A Thought Experiment
Imagine one billion people, each holding a coin.
If the heads side is facing up, we will consider it to represent a 1. If the tails side is facing up, it represents a 0.
Now imagine all these people flipping their coins, one side and then the other, in some coordinated sequence. If we could observe all the coins at once, we would see an enormous stream of zeros and ones constantly changing over time.
Now imagine that each person flips their coin in perfect synchrony with the firing of a particular neuron in a human brain — while that person is looking at the surface of the ocean on a clear, sunny day.
And imagine that every neuron has its "own" person, and therefore its "own" coin.
Would the coins somehow feel the blue of the ocean in the same way that you do?
Replace the coins with bits in an artificial neural network, and you get essentially the same situation.
Replace the coins with living neurons, and you still get the same situation.
If consciousness can be explained entirely by the activity of neurons, then something seems to be missing from this explanation.
Color, taste, sound, smell, the sensation of heat and cold — all of these seem to be something more than mere information.
A computer can simulate a living neural network and produce similar responses to stimuli. It can recognize an image, a sound, speech, or even a smell. It can analyze language and understand the meaning of a sentence.
And yet, it still does not see the color blue.
Nor can it hear the voice it has recognized.
Something is still missing.
Why Do I Think Something Is Missing?
Why do I think that something is missing?
Because our understanding of the processes taking place in the nervous system is, at its core, based on the firing of neurons and the connections between them. This is, of course, a vast oversimplification. Behind it lies an incredibly complex set of biochemical and electrical processes. But, ultimately, the effect is something like this: neurons activate and fire impulses that travel along their axons toward their destinations.
These signals eventually result in an eye blinking, a little finger moving, information about light being transmitted to other neurons, and, finally, in meaningful responses to the environment and the planning of future actions.
In fact, by simulating this simplified understanding of these processes, we have created AI.
AI can recognize the space around it, speech, sounds, and images. It can understand and plan. It can produce responses and take actions in order to accomplish a task.
Just like a human — and sometimes even better and more efficiently.
So we have created intelligence that cannot feel the blue of the ocean, yet can recognize that blue, describe it, and perform an action based on recognizing it.
It can recognize a white ship sailing across that ocean and say:
“Hey, there’s a ship!”
And yet it still does not experience the whiteness of the ship. It does not smell the scent of an orange peel. It does not feel pain.
That is why I wrote that something is missing.
Something about the consciousness of living beings still remains unexplained.
This is precisely where the concept of qualia comes in. Qualia refers to the subjective, first-person qualities of experience — what it is like to see a particular color, smell an orange peel, hear a particular sound, or feel pain.
This dimension of experience remains a mystery to science. And if we do not yet understand what produces it in living beings, then naturally, we have no way of simply transferring it into an AI system.
Another Interesting Observation in Our Thought Experiment
What can we notice in our thought experiment with a billion coins?
We said that a coin with the heads side facing up represents a 1, while a coin with the tails side facing up represents a 0. In this way, we can actually encode information.
If we arrange the coins in a row, we can write this entire text in binary form — as a sequence of zeros and ones. Someone who knows how each letter is encoded in binary could read this text from the sequence of coins.
But if nobody knows the code, then it is simply a sequence of randomly oriented coins.
Who Is Reading the Coins in the Brain?
In our thought experiment, suppose that we have one corresponding coin somewhere in space for every neuron in the brain. The state of each coin would mirror the state of its corresponding neuron.
If the coins were flipped — each one synchronously with its corresponding neuron — would those coins have consciousness?
If the output coins, corresponding to the output neurons, were connected to muscles, we would get the same movements. They would produce the same voices, words.
Does that mean that the coins would think the same thoughts that, for example, I am currently thinking while writing this text?
What is the functional difference between billions of coins and billions of neurons?
If the coins would not think, why not?
And if they would — then things become much more complicated.
To clarify, I deliberately chose coins as the example. I could just as easily have used the state of bits in a computer's memory. Functionally, the situation is essentially identical.
Information has to be represented somehow.
