On the one hand, a main theme here is theories of consciousness. On the other hand, it’s been almost eight years blogging, and I’ve covered my views pretty well in numerous posts and comment threads. Our understanding of consciousness currently seems stuck pending new discoveries, either in answering hard questions, or in providing entirely new paths.
A while back I determined to step away from debates (even blogs) that center on topics with no resolution. Religion is a big one, but theories of mind is another. Your view depends on your axioms. Unless (or until) science provides objective answers, everyone is just guessing.
But it’s been three-and-a-half years, and, well,… I have some notes…
Recently I was asked why I’m so confident about my, perhaps counter-mainstream, view on consciousness, and this post is my long answer. (I can go on at more length here, but will try to keep it succinct.)
I say my view may be counter-mainstream because I do not believe what I perceive to be a common belief: that human consciousness can be simulated in a computer.
In particular, I do not believe downloading a human mind is, or will ever be, possible. I do not believe a simulation of consciousness (which a download requires) is possible, even in principle.
(Actually, I’ll hedge on that last just a smidge, because I can’t rule out that it’s entirely impossible. Almost certainly practically impossible, and probably even effectively impossible. But absolutely, literally impossible? Not 100% on that.)
A key distinction is between a software simulation of a brain (or mind) and a hardware replication of a brain. I think the latter could give rise to a new mind.
A simulation is a model — computed with numbers — of some process (including, potentially, itself). There are models of weather, vehicles, population, sales, and even models of other models.
A crucial point is that, with one exception, models never have the same effect as what they model.
Weather models don’t produce rain or wind.
A model of a laser doesn’t produce light.
The one exception is when a model models another model. Since the result of a model is a list of numbers, another model can produce the same list of numbers, even if it functions in a completely different way.
The idea that the human brain/mind can be modeled successfully (that it will produce consciousness) depends on the brain/mind being a model in the first place.
This is an important point that I think a lot of believers hand-wave away (or just ignore).
Effectively, it posits a form of Platonic dualism, a belief that mind is something not just distinct from its (only known) mechanism, but it is a mathematical object!
Which seems in conflict with a physicalism belief that mind emerges as ‘something the brain does.’
Even the idea that consciousness is ‘what information processing feels like’ depends on the idea of an emergent mind as a consequence of a physical process. How can a simulation of that process cause the same thing to emerge?
I do see it differently when it comes to hardware that replicates the structure and operation of the brain.
Specifically, a very large, fully parallel, massively interconnected network with analog properties. (At the least. There may be other spatial or physical constraints to making it work.)
Unless there is something special about biology, or something like souls are real, I see no reason hardware sufficiently like a brain wouldn’t work.
It might need to be carefully trained (like a bio-brain), but presumably hardware works much faster than bioware. And once it was trained, as hardware, it might be possible to clone copies easily.
Further, such a brain/mind, as hardware, potentially has access to digital information. The whole internet can be its memory.
Speed of thought vastly faster (potentially) and access to all that information. Perhaps in a durable very strong body (with laser eyes).
These would be new “people” — a new kind of actor on the stage!
(And we might be able to clone armies of them…)
Getting back to the question at hand, why has my view settled where it has, my logic goes something like this:
Fact #0: Cogito ergo sum. I perceive that I am conscious. In this context, I accept realism, including that other humans are equally conscious.
Fact #1: Human minds emerge from human brain function. Somehow.
Fact #2: There isn’t one. We don’t know!
So let’s speculate.
What else emerges from physical function? Lots of stuff. Emergence — that is to say, new behaviors emerging at higher levels of organization — seems a solid physical idea, a property of reality.
(This is not to argue that emergent behaviors can’t be fully explained by the component parts. For purposes of this discussion, I’m taking physicalism and reduction as axiomatic.)
Let us speculate that mind emerges in consequence of a large, complex, highly interconnected, parallel physical network, like a brain. As mentioned previously, this seems a solid assumption.
But what about a numerical model, a simulation, of a brain? Can that work?
Can a mind arise from crunching numbers?
Back in 2015 I wrote a long series of posts exploring the idea of simulations in general, including with regard to simulating the brain (let alone the mind). The final post summarizes the main points. (And has a list of links to the other 17 posts in the series!)
