Consciousness is "deeply inscribed into the wider patterns of nature."
This book is a good non-technical synopsis on the cutting edge of consciousness research. However, this book offers little new insight. Seth manages to represent all sides of the argument fairly without giving up on his adherence to physicalism. Through this book, it is apparent that Seth is a proponent of Embodied Cognition, the idea that our bodies—not just our brains as passive information processors—play a crucial role in forming our conscious experiences.
Remark: When talking about consciousness, adherence to physicalism is necessary. Information is fundamentally physical (refer to papers by R Landauer and D Deutsch). There is nothing magical about information, emotions, intuitions and consciousness. They all obey the laws of physics. DHCA is hard evidence against the existence of an immaterial soul that leaves the body after death:
Deep hypothermic circulatory arrest (DHCA) is a surgical technique that induces deep medical hypothermia. It involves cooling the body to temperatures between 20 °C to 25 °C, and stopping blood circulation and brain function for up to one hour. It is used when blood circulation to the brain must be stopped because of delicate surgery within the brain, or because of surgery on large blood vessels that lead to or from the brain. DHCA is used to provide a better visual field during surgery due to the cessation of blood flow. DHCA is a form of carefully managed clinical death in which heartbeat and all brain activity cease."
This book might be frustrating to some readers because Seth is agnostic about several contentious open problems in the consciousness space, such as whether consciousness is substrate independent. Seth is prudent and reverent. I have seen many consciousness researchers treating the more outlandish hypotheses with ridicule, but he mostly remains agnostic as long as the jury is still out on the matter. Knowledge creation requires idea generation, not just idea judgement. Therefore, we should be more tolerant of the more outlandish ideas as long as there are no knock-down arguments or evidence against them.
This book covers the whole gamut of consciousness research. Seth covers topics ranging over:
- Measuring consciousness
- Integrated Information Theory
- Free Energy Principle and Active Inference
- Good Regulator Theorem
- Embodied Cognition
- Non-human consciousness (animal and machine consciousness).
Seth addresses the distinction between beliefs and reality early in the book. I was happy to see this because confusion about this distinction can lead to an infinite regress, where clever people get lost in their own beliefs about their own beliefs ad infinitum and forget that reality is the ultimate judge of the accuracy of our beliefs.
Can you imagine an A380 flying backwards? Of course you can. Just imagine a large plane in the air, moving backwards. Is such a scenario really conceivable? Well, the more you know about aerodynamics and aeronautical engineering, the less conceivable it becomes. It just cannot be done.
Seth follows this up with a knock-down argument against Chalmers' philosophical zombies:
In one sense it's trivial to imagine a philosophical zombie. I just picture a version of myself wandering around without having any conscious experiences. But can I really conceive this? What I'm being asked to do, really, is to consider the capabilities and limitations of a vast network of many billions of neurons and gazillions of synapses (the connections between neurons), not to mention glial cells and neurotransmitter gradients and other such neurobiological goodies, all wrapped into a body interacting with a world which includes other brains in other bodies. Can I do this? Can anyone do this? I doubt it.
Scott Aaronson has argued in his seminal paper Why Philosophers Should Care About Computational Complexity that philosophers should seriously consider the limits imposed by computation and the laws of physics. Physics constrains everything else in a way that everything else does not constrain physics. We cannot choose to change the laws of physics. The tendency to take armchair intuitions too seriously and privilege them above reality is a relic from the unavailing philosophies of Hegel and his ilk. This quote from Roger Penrose captures the relationship between physics and consciousness nicely:
We have a closed circle of consistency here: the laws of physics produce complex systems, and these complex systems lead to consciousness, which then produces mathematics, which can then encode in a succinct and inspiring way the very underlying laws of physics that gave rise to it.
