A Cognitive Toolkit for the Rationalist. 6/6: Matter Over Mind

Chapter 13: The problem of consciousness

Consciousness as an emergent property of the brain

Rene Descartes was a genius who gave us Cartesian geometry, used to this day. He also said the famous words “cogito ergo sum” — I think therefore I am. Without getting into the details of the argument, what it gives us is a picture of mind as something separate from matter. This gives rise to questions like -

·         Can consciousness exist without the body?

·         Can consciousness be transferred from one source to another?

·         Can consciousness survive death?

We can start with two diametrically opposed (but at this point equally valid) worldviews — the first one says that consciousness is fundamental and the material world is emergent from it (“idealism”), while the second describes consciousness as an emergent property of (living) matter.

https://www.scoopwhoop.com/video-of-freshly-removed-human-brain/

Hindu philosophical thought contains very advanced enquiries into the nature of consciousness. The Atman is the individual soul (or consciousness) and the Brahman is the universal consciousness. There is a school of thought Advaita which says that both are the same and it is Maya which divides them. But all the Hindu systems emphasize direct experience, rather than observation, as the way to know the nature of consciousness.

What about the materialist view? At this point in time, neuroscience is not even close to giving us a credible mechanism by which consciousness could emerge from matter. But scientists are working on it — Anil Seth gives a fascinating account of the latest advances. The problem then is whether we are prepared to accept that consciousness could have a materialistic explanation that is complete.

Dan Dennett outlines several obstacles to such an acceptance, like the hard problem of consciousness — is it possible to completely describe someone’s subjective mental state by objective analysis? A lot of studies show how we cannot have authoritative knowledge of the working of our own minds. Our brain is constantly playing tricks on us, as revealed by a number of illusions that have been systematically designed by researchers. So it would appear that the third person (objective) account of our mind is more reliable than our first person (subjective) impressions. This is reinforced by the realization that our brain tracks information on a strictly “need to know” basis. We know that we have a liver, a pancreas, a pair of kidneys etc. But we know this from third person observations and not from first person experience.

The debate on whether consciousness is fundamental or emergent is unlikely to be resolved anytime soon. So again, we can apply our heuristic rules from Chapter 8 to decide what kind of theory we like when it comes to understanding consciousness. Starting with consciousness as a fundamental property of the Universe (like spacetime or mass-energy) seems like a “top down” approach, leading in its most extreme form to panpsychism. It also does not make any specific, testable predictions that can be validated through experiments. And finally, the examples from Chapter 7, of phenomena assumed in the past to be fundamental but which are now better understood as emergent, gives us some inkling that the same may turn out to be true of consciousness.

The kind of theory we prefer is bottom up and it is empirical, i.e. provides a way to validate or falsify its predictions. We start with a model where the brain is the seat of consciousness and not a conduit for consciousness. We know that our brain has almost 100 billion neurons and that these neurons are connected. It is fed a constant stream of rich data to process, by the sense organs. Drawing an analogy to the ant colony from Chapter 7, we can at least contemplate how billions of neuron interactions could produce self-awareness even when none of the individual neurons have any. Individual ants don’t have to possess intelligence (or goals) for their interactions to produce intelligent (or goal-directed) behaviour. Individual neurons don’t have to be self-aware for their interactions to produce self-awareness.

The analogy between brains and ant colonies appears in Douglas Hofstadter’s 1979 classic Gödel, Escher, Bach. Hofstadter refers to “Strange Loops” as the crux of consciousness — an interaction wherein the top level of a system is built on lower levels but is able to influence the bottom level, and thereby itself. It is related to the concept of recursion which any computer programmer would be familiar with — think of two plane mirrors facing each other or a video camera pointed at a screen to which its output is connected.

A lot of this is speculation though, and we must acknowledge (once again) that neuroscience is not likely to provide a precise description of the mechanism anytime soon. But assuming that a complete and credible materialistic theory of consciousness will emerge at some point in the future, the answers to the three questions posed at the beginning of this Chapter would be, according to that theory No, No and No.

