On the Analogy Between Mind/Brain and Software/Hardware (1992) | Hacker News Reader

Hardware (1992)

massline.org

40 pointswsgeorge34 comments

The article makes an interesting point in fleshing out an analogy between (a specific reading of) the concept of "software" and the mind, but does so in a needlessly condescending tone, which is very distracting.

Take this example: "To begin with you may think of software as a series of instructions to the whole system (including both the hardware and the rest of the software) [...] This is as far as programmers in business data processing centers usually get. That is why so many of them keep insisting that "no computer can possibly think; if you know anything at all about computers you should know that they just follow directions". This is an incredibly naive view which only reflects the narrowness of their own conceptions and programming practices."

I would argue that defining software as "a series of instruction to the whole system" is perfectly valid. The author, instead, takes "software" to mean the functional properties of dynamic states of the system. Which also sounds like a valid definition, which leads to the mind / brain analogy.

edit: formatting

This 1993 article definitely reads like the Mind/Brain concept in the era of Gofai (good old fashioned AI).

It's quite interesting just for its historical outline, even if the outline is pretty sparse (you gotta start somewhere). The main thing that I think modern AI/Machine-learning has discarded is the concept that intelligence is based on logical specification.

And I think it's entirely reasonable to say that a mind isn't "like software" if you mean human-constructed, top-down, logical-intentions based software.

The thing is, though, that neural networks and similar advances haven't left us with any alternative analogies for the mind/brain relationship. Should we come up with a new one?

Also, the mind isn't like software in that it doesn't get copied to other minds or mediums. Yeah, there are sci fi scenarios or futuristic predictions for this sort of thing, but it's not a reality at this point. Another difference is that there is no machine code for the brain. There isn't some set of primitive instructions programmers could target to instruct the connectome.

Definitely. The metaphor breaks down at some point.

> Another difference is that there is no machine code for the brain. There isn't some set of primitive instructions programmers could target to instruct the connectome.

We don't know this yet. But what we do know is, by directly stimulating certain parts of the brain, we can recreate certain sensations.

I looks to me like this is close as close to some "primitive instruction set" as we've come.

> There isn't some set of primitive instructions programmers could target to instruct the connectome.

I wouldn't be so quick to rule that out. Think about the difference between short term memory and long term memory. Short term memory is on timescales far shorter than synaptic formation, so it stands to reason that it might take the form of electric patterns held in buffers in the brain (i.e. it is not encoded using synapses). If that is true, then why can't buffers that hold electric patterns also hold a set of instructions that can be retrieved from storage?

E.g. maybe short term memory is more like an FPGA (easily reprogrammable circuits) and long term memory is more like an ASIC (hardwired synaptic connetions that are far more dense and long-lasting).

Odd though, the author just recently reviewed another mechanistic consciousness book and recommended this old essay [1] in the review. There seems to be a real resistance towards accepting what's hard about the hard problem.

http://www.massline.org/ScienceGrp/ConsciousnessAndBrain.pdf

>There seems to be a real resistance towards accepting what's hard about the hard problem.

Well, there also seems to be a real resistance to even phrasing the "Hard Problem" in terms that both admit a nontrivial role for functional cognition (for "mere brains") and also don't presuppose dualism.

I'm in the camp (following somewhat Daniel Dennett) that the hard problem just simply doesn't exist. It's invented. Some have objected to Dennett in that he just wishes the "hard problem" away. I submit that dualistic philosophers, crystallized by David Chalmers, just "wished" the hard problem into existence. It's a made-up question, the same as "why are fairies so hard to get on camera?" It's made-up by philosophers who just want dualism to be true. There is zero evidence that dualism is true.

Consciousness and the brain boils down to this: if we reject dualism, then we implicitly accept some form of materialism. Materialism means that the brain and whatever physical processes that it carries out must give rise to consciousness, must give rise to subjective experience, even if that involves nonlocal quantum mechanics or whatever cutting edge physics you care to imagine. If it's physics, it's materialism.

From the other angle, we keep asking ourselves "what is it like to be a bat?" or what is it like to be this or that. Well therein lies the answer. There probably is something it's like to be a bat, or a fish. But this is what it feels like to be a human consciousness, with all those processes going on in the brain, feeding that consciousness information. Is that really so hard to believe? Consciousness is just one of many programs running in the human brain. This is what it feels like to be that program running inside your brain.

