Summary
Consciousness is something measurable and we are beginning to understand what it is. But there is still a lot left to learn and uncover. Theories like Integrated Information Theory (IIT) and the Free Energy Principle are promising starts, but still fall short of a complete picture.
To finally solve consciousness, we have to view consciousness as an engineering problem.
Main Ideas
We need accurate measures for consciousness, to make it a less slippery concept. Consciousness is something we have a very "intuitive" understanding about, but intuition is often not a good guide in scientific endeavours. The concept fo consciousness is tied to processing information about where our bodies should be. It's a controlled and controlling hallucination, held in check by the sensory stimuli the body provides. The central assumptions can be twofold.
Internal states should be changed by action. The brain doesn't want to update state representation, instead it's controlling internals to keep state within certain limits.
This is what feelings are, signals that make us change our internal states, to keep them within certain boundaries, boundaries which are necessary for survival. Outside perception on the other hand is hallucination reigned in by the outside stimuli. It's a two way ongoing communication, between our neuron's best guessing what is causing the stimuli, and then updating those guesses as new stimuli arise and by updating them also changing our perception of things.
There are a couple of interesting theories, worth checking out in more detail, like IIT - integrated information theory - and the "Free Energy Principle". In the detailed notes there are also a lot of connections with other books and links to further reading.
Overall a solid book, breaking new ground and making the edge of this exciting research more accessible than the papers. But it's still a long and far way, until having a definite understanding of consciousness could be claimed.
Detailed Notes
Prologue
Anaesthesia is the art of turning people into objects.
Without consciousness, existence doesn't matter.
Consciousness arises somehow from the combined activity of our neurons. The question is, how?
We go through phases of consciousness, it alters as we age.
The idea of a continuous "self" that exists through these changes might be wrong.
Eureka-like insight - a pleasant but usually inaccurate myth about how scientific understanding progresses.
Brains are embodied and embedded. They are chemical machines and have a body they are attached to and this is important for consciousness.
Conscious experience is about being a "beast machine", about how the brain controls and predicts the body.
Our conscious experiences of the world and the self are forms of brain-based prediction - 'controlled' hallucinations - that arise with, through, and because of our living bodies.
When life ends, consciousness will end too.
Part 1 - Level
Chapter 1 - The Real Problem
Starting with the problem of phenomenology - the philosophy of experiences and Thomas Nagels question: What is it like to be a bat?
Focus on experiences, instead of function. Alternatives:
Global Workspace Theory - mental content, when in the "workspace" of the cerebral cortex.
Higher Order Thought Theory - mental content is conscious if a though process uses it. Metacognition is a link.
Consciousness is defined as 'any kind of subjective experience whatsoever'.
Definitions, in science, evolve over time, as people understand things better.
The hard problem of consciousness: Why should physical process s give rise to a rich inner life at all?
Different positions on consciousness:
- Physicalism/Materialism: Consciousness arises from the physical world, both exist.
- Idealism: Consciousness gives rise to the physical world. Matter emerges from mind.
- Dualism: Different substances that somehow interact with each other.
- Functionalism: flavor of physicalism, belief that consciousness is substrate independent
- Panpsychism: consciousness as a fundamental property of the universe, stones are conscious, but to a lesser degree.
- Mysterianism: we won't find the answer to the hard problem, but it exists
Over time, mystery after mystery had yielded to the systematic application of reason and experiment.
Philosophical zombies, Humans that act just like we do, but don't have conscious experience. If possible, than there must exist something outside the workings of a brain to make consciousness happen.
But the more we learn about brains and how they work the more impossible it seems that we can create something that works this way. It might be impossible.
A science of Consciousness should be able to explain, predict and control conscious experiences.
It's about mapping brain workings to the qualities of an experience, what things in the brain make it so that we experience the color red the way we do?
It's about the search for NCCs - neural consciousness correlates. I.e. what neurons have to fire in what ways to produce a certain precept?
The better the theories of consciousness are the less of a problem the hard problem of consciousness will seem.