Now let’s move on to the question of why, according to Neo-Darwinism, did we evolve consciousness? Why as in “how come”? It is obvious that any complex organism needs to be able to differentiate its own body parts from its surroundings. A lobster can’t afford to claw itself. So self-awareness of the body would be an essential brain function. But what about self-monitoring of the mind? Why would that be useful or necessary?

The answer may be tied to another uniquely human adaptation — language. Language evolved because of the advantage it afforded to the individual. Recall from Chapter 6 that natural selection acts on individuals (actually genes) and not on groups or species. The use of language is as much for deception as it is for “true” communication of beliefs and intentions. An individual who indiscriminately communicates every thought to their fellow humans is unlikely to survive for long. This makes it not just useful but essential for the brain to have a self-monitoring ability. My thoughts, memories, beliefs, expectations, intentions must be tracked and represented separately. This would explain the feeling of “self” — what’s it like to be me? That in fact is Nagle’s definition of consciousness.

Does this explanation, due to the fact that it relates consciousness with language, imply that only human beings are conscious? Yes, but there is a different explanation which does not rest on language as a prerequisite. It starts by describing the brain as a prediction engine which uses sensory inputs to build a “model” of the external world (more on this in Chapter 14). In order to make this model as accurate as possible, the brain must include itself as part of the model. Instead of assuming an external “perceiver” or “experiencer” (aka soul) we may define consciousness as the brain’s high-level representation of itself. And consciousness, like the brain, may itself be an adaptation, as we shall argue in the next chapter.

Chapter 14: Artificial intelligence

Is it possible to have a sentient AI?

What is AI? AI is simply the ability of machines to do tasks that were so far assumed to require human intelligence. Self-driving cars are a popular example.

One of the applications of AI is Machine Learning (ML). Being a data scientist, here at last is a subject on which I can claim some expertise :-). Machine learning is a set of algorithms that can learn to perform a wide range of tasks without being explicitly programmed. Here “learning” refers to constant improvement by analysing more and more data; i.e. encountering more “cases”. So, while an elevator control system automates a task once performed by humans, it does not get better at it by analysing patterns. In other words, it lacks the ability to learn.

An ML algorithm which is trained to recognize images of cats will initially need to be fed many images of cats (labelled as “cat”) and other images (labelled “not cat”). Once it is “trained” it can start to identify cat pictures accurately. Replace cat with “malignant tissue” for a more useful application of ML, namely image-classification, also used by self-driving cars.

One of the most powerful ML algorithms (class of algorithms actually) is the neural network, which as the name suggests, learns in a way similar to the human brain. Our brains use Bayesian Inference to interpret data coming in from the sense organs (sight, sound, smell….). It has a “prior” expectation which it updates based on incoming data. What you see (or hear or feel…) is your brain’s best guess based on both the prior belief and the sense data. Sometimes the prior belief is so strong that it overrides the incoming information. Cognitive scientists like to demonstrate this through visual and auditory illusions.

We are now ready to try and answer the question we posed at the end of Chapter 10 — why can’t AI be considered minds separate from bodies? Or can an AI ever attain consciousness?

The analogy of the human brain as a necktop computer is useful in certain respects but misleading in some ways. The fact that computers are made of silicon while the brain is made of organic (“wet”) stuff is not important here. But they are fundamentally different for a different set of reasons —

·         The human brain has a bottom up organization, there is no centralized control. Every neuron is fighting for survival by making itself useful much as each worker survives in a market economy by finding jobs to do. But an individual transistor in a computer will not find itself “unplugged” if it stays idle for a period of time so it doesn’t need to actively seek out tasks to do.

·         The human brain has evolved in a partly hostile, partly cooperative environment where it has constantly needed to make decisions in life-or-death situations, competing and cooperating with fellow humans (as we argued in Chapter 12). Computers, on the other hand, live in a relatively sterile environment.

·         The emergence of consciousness in the human species is because we are survival machines (see Chapter 6) and not despite that fact. In other words, consciousness is not an attribute which makes us something “more than” survival machines; it is something that makes us better survival machines.

If an AI is ever to become a “conscious” mind unattached to a body it would need to have

·         Bottom up organization comprised of autonomous elements

·         A need to survive in a competitive environment with possibilities for cooperation. A need to practice deception, which requires self-reflexive thought. The tagline for the 2015 sci-fi movie Ex Machina sums this up pretty well — “There is nothing more human than the will to survive”.