Trying to explain consciousness via introspection is like trying to explain how Google works by doing Google searches. As Dennett puts it, we are the end-user of consciousness. Therefore, what we experience is a highly abstract user interface. In Every Thing Must Go, James Ladyman and Don Ross stake a strong claim about the uselessness of "esoteric debates based on prioritising armchair intuitions about the nature of the universe over scientific discoveries." Armchair intuitions about reality are unlikely to have predicted any of the results of quantum physics, such as the counterintuitive Bell inequality.
That being said, Seth does present the reader with the good old cloning thought experiment. Seth asks the reader to imagine a machine capable of creating a perfect clone of someone and then asks if the clone is the same person. He completely ignores the no-cloning theorem in quantum mechanics. I don't understand the point of these "thought experiments" when physics has ruled them out. Seth criticises Chalmers' philosophical zombies but then makes the same mistake.
Remark: Seth mentions the word 'emergence' a total of only 6 times in the entire main text, which I thought was impressive for a book on consciousness. Calling something we do not yet understand 'emergent' doesn't help us understand it any better than saying nothing.
Seth points out that to measure something, we must first find a fixed point, an unchanging reference against which we can measure it. We achieved this using absolute zero for temperature.
Seth refers to Tononi's Integrated Information Theory (IIT) as a candidate for a measure of consciousness, using the quantity Phi, symbolically . Phi is poorly defined, and its definition keeps changing. Vagueness and change are characteristic of poor theories. A good theory should be rigid and precise in its assertions about reality, making itself vulnerable to a critical test. IIT offers little explanatory power, and I was disappointed to see Seth ignore Scott Aaronson's incisive knock-down of IIT.
Aaronson asserts that IIT has already been falsified. It has failed to explain the very things it purports to explain. According to IIT, a large grid of XOR gates and error-correcting codes are conscious. When Aaronson pointed this out to Tonini, Tononi "didn't only bite the bullet, he devoured it." Tononi's response to Aaronson was something like: "Yes, a large grid of XOR gates and error-correcting codes are conscious, and your intuition about consciousness is wrong and needs to change." Tononi is retrofitting consciousness to his theory instead of providing a good explanation of consciousness. This is the equivalent of a theory of temperature according to which lava doesn't burn, and when you point this out, you get told: "Your intuition about what burning feels like is wrong."
The next part of the book blew my mind. I think I found an answer to the question, "What is your biggest fear?" It is called locked-in syndrome. This is a freaky condition, and I cannot begin to imagine what such an experience would be like. It is terrifying.
Locked-in syndrome [is] where consciousness is fully present despite total paralysis of the body. This rare affliction can follow damage to the brainstem, a region at the base of the brain (and at the top of the spinal cord) which, among other roles, mediates control of muscles in the body and in the face.
Some locked-in syndrome patients maintain the ability to make limited eye movement, offering a narrow channel for communication (the book The Diving Bell and the Butterfly was written this way!) and diagnosis, while others are entirely locked-in. It is estimated that one to two thousand undiagnosed locked-in patients languish forgotten in nursing homes and hospital wards worldwide.
I am going to treat these complicated topics with very broad strokes for the sake of keeping this review easy to read. Seth gives an excellent overview of the Free Energy Principle (FEP), Active Inference, Bayesian Inference and the Good Regulator Theorem, which I view as being closely related concepts.
I am going to skip giving an overview of Bayesian inference to a rationality community. There are plenty of excellent posts about Bayesian inference on LessWrong. Why is Bayesian inference relevant to life and cognition? Karl Friston thinks that all the processes of living systems boil down to solving inference problems to minimise predictive error and maximise evidence about the organism's own existence.
Let's take a top-down approach. Let's start from the top and ask this elementary question: If a system exists and actively maintains its own existence for periods of time, then what must it be doing? Before we tackle this question, let's take a step back and ask a more concrete question. Suppose you have an air conditioning system, and its goal is to maintain the temperature of a given building. The Good Regulator Theorem asserts that "every good regulator of a system must be a model of that system." Therefore, the air conditioning system must have some heat-map (e.g. via thermostats) of the building (i.e. a model). Similarly, for an organism to maintain its existence, it must have a model of the system it is trying to sustain, i.e. a model of itself and its surrounding environment. This way, the organism can remain within a narrow set of favourable physical states (the organism's attractor set), which allow it to stay alive.
