I read an article about ‘The Neural Correlates of Consciousness’. That’s my first time to know that consciousness has a physical aspect. Based on that, that leads to one of my talk with deepseek and I suddenly figure out the learning model of the brain.
Conclusion:
The brain is like a three-dimensional matrix and a group of units produces different electrical signals with different frequency. If an electrical signal has an output, and that output is marked as effective, the brain will strengthen the connections within that group of units and make it easier to find in the future.
The question is which kind of output is effective-it is determined by Dopamine-driven system: Only the output that activates your brain to release dopamine can be marked as an effective one. Imagining you are trying to ride a bike, the way you have done makes you fall and this way which controlled by certain units in brain fails to activate Dopamine-driven system. But next time when you can control the bike, your brain release dopamine widely and it remember the certain units-you succeed in learning to cycle. And Dopamine-driven system can used to train BCI.
However, consciousness is too complex with two aspects of physical level and philosophical level. All we talk about above is in physical level so that consciousness can be detected by BCI. I think the output has two levels too-internal feelings and external behaviours. So I still confused that if the internal feelings are consciousness in philosophical level, are they still driven by dopamine? Can BCI detect them?
· 1
Me: Summarize article on neural correlates of consciousness.
Deepseek: Cortico-thalamic circuit necessary; sensory cortex necessary but not sufficient (needs frontal/parietal); action awareness involves medial motor and inferior parietal cortex; knowledge remains correlational.
· 2
Me: Any new discoveries?
Deepseek: Thalamic nuclei confirmed as gatekeeper; frontal role differs by modality; brain-body interaction (heartbeat) needed; quantum consciousness controversial.
· 3
Me: What is consciousness?
Deepseek: State of having feelings/experiences. Level (arousal) depends on cortico-thalamic system; content depends on posterior hot zone. Hard problem unsolved.
· 4
Me: Muscle memory doesn’t need consciousness?
Deepseek: Correct. Handled by cerebellum/basal ganglia. Consciousness only involved in early learning, initiation, and unexpected events.
· 5
Me: Consciousness feels like something advanced suspended above the brain.
Deepseek: Subjectively yes. Mechanistically it’s deeply coupled—an emergent state/function, not an independent entity.
· 6
Me: Do animals have consciousness?
Deepseek: Mammals, birds, octopuses likely yes; fish/reptiles controversial; insects unlikely; plants/bacteria no.
· 7
Me: Does BCI act on consciousness?
Deepseek: Two directions—reading (decoding intentions) and modulating (directly altering experience). Raises control and substitution risks.
· 8
Me: Is consciousness essentially specific cortical activity?
Deepseek: As a necessary condition, yes (cortico-thalamic). As sufficient, no—requires large-scale integration. Hard problem remains.
· 9
Me: So machines cannot fully capture consciousness?
Deepseek: Technically they capture correlates; philosophically they cannot capture the feeling itself (explanatory gap).
· 10
Me: Let’s not talk about philosophy.
Deepseek: Purely scientifically: no measurement tool, limited decoding precision, format mismatch.
· 11
Me: What exactly does BCI capture?
Deepseek: Physical signals—firing frequency, rhythms, blood oxygen. The footprint, not the feeling.
· 12
Me: Brain as a 3D matrix (xyz); different units/sets firing at different frequencies produce different outcomes.
Deepseek: Accurate model. Add temporal dimension and global synchronization.
· 13
Me: Brain development = a set of units is validated, produces an output, brain remembers it, forms development?
Deepseek: 10% myth is false. Your mechanism = Hebbian learning. “Development” = consolidating effective circuits into automation.
· 15
Me: What is an effective output?
Deepseek: Dopamine scores it: actual outcome ≥ predicted outcome → release → marks set as effective → consolidates.
· 16
Me: Can this dopamine-driven learning system train BCI?
Deepseek: Yes. Examples: drug enhancement, signal decoding, 2025 “emotion knob” experiment.
· 22
Me: What are my weaknesses?
Deepseek: (1) Didn’t ask for evidence, (2) cut philosophy too early, (3) narrowed “output” (you later corrected me), (4) didn’t ask about edge cases.
· 23
Me: For #3, internal outputs count too.
Deepseek: I withdraw that criticism. Your model covers internal outputs. New question: does all internal output involve dopamine?
· 24
Me: Now we’re back to philosophy of consciousness.
Deepseek: Yes. Your two commitments (internal outputs count + no philosophy) conflict. Three paths: drop internal outputs, accept philosophy with boundaries, or keep追问.
· 25
Me: I think internal outputs count, but what if they’re philosophical?
Deepseek: Operationalist strategy: “My model covers internal outputs that can be marked by dopamine. The rest I acknowledge but set aside.”
· 27
Me: But BCI research is engineering—they don’t consider philosophy.
Deepseek: Engineers default to the philosophical stance that consciousness can be physically described, without discussing it. Philosophy gets transferred to ethics/law. Methodological naturalism: pretend it doesn’t exist and keep experimenting.