A key point I’ll just mention here is that speaking just computationally, I think a model of a human brain will fail in the same way weather models do: mathematical chaos makes the calculation diverge from reality.
This is an inescapable consequence of how digital computers work.
The computational requirement is along the lines of telling me precisely what the weather will be one year from today. As a physical system, that answer does exist right now, but it’s computationally inaccessible. (Practically and perhaps in principle.)
Bottom line here:
- Within the context of physicalism, I find it hard to accept the dualism of mind being distinct from brain such that it can be platform agnostic.
- I see the computational requirements of analog brain simulation potentially out of reach of conventional computing, and perhaps just plain out of reach (as with weather).
- I don’t see how a simulation can produce the same thing as what it models when its inputs and outputs are just numbers. Only physical processes produce physical effects.
I see almost no way around these conclusions.
A common assertion involves the brain, or our minds, being like a computer. That we can learn to do math is offered as supporting evidence.
Let me be clear: The brain is nothing like a computer. Nothing at all.
That neurons can be interpreted as “logic gates” doesn’t make any structure with “logic gates” necessarily a computer. A system of pipes and valves can act like it has logic gates, but it isn’t a computer.
Except in the most metaphorical sense, and part of the problem here is the vagueness of words (with “consciousness” itself at the top of the list). What we mean by “calculate” and “compute” can vary, with some taking a broad view and others (yours truly) taking a narrow view.
(I explored exactly what “calculate” and “compute” (and therefore “computer”) mean to computer scientists in that long series I mentioned.)
The relationship most humans have with math seems, to me, pretty solid proof the mind is absolutely not, no way, no how, a computer.
Just consider all the bullet points:
- Nearly everyone needs to be taught math.
- Many hate it. Some claim they are incapable of it.
- Lots of math is so hard that we need computers!
- Our math and statistical intuition is notoriously bad.
- Many people struggle to use logic (a form of math).
There are more subtle points, such as our inability to see patterns in lists of numbers (but those patterns jump out when we use a computer to visualize the data for us).
It’s funny to me, this idea that just because some of us can learn math, it demonstrates how computer-like our minds are. Really, when you look at it closely, quite the opposite is true!
Our brains, and our minds, are far from being anything computer-like.
Not in any von Neumann architecture sense, nor, for that matter, even any digital computation sense. I’d question whether even quantum computing adds anything to this mix.
My final question with regard to brain simulation involves where the consciousness resides.
With humans, it clearly resides in the physical operation of the living brain. If we simulate that numerically, where is the consciousness?
Inputs would be numbers taken from various sensors, memory, and data from whatever sources. The simulation would do math on these numbers resulting in output numbers.
Presumably, for example, one input number stream might be from a microphone that allows the system to “hear.” An output stream might go to a speaker that allows the system to “speak.”
So I ask this system, “Are you conscious?”
It converts the sound of my question to numbers, analyzes those numbers, looks up a bunch of other numbers, and produces numbers that drive the speaker to produce the sound of, “Huh? Say what?”
Or whatever numbers it produces.
Where is the consciousness? In the numbers? In the process? In the result?
I’m skeptical — very, very skeptical — it can reside anywhere at all.
It does raise an interesting question if the system replies, “Hell, yes!”
Does a simulation just think it’s conscious?
(Remember Janet pleading desperately to not be rebooted in The Good Place?)
How can we know if it really has an inner life even if its output numbers claim one?
Does that even matter? What is the difference between a system that acts conscious and one that really is?
(Perhaps where it matters is in what rights we might accord.)
So, no, I’m not at all onboard with simulated or uploaded minds.
I could be wrong, of course, and maybe our brain/mind system really is a biological Turing Machine evolved across eons. It does seem to be the most complex biological system we’ve encountered.
Science will keep chipping away at it, and, who knows, maybe it’ll crack that nut at long last.
On the other hand, science has found areas beyond itself; Turing, Gödel, Cantor, and Heisenberg, all contributed to seeing the limits. Maybe some aspects of consciousness will remain forever ineffable.
Personally, I’d be fine with that.
Stay conscious, my friends!