Karl Friston is widely considered to be the most influential and cited cognitive scientist alive. Friston is a physicist at heart, and therefore, his theories about biology and cognition are grounded in physics. Friston originally came up with the inspiration for FEP when he was eight years old while watching woodlice in a garden. He observed that woodlice moved faster while exposed to sunlight and moved slower under shade. He then hypothesised that the reason for this is literally that in the sunlight, the lice have more energy! This insight is very obvious but it is nevertheless very insightful. This inspired Friston to pursue a grounds-up physics-based explanation for life.
FEP is remarkably elegant. In very broad terms, FEP says that all the processes of living organisms are solving inference problems to make decisions to move uphill on the probability distribution for the evidence of the organism's existence. In other words, organisms are attracted (both in the literal sense and in the sense of a dynamical attractor) to resolving uncertainty about their own existence. FEP is not falsifiable, and that is ok. It is more like a law of physics. It is a deduction from the fact that living systems exist and persist over time. We can ask questions like, "Does this system conform to FEP?" It nevertheless provides immense explanatory power, and theories based on it will provide us with falsifiable predictions.
Bayesian optimality becomes computationally intractable as you throw large volumes of data at it. An organism must model its environment efficiently. This can be achieved by minimising variational free energy. Minimising variational free energy provides an algorithm agnostic and computationally tractable way of approximating Bayesian optimality. Richard Feynman introduced variational free energy in 1972 to convert an intractable integration problem into a tractable optimisation problem while working on quantum electrodynamics.
Remark: Friston likes to joke and say maximising Bayesian Model Evidence is the solution to all problems. This is so true. Whatever your goal, you have a better chance of achieving it by maximising the evidence showing that you’re achieving it (easier said than done).
FEP can be generalised to action through Active Inference. According to Active Inference, actions are a form of self-fulfilling perceptual prediction. When you go around searching for something, you are moving uphill on the probability distribution of finding what you are looking for. We use actions to manipulate our surrounding environments to maximise the evidence for our existence. Thinking about action in this way underlines how action and perception are two sides of the same coin. Rather than perception being the input and action being the output, with the brain being a passive information processor, action and perception work together to solve inference problems.
Seth is interested in marrying together the Free Energy Principle and Integrated Information Theory, but he thinks we are far from achieving such a feat. "FEP starts from the simple statement that 'things exist' and derives from this the whole of neuroscience and biology, but not consciousness. IIT starts from the simple statement 'consciousness exists' and launches a direct assault on the hard problem. It's not surprising that they often talk past each other."
Emotions are not magic. Again and again, I come across people who think emotions somehow reveal the truth of the universe to them. Everything Seth says about emotions (like most of the rest of the book) is a well-articulated summary of other people's ideas. His treatment of emotions is based chiefly on the works of Antonio Damasio and Lisa Feldman Barrett, who have put emotions on solid grounds.
Confusion about emotions arises because what science tells us about how emotions work appears counterintuitive with respect to what emotions feel like from the inside. For example, our intuition about emotions tells us that we cry because we are sad, but science tells us that "we are sad because we perceive our bodily state in the condition of crying." Science has, again and again, subverted our intuitive, how-things-seem notions about the world and ourselves. Our intuitions about emotions tell us that emotions cause bodily responses, but science asserts that the relationship is the other way around.
Emotions are closely linked with changes in bodily state. However, emotions are not as simple as just expressions of distinct bodily states. Appraisal theories of emotions claim that a context-based cognitive inference process takes place, such that the same bodily state can result in different emotions based on context. Barrett's research asserts that emotions are best-guesses that compromise accuracy for speed. Furthermore, according to Barrett, we can control, to some extent, what emotions we create. Given that emotions are inferences, they can readily be explained by FEP (but is there anything FEP doesn't purport to explain?).
