. However, how good a given deck is depends on two things: The synergy between its cards, and the nature of the deck it plays against. Thus, decks cannot be analyzed in isolation. Their competitiveness depends on the distribution of other deck types in tournaments.
Over the months, therefore, the density of various deck types evolves in response to past distributions of deck types. This distribution, and evolutionary process, is often referred to as the Metagame. The connection to evolutionary game theory is straightforward: After gauging the frequency of various deck types at a tournament, one may strategically decide to switch one’s deck type in order to have a higher expected performance.
Rogue
Some number of players tend to find playing common deck types boring or too cliché. In practice, the monetary cost for acquiring certain key cards for a given type may also push some players to develop their own unique deck type. It is rare for these decks to be top performers, but they cannot be ignored since they meaningfully contribute to the Magic ecosystem.
The Cards and Deck Types of Consciousness Theories
To make the analogy between Magic decks and theories of consciousness, we need to find a suitable interpretation for a card. In this case, I would posit that cards can be interpreted as either background assumptions, required criteria, emphasized empirical findings and interpretations of phenomena. Let’s call these, generally, components of a theory.
Like we see in Magic, we will also find that some components support each other while others interact neutrally or mutually exclude each other. For example, if one’s theory of consciousness explicitly rejects the notion that quantum mechanics influences consciousness, then it is irrelevant whether one also postulates that the Copenhagen interpretation of quantum mechanics is correct. On the other hand, if one identifies the locus of consciousness to be in the microtubules inside pyramidal cells, then the particular interpretation of quantum mechanics one has is of paramount importance.
In this particular conference, it seemed that the metagame was dominated by the following 8 theories, in (approximate) order of popularity (as it seemed to me):
- Integrated Information Theory (IIT)
- Orchestrated Objective Reduction (Orch OR)
- Prediction Error Minimization (PEM)
- Global Neuronal Workspace Theory (GNWS)
- Panprotopanpsychism (not explicitly named)
- Nondual Consciousness Monism (not explicitly named)
- Consciousness as the Result of Action-Oriented Cognition (not explicitly named)
- Higher Order Thought Theory (HOT)
David Pearce and I, together with perhaps up to ten other attendees, seemed to be playing a particularly rare rogue strategy: Panpsychism + Wavefunction realism + Quantum Coherence to Bind.
Other notable rogue types included: Transcendentalism + semantic nihilism, timeless + perspective-free functionalism, and, oddly, multi-draft theory of consciousness (which seems to have fallen out of favor for some reason).
Finally, it is worth mentioning that as far as this conference goes, it did not seem to be the case that any one theory was held by the majority of the participants. The plurality seemed to be held by IIT, which has a lot of interesting developments going for it.
Coming next: In the next article I will provide a chronology of the events in the conference. I will also discuss the most prominent theories of consciousness explored in Tucson this past week (25 – 30 April 2016) in light of their implicit components. Finally, I will also elaborate on some of the strengths and weaknesses of these theories relative to Qualia Computing. We will be assessing these theories in light of today’s points, and making sense of the implicit background assumptions of their proponents. (More specifically, inquiring into: the conceptions of consciousness, the criteria for theories of consciousness, the knowledge bases, and the implicit background assumptions of the various attendees who participated in this event.)
Reblogged this on Quaerere Propter Vērum.
Hey, I am the author of a paradigm involving Supermathematics (lie Superalgebras) and artificial general intelligence.
I wonder if qualia is at all pertinent now, given the problem space involved in developing agi.
In Yoshua Bengio’s recent paper, he describes that he particularly avoids “illusive” descriptions of consciousness, precluding “qualia” etc, but instead he focuses on the definition in the realm of performing complex goals.
What are your thoughts on Bengio’s paper: https://arxiv.org/abs/1709.08568
I think qualia is in fact extremelly important, and that defining consciousness in terms of performing complex goals is counter-productive, misleading, and ultimately missing the entire point. I would point out that we I am not a functionalist (although I was for a long time), and that Dennett style approaches to consciousness are perhaps as confused (if not more) than the somewhat more naïve approach to simply treat qualia at face value. That said, I do think that in order to understand the problem space properly it’s probably a good idea to have spent some time thinking in functionalist terms and only then transcend it by a deep understanding of the problems with that philosophy of mind. I don’t think many people ever make it to the other side, though.
I highly recommend reading this article my colleague wrote about the subject: https://qualiacomputing.com/2017/07/22/why-i-think-the-foundational-research-institute-should-rethink-its-approach/
Another question that should be asked is whether consciousness is necessary for Artificial General Intelligence.
Indeed. At least in the context of the Bay Area rationalist cluster, I have generally found that people think that consciousness, even if real, is not actually needed for super intelligence. What they argue is that it suffices to create a system that can explore large state-spaces of possible solutions in order to maximize (or satisfy) a given utility function. Eliezer Yudkowsky and Stuart Armstrong are not particularly worried about whether recursively self-improving intelligences will be conscious. What matters to them is whether we can align the outcomes of what the intelligence would do with our own human values.
At Qualia Computing we are open to both possibilities. Orthogonality thesis (https://wiki.lesswrong.com/wiki/Orthogonality_thesis) may be true even if intelligence requires consciousness, and as rationalists, we need sufficient evidence to assert that consciousness is indeed necessary. We do, in part for that reason, explore what the computational benefits of consciousness may be. The phenomenal binding problem and the unitary nature of consciousness are strong hints that consciousness’ computational role may be related to the quick unification and propagation of information by establishing a fully connected state. Our consciousness may be less unitary than they seem, but they still are unitary, and ultimately, the way our attention integrates pieces of information by interacting with the contents of the mind (saliency ranking) suggests that consciousness may be responsible for creating states of highly integrated information throughout the nervous system of an organism.
Whether one takes the features of one’s experience seriously (including the unitary nature), however, depends on the person. I do take seriously the features of my experience, and for that reason I write about what the implications of the structure of consciousness are. Others, however, may be too cautious and restrict their curiosity out of a sense of pessimism: How could introspection reveal the way introspection works? Well, the answer is, by bootstrapping its way out of ignorance.