What counts as mysterious now may not always count as mysterious.
Consciousness as a problem can be divided into three: level, content and self.
A combination of different amounts of the three constitutes what we call consciousness. The three can exist somewhat independently from each other.
There are ad many different ways of being conscious as there are different conscious organisms.
Chapter 2 - Measuring Consciousness
Book Recommendation: Inventing Temperature - Hasok Chang
To explain consciousness, we need to be able to measure it first.
Consciousness seems to depend on how different parts of the brain speak to each other and not the brain as a whole.
Zapping the brain and then recording the response is a useful tool, the algorithmic complexity, in other words, how compressible the patterns of response are, directly correlate with levels of consciousness. This kind of data is know as perturbational complexity index. It's essentially measuring how random neurons in the brain fire.
Complexity of brain activity goes down when we are "less" conscious.
Book Recommendation: The Diving Bell and the Butterfly - Jean-Dominique Bauby
Consciousness is not tied to responsiveness.
Hallucinogens don't just change the contents of consciousness but also the level of consciousness. They increase it as measures by complexity indexes.
Complexity is not pure randomness. It exists at the border between order and chaos. These ideas are tackled a lot in Mitchell Waldrops book Complexity
All conscious experiences are both informative and integrated.
Conscious experiences reduce down the space of possible experiences to only one at a time.
Conscious experiences are therefore defined by all the things they are not.
Redness is redness because of all the things it isn't.
Integration means that conscious experiences are experienced as a whole, not splintered into colors and shapes and sounds and touches separately. But all at once, unified.
Integratedness and Information are at odds with each other. Perfect integration carries minimal information and no integration (randomness) maximizes information. Consciousness exists at the sweet spot, where complexity is highest.
This point is also made in Mitchell Waldrops Complexity - again.
Causal density and neural complexity are two measures for how "conscious" systems are, based on their closeness to the edge of chaos. I.e. how much information they process and how much connectivity they have.
IIT - integrated information theory of consciousness. This is also mentioned in Max Tegmark's - Life 3.0.
Chapter 3 - Phi
Consciousness is integrated information.
A system is conscious to the extent that its whole generates more information than it's parts.
IIT posits that information exists in a meaningful way, that small changes to brains like adding neurons that never fire would change the nature of all conscious experiences of a brain, and that everything that integrates information - which is literally everything - is to some, albeit very small degree, conscious. IIT however is not testable.
It's tempting to think that everything is a kind of hallucination.
Part II - Content
Chapter 4 - Perceiving from the Inside Out
What we see, hear, and feel is nothing more than the brain's 'best guess' of the causes of its sensory input.
Consciousness is a controlled illusion. Controlled and held in balance by the input the brain receives from it's neurons and by what it has experienced as likely in the past. Assumptions + stimuli merge in a controlled hallucination.
Book Recommendation: De Anima - Aristotle
Raw sensory data without prior concepts makes no sense. Noumenons - things in themselves - make no sense. Brains need to learn what they are first to perceive the sensory data in a useful way. That's what babies do. Our perception of reality is an inference over sensory stimuli. And we use real world stimuli to correct our interpretation and prediction of the real world data.
Perception happens through a continual process of prediction error minimisation.
This idea is known as predictive processing or predictive coding in the literature.
We don't perceive color from reality. Color is a property of the reflective properties of something + the light with which it is illuminated. Our brains subtract the surrounding illumination color to infer the "real" color of a surface. Something "white" looks white to us, whether or not it is under yellow or blue or red light.
The surface of a chair has a particular property, the way-in-which-it-reflects-light, that my brain keeps track of through its mechanisms of perception. Redness is the subjective, phenomenological aspect of this process.
When we agree about our hallucinations, that's what we call reality.
Sensory illusions show how this hallucinatory machinery in our brains works.
We never experience the world 'as it is'.
Chapter 5 - The Wizard of Odds
Brains do Bayesian inference on the sensory input to infer the most likely causes of the stimuli and get the best possible idea about the world "as it is" possible, given the data.
Science is a process of ongoing Bayesian inference updating.