The chapter on animal consciousness was exciting and insightful. This was one of the funniest things I've ever read:
Plagues of rodents, locusts, weevils, and other such smaller animals were less easy to deal with via legal proceedings. In one celebrated sixteenth-century case, the French lawyer Bartholomew Chassenée successfully exonerated some rats with the clever argument that they could not reasonably be expected to turn up to trial, given the dangers posed to them by the many cats lying in wait along the route. In other cases, including various weevil infestations, the offending animals were issued with written orders to leave a property or a barley crop, often on a specific day and even by a specific hour.
Seth talks about the mirror test, which I thought was insightful. The mirror test is an experiment in which animals are anaesthetised and marked with a dye on their bodies and then, upon waking up, are placed in front of a mirror. Upon seeing the mark on their bodies in the mirror, those animals that examine the mark on their own bodies pass the mirror test.
Who passes the mirror test? Among mammals, some great apes, a few dolphins and killer whales, and a single Eurasian elephant. A parade of other mammalian creatures, including pandas, dogs, and various monkeys, have failed—at least so far. Given how intuitive mirror self-recognition is for us humans, and how otherwise cognitively competent many of these non-self-recognising mammals seem to be, this pass list is remarkably short. There is no convincing evidence that any non-mammal passes the mirror test...
The rest of this chapter is dedicated to octopuses. Octopuses are of great interest to consciousness researchers. Other Minds: The Octopus, the Sea, and the Deep Origins of Consciousness is an entire book dedicated to exploring octopus consciousness. This is because octopuses and mammals shared a common ancestor a long evolutionary time ago. It is hypothesised that we shared a flatworm that trawled the seafloor 750 million years ago as a common ancestor. Therefore, octopus consciousness is the most alien kind of nontrivial consciousness we can find on this planet. Octopuses use a remarkably different biological infrastructure to instantiate their consciousness. Their "brains" are distributed through their bodies. Therefore, Seth speculates about "what it is like to be an octopus arm."
Seth doesn't say a lot about machine intelligence. I thought this was a mistake. There is a lot we can learn about consciousness from the cutting edge of machine intelligence. I bet that AI will help us make the most significant advances towards understanding consciousness. As Feynman said, "What I cannot create, I do not understand."
Seth mentions in passing that he believes that intelligence and consciousness are distinct and that you can have an "intelligent" system that is not "consciousness" and vice versa. I found this to be poorly justified. There are good reasons to believe that both consciousness and intelligence are processes of inference/modelling. My hunch is that intelligence and consciousness are deeply intertwined (but Seth is the world-class cognitive scientist, not me). This distinction is a debate about the definitions of words, not the workings of intelligence or consciousness and is therefore unlikely to be fruitful.
I thought this was an exciting way to capture our experience of the exponentials: "Where are we on this exponential curve? The problem with exponential curves—as many of us learned during the recent coronavirus pandemic—is that wherever you stand on them, what's ahead looks impossibly steep and what's behind looks irrelevantly flat. The local view gives no clue to where you are." This makes sense because the derivative of is itself. Hence you cannot tell where you are on the curve.
Overall, this is a decent book. It offers no new insights, but it nevertheless offers a good non-technical synopsis. Seth covers a wide range of topics with great depth while presenting the reader with easy-to-grasp explanations. The lack of conclusive answers might be disappointing to some, but I don't expect us to find ultimate theories explaining the whole of consciousness any time soon. That being said, we shouldn't treat every momentary difficulty as insurmountable or evidence that consciousness is somehow non-physical. Instead, let's keep working towards finding a precise explanation.
If you are interested in a more technical synopsis on consciousness, I think this is a good technical survey paper, The rise of machine consciousness: Studying consciousness with computational models.