In a neuron, we might describe this in simplified terms as an active (excited) or inactive neuron. In a computer, it might be a 1 or 0 stored in a bit. It could be an electrical charge on a medium, a pulse of light travelling through an optical fiber, or simply a coin whose position stores the information.
So, let us imagine that our brain, consisting of billions of neurons, has an equivalent system of billions of coins. Each coin is either heads or tails, changing synchronously with the corresponding neuron in the brain.
And suppose that this collection of billions of coins really does have consciousness. Suppose it actually thinks.
It does not see the color blue — but it thinks.
And it thinks because, at the beginning of our experiment, we defined tails as 0 and heads as 1.
But who are we to make that decision?
What if we define the opposite rule? Tails as 1 and heads as 0.
Or some completely different rule?
Would the coins suddenly lose their consciousness because they are no longer synchronized with the neurons in our brain?
And yet, at exactly the same time, would they still have consciousness because, according to our original rule, they are synchronized?
Does Time Matter?
Over what period of time must the coins remain synchronized?
What if they are synchronized, but everything happens at half the speed? Does consciousness still exist?
What if they are synchronized, but in the reverse temporal order?
Would consciousness still exist — but instead of thinking from the past toward the future, would it be moving from the future toward the past?
Does Physical Distance Matter?
Do the coins have to be physically close to the neurons?
In the brain, this seems necessary because neurons need to be connected and communicate with one another.
But imagine that we could physically separate the neurons while keeping them alive, and connect them using optical fibers in exactly the same pattern in which they were connected inside the brain.
Our brain could then occupy an area several kilometers across and still function in exactly the same way.
So, if physical distance is not fundamentally important, our coins could be separated by light-years.
As long as they remain synchronized with our neurons, they would have consciousness.
They would think.
Now replace the coins and their heads-and-tails states with the spin states of electrons.
Somewhere in the universe, there would presumably be a collection of electrons whose state happened to be equivalent to the state of the neurons in our brain.
Would those electrons — that collection of particles scattered across the universe, perhaps separated by light-years — be thinking our thoughts?
Would they, at this very moment, be trying to scratch the tip of our nose?
And who decides which rule should be applied to those electrons scattered throughout the universe so that they are, right now, thinking the same thoughts?
Perhaps the most unsettling question is not whether the coins can think, but rather:
And who decides which rule should be applied to the electrons scattered throughout the universe so that they remain synchronized with our neurons?
Maybe nobody does.
Maybe different parts of the universe are actually thinking exactly the same thoughts that we are thinking right now — but under different rules of “reading” the information.
Disk drive brain
Now imagine that it were possible to record every firing of every neuron in your brain onto a disk drive.
It would be an extremely, extremely long binary file.
If we played it back sequentially, one “frame” at a time — with each frame representing the set of bits corresponding to the state of every neuron at a particular moment — the sequence would reproduce every thought you had throughout your entire life.
Let me spice things up a little. I don't think the story is confusing enough yet.
When I say “play it back,” you probably imagine a movie being projected onto a screen.
And it could be exactly that.
Imagine that each set of bits is projected onto a huge screen containing the equivalents of all the neurons in your brain. In this way, our original thought experiment with the coins could be mapped onto bits displayed on the screen.
Of course, over time, these patterns would change, one after another, and in doing so they would form the sequence of “thoughts” that passed through your mind throughout your entire life.
But here comes the question again:
Why project it onto a screen at all?
Why isn't the recording on the disk already enough?
Who is reading that recording and deciding that it has to be projected onto a screen?
And why does it have to be played back through time?
The Core Idea of the Thought Experiment
The important point in this thought experiment is that the rule we use to interpret the coins is our rule, not the coins' rule.
We arbitrarily decided that heads means 1 and tails means 0. This allows us to read the state of the coins and compare it with the state of the neurons in a brain. But the coins themselves do not know anything about this convention. Their physical states simply exist.
We could just as easily define the opposite rule, or use a completely different algorithm to interpret the same physical states.
This leads to an important distinction between the physical process itself and our interpretation of that process.
If consciousness is a property of the underlying physical process, then it should not depend on how an external observer chooses to encode or decode that process.