Let's return to our imagined brain, quiet and dark inside its skull, trying to figure out what's out there in the world.
How things seem is a poor guide to how they actually are.
Actions alter perceptions. Perceptions help to do actions.
We don't perceive the world as it is, we perceive it as it is useful for us to do so.
In a way the brain minimizes prediction error by doing actions to alter perceptions in the right way.
Actions are self- fulfilling predictions about where the body is going to be.
Action and perception are both forms of brain-based prediction.
Our perceptual world alive with colours, shapes and sounds is nothing more and nothing less than our brain's best guess of the hidden causes of its colorless, shapeless, and soundless sensory inputs.
Chapter 6 - The Beholder's Share
Book Recommendation: The Age of Insight - Eric Kandel
The brain reaches out to the world from inside the skull.
Our brain makes the low level details of vision and other senses work better, when it is expecting the "right" thing. I.e. we are better at "seeing" words that we know than we are at "seeing" jumbled letters we don't know.
Everything we perceive is a hallucination, however of different degree than "actual" hallucinations, where predictions overpower sensory feedback completely.
Hallucination is a form of uncontrolled perception.
Perception of time is based on perception of change, it's a measure of how much things changed over time for sensory input and the measure of neural networks of that amount of change corresponds to how humans perceive time.
What we experience as causality is a perceptual inference.
This type of outward inference is known as Humean Projection.
Things seem to be and the property of seeming to be, both are perceptual best guesses of brains. Our generative models are not experienced as models, but we see the world with and through them.
The hard problem of consciousness is a non-problem, because conscious precepts aren't a thing "out there" at all. This is an illusion.
A chair has mind-independent existence. Chairness has not.
On this there is an absolutely awesome Vsauce video on the nature of chairs and if things exist.
Part III - Self
Chapter 7 - Delirium
But what is a 'self'? Is it the sort of thing that can be departed from and returned to?
Chapter 8 - Expect Yourself
The many and varied elements of selfhood are Bayesian best guesses, designed by evolution to keep you alive.
In the teleportation paradox both are the "real" person from before.
We think that the property of being you is a thing in itself instead of a bunch of related inferred perceptions. This is wrong.
Book Recommendation: Being No One - Thomas Metzinger
Being you is not as simple as it sounds.
We feel like we own our bodies. This is known as embodied selfhood.
We also feel like we're observing things from a first person point of view. This is known as the perspectival self. But who is there to do the perceiving - why is it "located" in the skull? Isn't that the idea, that this is wrong, that Sam Harris champions in his Waking Up Lectures? Agency or the volitional self is another part of self. Another is that of a story, a woven autobiography, a narrative self. Another is the social self when I think about how others perceive me.
Just as there is no "redness" there is no "actual self".
The experience of being me, or of being you, is a perception itself — or better, a collection of perceptions — a tightly woven bundle of neurally encoded predictions geared towards keeping your body alive.
Experiences of what is my body and where am I are different but tightly coupled.
Hypnotizability is determined by the strength of the top down generative models that drive the predictive inferences. If those are stronger, people are more hypnotizable.
Book Recommendation: Forever Today - Deborah Wearing
Social perception, theory of mind, can be automatic and direct. Guessing what other people might think comes effortless, like seeing colors. Though it is often less accurate.
When we speak we are trying to insert thoughts into another's mind.
The experience of being a self seems to be an enduring centre within a changing world.
Our selfs change all the time, but we are perceptually blind to that change. This is the reason why we experience the self as an immaterial, unchangeable thing, it's the origin of the concept of a soul.
We do not perceive ourselves in order to know ourselves, we perceive ourselves in order to control ourselves.
Chapter 9 - Being a Beast Machine
We do not see things as they are, we see them as we are. — Anaïs Nin
Descartes Wound - Dualism, the divide between res cogitans and res extensa. Mind and Matter.
Our conscious experiences of the world around us, and of ourselves within it, happen with, through and because of our living bodies.
Bodies are a central ingredient for consciousness. Our consciousness can only be understood through that lens.