Chapter 15: The brief history of mind

A story should have a beginning, a middle, and an end... but not necessarily in that order

Recall from Chapter 8 that the heliocentric model of our solar system took more than 1000 years to be accepted from the time it was first proposed by Aristarchus of Samos in the 3rd century BCE. It took further work by Copernicus and Kepler in the 16th Century to construct a mathematical model, and Galileo’s observations through the newly-invented telescope, to validate the model. Was the new theory greeted with enthusiasm and excitement? Of course not! The disgraceful treatment of Galileo by the Catholic Church should continue to remind us that new ideas which threaten to turn our worldview upside-down are likely to be met with strong opposition.

While some of the opposition may be genuine scientific skepticism, a lot of it is simply blind rejection of new thinking that is seen to threaten faith beliefs. In the case of the heliocentric model, there was a lot at stake - one of the pillars of the anthropocentric worldview (see Chapter 9) was the belief that the Earth enjoys a central, supreme position in the Universe. It was heresy to suggest otherwise.

We are right now facing a crisis of even greater proportions with respect to our worldview, and it is this -

·                     The question of Origins has had many different answers offered by religious and spiritual traditions over the centuries (Chapter 8 contains a link to a list of “creation myths”). But at a fundamental level they are all parallel narratives which, translated to modern language, go somewhat like this - first there was a mind, then came matter followed by lifeforms.

·                     With the latest advances in science we have reached a point where we can say quite confidently that the above sequence is wrong. It was matter that came first, followed by lifeforms emerging from (inanimate) matter and finally, minds emerging from lifeforms.

What do we mean by “minds”? A mind is something that has a capacity for any of the following - rationality, goals, beliefs, intentions, motives, reasons and purposes. The regular English meanings of these words will serve us perfectly well here. What we are essentially saying is that none of these played any role in shaping our world until very recently, that is until the appearance of Homo Sapiens. In other words, the 4.5-billion-year history of the Earth (and the 10 billion year prior history of the Universe) is best understood in terms of blind, mechanistic, purposeless processes. Neither the equations of physics nor the algorithmic sequence described by Evolution requires any assumption of a rational agent working behind the scenes - one that has goals, purposes, intentions or beliefs.

But most spiritual belief systems rest on a contrary set of assumptions -

1.           Minds pervade vast regions of space rather than being localised in the place between our ears

2.           Minds have existed since the beginning of time rather than having evolved relatively recently

https://www.pmfias.com/origin-evolution-life-earth-biological-evolution/

This picture provides a clearer perspective on what we mean by “relatively recently” - the history of minds does not account for more than 0.01 billion (or 0.2 percent) of the 5-billion-year history of our planet.

But so what if the assumption of a primordial mind is not required, can we not retain it still? The general heuristic of Occam’s Razor would eliminate any redundant assumptions, i.e. those that do not add explanatory power to our model of reality. But we will go further and invoke another heuristic, already introduced in Chapter 10 - the Intentional Stance from Dan Dennett.

Dennett describes three levels of analysis for modelling (or predicting) the behaviour of things of varying complexity. The Physical Stance describes a thing (or system) in terms of its physical components and their interactions, making predictions based only on an understanding of these. It is, in principle, the most accurate approach and the one used in Physics and Chemistry. For instance, the behaviour of a stone can be accurately predicted based on the fact that it is held together by electrostatic forces and is subject to the gravitational force. But if we replace the stone with a live bird, the Physical Stance would not continue to serve us well. Though it is true that a bird, like any other object, is made up of atoms and molecules and is subject to the laws of physics, we need a higher level of abstraction to predict, that when released it will fly up and not fall down like a stone. This second level of analysis is the Design Stance which assumes, in this case, that birds are “designed for flying”. Never mind that it is design without a designer (see Chapter 6). The higher complexity of a living thing (bird) compared to an inanimate object (stone) warrants the adoption of the Design Stance. It is what we implicitly use in Biology and Engineering.