In other words, the consciousness of a system should not require anyone to say:
“Heads is 1, tails is 0.”
The rule is necessary only if we want to read the output.
For example, if the output coins corresponding to motor neurons were connected to muscles, we could use our encoding rule to reproduce the same signals and therefore the same movements as the original brain. We could make the system say the same words, move the same finger, or perform the same action.
But the consciousness itself should not depend on this output or on our ability to read it.
This also means that there could potentially be many different physical systems implementing the same underlying process: neurons, computer bits, electrical signals, optical signals, coins, or even some completely different physical medium.
If two systems have the same relevant causal and functional organization, why should consciousness exist in one but not the other?
And if the answer is that consciousness does depend on the physical substrate, then we have to explain which physical properties matter and why.
The most unsettling possibility is that the same conscious process could be physically realized in many places at once — perhaps even in completely different parts of the universe (or time) — without any of those systems knowing that they are being interpreted as copies of one another.
The observer's choice of an encoding scheme would then be irrelevant to consciousness.
The question would no longer be:
“How do we decode the system?”
but rather:
“What physical conditions make a system conscious in the first place?”
How Do We Recognize Intelligence?
How do we recognize intelligence?
We usually recognize it through intelligent behavior.
We see animals organizing themselves into herds for protection against predators. We see predators developing strategies that make it easier to catch their prey — for example, selecting the weakest individual in a herd.
We communicate with other people and observe meaningful responses.
And finally, we interact with AI: we ask questions, and it gives us meaningful and informative answers.
In all these cases, we observe some kind of output — the result of a process taking place inside a nervous system or an artificial system.
From that output, we infer that something exists behind it: some internal process that produces the movement, the speech, the strategy, or the answer.
This is also why, in our coin thought experiment, we might initially assume that we need to define what represents 0 and what represents 1. We need some way to interpret the states of the coins and connect them to an output.
But this brings us to a very particular phenomenon in living beings:
Dreams
When a person is dreaming, they can become almost completely disconnected from the real world around them.
The only world that exists for them is the world created by their own mind.
They may receive no meaningful signals from the outside world — assuming nothing is disturbing their sleep — and they may produce no meaningful output to the outside world either.
Imagine that someone dreams of a blue ocean with a white ship on the horizon.
That image is being created entirely within their mind. Nobody else can see it.
If they dream that they are swimming toward the ship and move their arms through the water, their real arms remain still. Their physical body is lying motionless in bed. Only their “virtual” arms are moving through the water of the dream.
And yet, they are still conscious.
In an extreme case, that entire experience might remain forever inaccessible to everyone else.
Imagine, rather darkly, that the person dies in their sleep.
Nobody would ever know that they had been dreaming of a blue ocean, or that they had been swimming toward the ship.
There was no meaningful output to the outside world.
And yet, while it lasted, they were conscious. They were there.
They were swimming in the ocean.
Why Does This Matter?
I bring this up to make a simple point:
Consciousness — and even subjective experience, or qualia — does not necessarily require an output to the external world.
The experience can exist entirely within the system.
So, if we return to our billions of coins, perhaps we do not need to connect them to muscles, speakers, screens, or any other external output for them to be conscious or to experience something.
The coins would not need to tell us what they are experiencing.
The only remaining question would be:
Who — or what — determines what their states mean?
Who decides which state is 0 and which is 1?
And perhaps the even deeper question is:
Does anyone need to decide at all?
What if the coins are conscious and unconscious at the same time?
What if their consciousness depends entirely on interpretation?
Under one interpretation, they form a conscious system — a mind thinking thoughts, perhaps even experiencing the blue of the ocean.
Under another interpretation, they are nothing more than billions of pieces of metal, randomly flipping between heads and tails.
The Good Old Record Player
How did the good old record player produce sound and play music?
A needle moved through grooves spiraling across the surface of a vinyl record, “reading” the physical pattern encoded in those grooves and turning it into sound.
Data was recorded and later read from magnetic tape in a similar way. A magnetic head remained in place while the tape moved beneath it, and the magnetization of the tape represented zeros and ones. Those bits were then transferred to a computer, and we could start playing Jet Set Willy on a ZX Spectrum.