Interoception - body perception Proprioception - perception of body position within environment Exteroception - environment perception
Insular cortex process visceral stimuli. Visceral is another word for inside the body. Those stimuli are a signal of "how well the brain is taking care of its body".
Emotions are proprioceptively motivated.
We don't cry because we are sad, we are sad because we perceive our bodily state in the condition of crying.
Emotions are "embodied".
But we can still interpret the embodied state and change similar body states into more than one emotion. Idea of misinterpreting bodily state of arousal - because of a scary situation - with that of sexual arousal - has been shown in experiments.
Emotional experiences depend on how physiological changes are evaluated by higher-level cognitive processes.
Interoception is still only the best guess at what's going on in the body. Hence interoceptive inference is what the brain does. Best guessing the state of the body, then correcting by the actual stimuli gotten.
Paper: Every Good Regulator of A System Must Be A Model of That System
Perception is not for figuring out what's there, it's for control and regulation.
Interoception is active inference. We perceive to act and act to survive. Hence interoception - emotions - feel so different from exteroception - like hearing or vision.
Perception is structured by affordances - opportunities for actions. We see and comprehend reasons within the world.
We behave so that we end up perceiving things in a particular way.
Brains exist to help organisms survive better. Controlling the variables that signal "alive" through actions as best as possible. The goal is allostasis. Stasis through change.
All of our perceptions and experiences, whether of the self or of the world, all are inside-out controlled and controlling hallucinations, that are rooted in the flesh-and-blood predictive machinery that evolved, develops and operates from moment to moment always in light of a fundamental biological drive to stay alive. We are conscious beast machines through and through.
For as long as we live, the brain will never update its prior belief of expecting to be alive.
Maybe it sometimes does? When people "want" to die because they are old and unhappy and sad and they soon die, maybe that's exactly what happens? Their brains stop "caring", stop controlling to keep that prior true?
We see us as stable over time, because we need a strong prior to misperceive actual changes ass not that drastic, so that action is more important than updating our perceptions.
The self as such does not exist, just like "redness" as such does not exist in reality. The self is instead a quasi permanent quality arising from our interoceptive inferences, that has to be stable, because we use it as a best guess to keep alive. This constant sense of being alive, is also what gives rise to the idea of the soul. Ātman. The breath of life.
We are not cognitive computers, we are feeling machines.
Chapter 10 - A Fish in Water
Post Mention - Slatestarcodex - "God Help us Let's try to understand Friston on free energy"
Existence of a thing means a over time persistent boundary between that thing and everything else. Living things maintain their boundaries actively. To maintain those, they expect to be in certain states, otherwise they wouldn't exist. Certain states are more likely than others, and those states are what living things want to keep experiencing - their perceived entropy - also called sensory entropy is kept low by the actions of the organism.
Free energy is the sensory prediction error. It allows a living thing to measure how close it is to states it likes to be in. How good it is at surviving.
Living systems have - or are - models of their environment.
Living things model the environment and their actions within that environment such that their actions leaf to revisiting those states a lot over time. Which means that their boundaries, their preferred states of being keep existing, that they stay alive.
In order for organisms to stay alive they need to behave so as to maintain themselves in the (low entropy) states they 'expect' to be in.
The key word here is maintain, the states in the future are important, hence temporary novelty in sensory states, higher sensory entropy - exploration and curiosity - are better for the long term sum of sensory entropy. And this is what it means to exist, I expend energy now to find food, so that I exist longer.
Chapter 11 - Degrees of Freedom
It seems like our selves can decide on actions, like we can decide to do something and then do exactly that because we decided to do so.
Benjamin Libets Volition Experiments
Voluntary actions are voluntary because at that moments they are what we want to do. However we couldn't have chosen to want differently, because we would have had to be a different person in that moment to want differently.
Man can do what he wills, but he cannot will what he wills. — Arthur Schopenhauer
Volition has 3 defining Features:
- We are doing what we want to do.
- We could have done otherwise.
- Actions seem to come from within.
