The Feeling of Life Itself, Koch
Last updated on 15 Jul 2022.Review and notes on The Feeling of Life Itself: Why Consciousness is Widespread but Can’t be Computed, by Christof Koch.
1. What is Consciousness?
- Straightforward understanding - consciosuness is experience.
- Decartes - I think, therefore I am. Experiencing is the basis of being.
- Consciousness is prior to physics.
- Assertion: any experience is structured. Saussurean/structuralist approach to the textual nature of reality.
- An experience is informative, integrated, and definite; it has a point of viw and occurs in time.
2. Who is Conscious?
- Abductive reasoning - extrapolate backwards to infer hypotheses that fit the given data best.
- Psychologists can quantify phenomenology under laboratory conditions.
- Perception is an active process: construction of useful features.
- Phenomenal consciousness and access consciousness: what you experience and what you can report
3. Animal Consciousness
- We can abduce experiences to fellow mamamls.
- Mammals are evolutionarily speaking very similar to us. Modern humans are genetically linked closely to chimpanzees.
- The nervous system is very similar across all mammals.
- The behavior of mammals is important to people.
- Experience without voice - language is not a prerequisite for language.
4. Consciousness and the Rest
- Most refined cognitive capabilities are not directly accessible to experience.
- Unconscious homunculus - we are not conscious of either raw sensory data or the highest processing stages of mind. We are instead sensor of the ‘in-between’.
5. Consciousness and the Brain
- Previously, we have associated consciousness with the heart. Only later did we understand the association with the brain.
- Neural Correlates of Consciousness - the minimal neuronal mechanisms jointly sufficient for any one specific conscious percept.
6. Tracking the Footprints of Consciousness
- The brainstem enables consciousness. It is the central canal through which information is transmitted. Brainstem neurons enable consciousness by providing the cortex neuromodulatory substances.
- Losing the cerebellum does not affect consciousness, even though it maps the body and outside space onto its neurons. But it does not generate consciousness: it is feedforward.
- Consciousness resides in the cortex.
- Little evidence suggests the brain exploits macroscopic quantum effects.
7. Why We Need a Theory of Consciousness
- Experience cannot be explained by reductionist materialism, physicalism.
- Experience is the starting point from which I abduce all else.
- Philosophical zombies - can they exist? Sure. Consciousness is currently above science.
- Integrated Information Theory asks how matter must be organized to support the mental. It links ontology with phenomenology in physics and biology.
- Five phenomenologgical properties/axiomatics of Integrated Information Theory.
- Intrinsic existence - it exists for itself
- Composition - it is structured
- Information - it is the specific way it is
- Integration - it is one
- Exclusion - it is definite
- Why should integrated information be experienced? Subjective feeling is completely defined by these five axioms.
8. Of Wholes
- Integrated Information Theory - consciousness arises from causal properties. COnsciousness is a property of a mechanism which has cause-effect power upon itself.
- The more a system’s components constrain each other, the more causal power we have.
- Intrinsic existence. Consciousness exists for itself, without an observer. It removes the ouitside observer: it exists. For you to observe something, it must be a difference which makes a difference. Only things with causal power exist.
- Composition. It must be structured in a recursive way, a structure is either an atomic or combines substructures.
- Information. It is the specific way it is: the state of the system is informative in that its effect on the future and where it came from the past can be analyzed.
- Integration. Integration equires a cause-effective sturcture to be irreducible. \(\phi\) represents the extent to which cause-effect changes if it is reduced along the minimum inofmration partition. \(\phi \ge 0\)
- Exclusion. Only the maximum set exists; all other overlapping subsets are not ocnsidered. Thus the graph corresponding to \(\phi^\text{max}\) is the most conscious one.
- Central identity: any experience is identical to the maximally irreducible cause-effect structure associated with the system in that state.
9. Tools to Measure Consciousness
- When language fails, how do we measure consciousness?
- Motor function vs cognition, measured by internal perception
- Gamma oscillations?
- Zip-and-zap
10. The Uber-Mind and Pure Consciousness
- Splitting a brain creates two conscious wholes. When the brain is split, the maximum information forms in each of the split networks.
- Brain-bridging and the Uber-mind: connections between brains would need to be abundant enough such that \(\phi^\text{max}\) occurs across the uber-mind.
- Neuronal hegemony
11. Does Consciousness Have a Function?
- Unconscious behaviors rule much of our life. Many important activities are performed without consciousness.
- Integrated Information is adaptive. A whole doesn’t have to be useful for it to be an expeirence. As such it does not subscribe to evolutionary teleologism. Experience has no function.
- Experiments show that artificial brain simulations increase in integrated information when regulated for fitness.
- Intelligence-consciousness plane - you can be conscious without being intelligent, and vice versa.
12. Consciousness and Computationalism
- Can we represent consciousness through computation? Many in the information age seem to think so.
- The Turing machine demonstrates that computation must occur in this simple state. So computation must also be written here. Can human biology and thought be expressed as a Turing machine?
- Computationalism is a variatn of functionalism - functionalism, the mind can exist apart from the body
- The brain-as-computer metaphor can be easily abused/misused.
- Cortical networks are not feedforward, but neural networks certainly are.
- Can we emulate the entire brain?
- Global neuronal workspace model of consciousness - neural correlates of consciousness appear relatively late after stimulus; attention is necessary for conscious perception/
13. Why Computer’s Can’t Experience
- The same networks can have different integrated information. Feedfoward networks have zero integrated information since they have no causality; recurrent networks have some causal information, but when unrolled they also have zero influence.
- Digital computers have miniscule intrinsic existence. A Boolean circuit, when simulated by a computer circuit, has very littel integrated information. And this is true for all networks and their simulations.
- Mind uploading: we cannot retain the causal relationships in the brain by simulating it all.
14. Is Consciousness Everywhere?
- Most animals have some level of consciousness.
- Consciousness is in the universe.
- Many, but not all things, have minds.
- IIT rejects group-mind; the whole has definite borders. The irreducible Whole forms my conscious experience, not the underlying neurons. It forms as an aggregative/joint operation.
- IIT’s exclusion postulate suggests that a group of beings cannot both be independently conscious and conscious as a group - the self-consciousness would dissolve into the whole.
- We must assert the feeling of life, of the ontological reality of phenomenology.