Something similar happens inside a hard disk — at least the old-fashioned kind that actually contains a spinning disk with a magnetized surface.
Why am I telling you this?
Imagine that we could record the state of a single neuron — biological or artificial — onto such a medium.
Suppose that reading the recording would give us the state of that neuron at a particular moment in time.
Now imagine having as many such recordings as there are neurons in a human brain, with each one containing the state of one particular neuron.
If we could read all of them simultaneously, we would reproduce the state of the entire system at that particular moment.
And if we follow the logic of our coin thought experiment, perhaps we would also reproduce consciousness.
There is an interesting difference here.
We no longer need one coin that someone is continuously flipping between heads and tails. Instead, we have a large collection of coins, each already positioned on the appropriate side, with that state corresponding to the state of a particular neuron at a particular moment.
Reading those states gives us the state of the conscious system.
But this raises an important question.
Do we actually need the process that produced those states?
If we already have the complete state of the system, why do we need to reproduce the mechanism that originally generated it?
The connections between neurons may be essential because they determine what happens next — they are part of the causal dynamics that transform one state into another.
But if we already have the resulting state, perhaps we no longer need the mechanism that calculated it.
This is especially interesting when we think about artificial neural networks. The “connections” between artificial neurons are ultimately represented by numbers and algorithms. They determine how the next state of the network is calculated.
But what if we already have that next state?
What, exactly, would be missing?
Perhaps consciousness does not depend on the mechanism that calculates a state, but on the state itself.
Or perhaps it depends on the dynamics.
Or perhaps it depends on something else entirely.
And this brings us back to the coins.
It does not even have to be one coin.
It could be an atom, a photon, an electron, or some other physical object. There could be billions of them, each occupying a particular physical state.
They could be inside a single drop of water.
They could be distributed across a laboratory.
They could be scattered across the universe.
What matters, perhaps, is not what the objects are, but whether their physical states correspond to the relevant states of the neurons in a conscious system.
And now we arrive at a much stranger possibility.
Is consciousness the information, the physical representation of the information, or the process of reading/transforming that information?
What if consciousness is not confined to a particular physical object?
What if the same conscious state can be physically realized in many different places?
What if the physical equivalents of the states of my neurons exist somewhere else in the universe — perhaps thousands, millions, or billions of light-years away?
Would that distant arrangement also be me? Would it experience the blue of the ocean? Would it be thinking the thoughts I am thinking right now?
And what if we include time?
What if equivalent states exist not only somewhere else in space, but at different points in time?
Perhaps some physical arrangement in the distant past corresponds to a state of my consciousness.
Perhaps another exists somewhere in the future.
Could my consciousness, in some physical sense, extend from one end of the universe to the other — and from one end of time to the other?
Not as a mystical entity floating through the universe, but simply as a consequence of the physical realization of the same informational or functional state.
Perhaps a single atom, photon, or electron could play the role of one element in such a system. They could be arranged inside a drop of water, scattered across the universe, or distributed across spacetime. If their physical states corresponded to the relevant states of our neurons, would the resulting configuration be enough to produce a tiny fragment of consciousness?
Or would we still need something else — some process, some physical interaction, some property of matter that we have not yet discovered?
Then the question becomes much larger than whether an AI is conscious.
It becomes a question about the nature and location of consciousness itself.
Perhaps consciousness is a property of a particular biological substrate.
Perhaps it requires a particular causal process.
Perhaps it requires both the state and the dynamics that produce the next state.
Or perhaps the state itself is enough.
If the latter is true, then the universe could contain countless physical realizations of conscious states - distributed across space and time, without any of them knowing about the others.
But there is another possibility.
Perhaps a perfect description of a conscious state is still only a description. Perhaps a recording of a state is not the state itself. Perhaps the pattern on the tape is no more conscious than the grooves on a vinyl record are music.
The record contains the pattern. The needle turns that pattern into vibrations. The vibrations become sound.
But what is the equivalent of the needle for consciousness?
And if no needle is necessary - if the physical state itself is enough — then perhaps we have to ask an even stranger question:
Where does one consciousness end and another begin?