Volition is the conscious manipulation of degrees of freedom within our brains. What, when and whether is handled by different, interlocking processes.
Free will is important for learning, we could have decided differently is a good way for the brain to learn for the next time.
No man ever steps in the same river twice, for it's not the same river and he's not the same man. — Heraclitus
Book Recommendation: The Illusion of Conscious Will - Daniel Wegner
The experience of free will is not an illusion, even though the decisions are not free - we would decide the same over and over again if placed in the exact circumstances - we have an ability to change our brains to exercise degrees of freedom, influencing our volitions.
We project causal power into our experiences of volition just the same way that we project redness jnto our perceptions of surfaces.
Part IV - Other
Chapter 12 - Beyond Human
Animals have consciousness, but figuring out how similar their consciousness is to that of humans is a difficult task. It might be very very different. And intelligence is not required for consciousness but is required for certain conscious experiences.
All mammals are conscious. They share a lot of functionality and brain circuits with us. But their conscious content and experience of self is very different from ours, simply because they inhabit different environments. The same is true for other animals, even more different from us.
Octopus bodies can change their color locally without their "brains" getting involved at all. What that means for what it feels like to be an octopus is a very interesting question.
Octopuses might not have a distinct conscious representation of their bodies, because their arms can move largely independent.
There might even be something it is like to be an octopus arm.
Anaesthesia drugs seem to have a pain killing effect on all animals.
We inhabit a tiny region in a vast space of possible conscious minds.
Consciousness can be severed from intelligence, one might exist without the other.
Chapter 13 - Machine Minds
Functionalism - the idea that consciousness is substrate independent and that consciousness is solely input/output based. Have the right system, processing information in the right way - ta-da - you have consciousness. This view might or might not be true. We simply don't know yet. Max Tegmark describes this in much more detail in his book Life 3.0.
Just making computers smarter is not going to make them sentient.
A mechanized beast machine, according to Anil Seth wouldn't be conscious.
Even though the same circuitry that we have is there, and it has a model of itself and the environment and generative, controlled and controlling illusions about the world and it's body, it nonetheless wouldn't be conscious? Because it lacks cells that "all the way down" "care" about being alive? To me this makes no sense, here he spend all this time building up a nice theory just to pull it to the ground by saying that the thing that really makes us conscious is our existence as bodies made up of cells. That "all of us" cares about survival.
The Gartland test: Can humans be fooled into feeling that an AI is conscious even if they know that it isn't?
Could it be possible to torture a robot while feeling that it is conscious and simultaneously knowing that it is not, without one's mind fracturing?
Developing conscious machines, or things in general is something that might go horribly wrong. If something can be conscious and suffer in ways we don't understand, we could create a lot of suffering unnecessarily.
Interesting Science: Cerebral Organoids - prices of stem cells grown into brain like structures.
The idea of machine consciousness and mind uploading is due to humans wanting to become immortal. We want to transcend our fleshy origins and become "perfect", hence flee into the realms of machine consciousness because that holds the door open.
Anil Seth's "Human Beast Theory" is contradictory to that. Consciousness arises because we are living agents that want to persist. We are firmly part of nature, not apart from it.
Epilogue
Anil Seth describes watching a brain surgery:
I was left in awe of the material reality of this most magical of objects. It felt almost transgressive. A curtain had been pulled back revealing something too intimate to be so openly on view. I was looking directly into the mechanics of a human self.
Every conscious experience is simultaneously unified and distinct from all other conscious experiences.
Different experiences of internal and external perceptions can be explained because of the different predictions they try to make and that one is controlled by actions, while the other updates the internal models, by finding out more about the world. Both are tied to our existence as biological machines that need to survive and are hence shaped by evolution, both try to minimize prediction error of the assumption of "alive" vs. "dead". This part of consciousness exists in all conscious animals. And places us further away from machines than we often assume.
Everything in conscious experience is a perception of sorts, and every perception. Is a kind of controlled - or controlling - hallucination.
We will see how our inner universe is part of, and not apart from, the rest of nature.