When it comes to predicting the behaviour of humans, even the Design Stance will generally not suffice. There isn’t anything in particular that a person is “designed” to do. Of course, the same goes for birds but most people would agree that a bird’s behaviour patterns are less complex, and therefore easier to predict, than those of a person. The third and highest level of abstraction then, is the Intentional Stance.

Here is how it works: first you decide to treat the object whose behaviour is to be predicted as a rational agent; then you figure out what beliefs that agent ought to have, given its place in the world and its purpose. Then you figure out what desires it ought to have, on the same considerations, and finally you predict that this rational agent will act to further its goals in the light of its beliefs. A little practical reasoning from the chosen set of beliefs and desires will in most instances yield a decision about what the agent ought to do; that is what you predict the agent will do.

Having defined the three levels of analysis, here’s the main insight. Jumping to the next (higher) level involves sacrificing accuracy and reliability in the interest of “zooming out” irrelevant details.

Consider rivers. A river can be described using the Physical Stance as water flowing downhill. This approach affords us a reasonably good understanding of not only its “normal” state but also flooding or drying up. But skipping to the next level, we would need to assume that rivers were designed “for” something - perhaps to provide water which sustains life (see Chapter 9 for more examples of this type of reasoning). But that wouldn’t explain why it floods or dries up. To explain that, we would need to attribute desires, goals and intentions (in short, minds) to the river. What we end up with is a River Deity whose “actions” we may be able to influence by appealing to its mind through prayers and rituals. And since the only minds we know are human minds, we inevitably converge on an anthropomorphic representation, usually female in the case of rivers.

In Chapter 10 we had encountered a similar scenario relating to lightning and thunder and also formed a hypothesis around why the Intentional Stance would be an evolutionarily “safe” strategy. It seems that the Intentional Stance is more likely to be applied to things or systems which are both -

·                     Either essential for or dangerous to our survival, and

·                     Beyond our control by physical means

Coming back to the question we set out to answer, we already have a reasonably accurate description of the processes that got us here (Big Bang, Relativity, Quantum Mechanics, Evolution…), and these descriptions are all based on the Physical Stance. In the past we had neither the methods nor the technology to be able to come up with models that describe Nature from the bottom up (see Chapter 7). Of course, there will always be gaps in our knowledge. We agreed in Chapter 3 that every scientific theory is tentative and susceptible to falsification. But because of this, in terms of agreeing with empirical data it easily outperforms anything we had before. So it doesn’t seem like there is anything to be gained by switching to what is essentially a more “top-down” approach; i.e. the Design or Intentional Stance.

The Intentional Stance applied to Nature has not, over the years, given us a reliable understanding of any aspect of it. But shedding it would have a domino effect on many “meta-beliefs” of the kind we went over in Chapter 12 -

·                     The idea that everything has a cause or happens for a reason

·                     The idea that Nature (or the Universe) has goals and purposes

·                     The idea that morality has any influence beyond the domain of human societies

·                     The idea that introspection can reveal “truths” about the outside world

Obviously, a great deal is at stake for some of us. Let’s look at one of the popular arguments in defence of the primordial mind - the one which starts with the doctrine of consciousness as a fundamental property of the Universe (see Chapter 13). It has been used to construct narratives that are, in essence creationism disguised in sciency-sounding language.

One of its unwitting allies is Quantum Mechanics (QM); in particular, the concept of “observer effect” based on the famous double-slit experiment. The concept of observation (better understood as “measurement”) in QM is a subtle one – it is any interaction between the quantum system and its environment which leaves a record in the environment. The observer can be a device (like a camera or screen) or even another particle; it does not have to be a “conscious” entity. More on this age-old debate here and here.

To get science across to the layperson, scientists must provide interpretations in plain English. This is hard enough to do in the case of QM, which is highly mathematical and difficult to explain in terms of anything familiar (in fact it’s downright weird). To make things worse, we have plenty of ideologues waiting to take advantage of the fact that words in English (or any human language) don’t have meanings that are unambiguous and context-free, which allows them to twist meanings of words to support whatever it is they want to believe and want others to believe. It seems like a Catch-22 - be misinterpreted or be ignored! Perhaps we need fewer scientists and more science popularizers.

Written by Ambar Nag.

ambarnag@gmail.com

(Concluded)