Perhaps the real mystery is not whether an AI can be conscious.
Perhaps the mystery is whether consciousness is something that belongs to individual brains at all.
But Is the Neural Signal Really Digital?
At first glance, there seems to be a problem with our simplified idea that a neuron is just a binary device.
If a neuron either fires or does not fire, how can the nervous system control something as seemingly analog as the strength of a movement?
How can I produce a very weak grip with my hand, a stronger grip, or an extremely strong grip if the signal coming from each neuron is apparently just 0 or 1?
It turns out that this is a well-studied area of neurophysiology. The control of muscle force depends on both the recruitment of motor units and, importantly, on the rate at which those motor units discharge action potentials — a mechanism known as rate coding.
(Rate Coding and the Control of Muscle Force)
For example, the activity of a motor neuron could be represented as a sequence of discrete events:
Weak activation:
1---------1---------1---------1
Stronger activation:
1---1---1---1---1---1---1---1
Fast / explosive activation:
1-1-1-1-1-1---1----1-----1The individual action potentials are still essentially discrete, all-or-none events. What changes is when they occur and how frequently they occur, as well as which motor units are recruited. In voluntary contractions, motor-unit discharge rates commonly fall in the tens of pulses per second, while much higher instantaneous rates can occur during rapid, ballistic contractions.
So perhaps the signal is not really analog after all.
Perhaps it is still fundamentally a sequence of binary events — a neuron either fires or it does not — but the information is encoded in the temporal pattern of those events.
And this is enough for our thought experiment.
We do not need to model the membrane potential, the synapses, the biochemical processes, the thresholds, or the mechanisms that cause a neuron to fire.
After all, nothing is fundamentally binary.
Even in a computer, the physical quantity representing a 1 is not some perfect, discrete entity. A voltage or charge can vary within an allowed range and still be interpreted as the same logical 1. The same is true for 0.
Binary is not a fundamental property of the physical world. It is a concept we impose on a physical system.
We define a range of physical states as 0, another range as 1, and deliberately ignore the differences within each range.
The same idea applies to my thought experiment with neurons.
A neuron is not fundamentally a binary object. Its membrane potential, ion concentrations, neurotransmitters, synaptic states and countless other physical variables are continuously changing. But if, for the purpose of the experiment, I care only about whether the neuron fires an action potential or does not fire one, I can represent it as:
0 = no action potential1 = action potential
The question is therefore not whether the brain is really binary.
It isn't.
The question is whether the information relevant to consciousness can, in principle, be represented at this level of abstraction.
If it can, then billions of neurons become a gigantic sequence of binary states evolving through time.
And that brings us straight back to the coins.
We simply define the state of a neuron as:
0 = no action potential
1 = action potential
The only additional parameter we need is time.
We choose a sufficiently small time interval in which the state of a neuron can change.
For example, if we want to distinguish firing rates up to 200 Hz, the interval could be:
1 / 200 second = 5 milliseconds.
If we wanted to distinguish events occurring at up to 1,000 Hz, we could use:
1 / 1,000 second = 1 millisecond.
For the purposes of the thought experiment, the interval can be made arbitrarily small.
At every such moment, every neuron is simply in one of two states:
0 or 1.
With billions of neurons, we therefore have a gigantic binary vector at each moment in time:
Time t₀:
010011001010...
Time t₁:
010011101010...
Time t₂:
110011101010...
Time t₃:
110111101010...The complete neural activity is therefore not a single binary number, but a sequence of binary states evolving through time.
And that brings us right back to our original thought experiment with the coins.
The coins do not need to reproduce the chemistry of the neurons.
They do not need to reproduce the voltage across the membrane.
They do not need to reproduce the mechanism that caused a neuron to fire.
For the purposes of this thought experiment, we only need to know whether each neuron fired within the chosen time interval.
Heads = 1.
Tails = 0.
And if we can reproduce the complete sequence of those binary states, we have reproduced the neuronal activity pattern that we are using to describe the system.
The question remains:
Would that be enough to reproduce consciousness?
No comments:
Post a Comment