>From the other angle, we keep asking ourselves "what is it like to be a bat?" or what is it like to be this or that. Well therein lies the answer.

Well, the thing is, I'd imagine that being a bat is probably like the flow of air through your wings when you flap, the sound of mosquitoes you can eat, the taste of fruit you settled for, the relaxed feeling of finding a branch or something to hang on and sleep, et cetera and et cetera.

That is, I expect that "what it's like to be a bat" is exactly what happens to a bat in life, as experienced through the bat's model of the world as constructed from the (exteroceptive and interoceptive) information brought in by its own sensorium and the dynamics of its body.

I don't know for sure, but I kind of suspect that bats don't have any sort of analogous "consciousness program" in their brains like humans. They probably go through life very automatically, without a conception of self, without a reflection on their existence, very much like a person who sleepwalks is unconscious, but in a far dimmer way. The same probably goes for most animals, and certainly for insects and worms. Only advanced primates, dolphins, whales, elephants, and maybe some birds are likely to have "consciousness" programs that are in any way analogous to what you and I experience.

After all, think of it: you don't experience the regulation of your heartbeat or other autonomic functions, or the rods and cones detecting light in your eyes, or the individual variations in sound intensity that make up audio. Instead, you experience a very highly processed and organized show of sorts that is presented to your consciousness by many various brain subsystems. Studies have started to show that even the decisions we make are determined by the brain far before we are consciously aware. In some sense, consciousness is the story our brain tells itself while all the systems are off busy doing their things. Maybe bats just don't have that, so there is nothing that our consciousness can conceptualize that is similar to their very automatic behaviors.

> I don't know for sure, but I kind of suspect that bats don't have any sort of analogous "consciousness program" in their brains like humans. They probably go through life very automatically, without a conception of self, without a reflection on their existence, very much like a person who sleepwalks is unconscious, but in a far dimmer way. The same probably goes for most animals, and certainly for insects and worms. Only advanced primates, dolphins, whales, elephants, and maybe some birds are likely to have "consciousness" programs that are in any way analogous to what you and I experience.

Nah, I suspect that most mammals are conscious, emotive, and at least somewhat intelligent. But of course, I'd have to understand consciousness a lot better to really say!

>Instead, you experience a very highly processed and organized show of sorts that is presented to your consciousness by many various brain subsystems. Studies have started to show that even the decisions we make are determined by the brain far before we are consciously aware. In some sense, consciousness is the story our brain tells itself while all the systems are off busy doing their things.

Welll.... not really. Consciousness, AFAWK so far, is what happens when parallel sensorimotor systems sample a maximum-a-posteriori perceptual, interoceptive, and affordance model of the world into a global workspace where certain precise, mostly deterministic sampling computations can be done as a part of active inference to determine which actions to emit.

Decisions aren't made ahead of time; (low-precision) affordances are computed ahead of time. It's not a show; it's the best model.

> Welll.... not really. Consciousness, AFAWK so far, is what happens when parallel sensorimotor systems sample a maximum-a-posteriori perceptual, interoceptive, and affordance model of the world into a global workspace where certain precise, mostly deterministic sampling computations can be done as a part of active inference to determine which actions to emit.

There seem to be some levels of abstraction here which I was trying to get at in my previous comment. Clearly we don't perceive the raw data of our sensory inputs getting dumped wholesale into our consciousness. We don't perceive the "pixels" of our eyes or the pressure of sound on our eardrum. That data is clearly processed by lower levels of the brain and turned into a more symbolic form. The visual system abstracts over the pixels, lines, borders, even shadows of visual input. The audio system recognizes pitch, volume, associates memory with patterns to classify sounds into churchbells, speech, scratching, animal sounds, music, etc. These things happen below the level of consciousness, since, try as we might, we cannot will ourselves into perceiving these details. Similar with our sense of time, our grasp and use of language, memory, etc. These all exhibit a form of abstraction that suggests that consciousness sits farther away from the base processing of the brain than what your comment suggests.

In my view, consciousness is kind of like the President of your brain. It doesn't see the full details of all the processing layers underneath; that would be overwhelming. It tries to make decisions and decrees and influence those subservient processes, but always along the chain of command there are those who have other ideas or intentions. So President Consciousness tries its best to steer the ship of hundreds, maybe thousands or millions, of diverse processes that are all loosely coupled. It doesn't see conditions on the ground, but always, always receives filtered reports that leave out details, get facts wrong, misrepresent and misconstrue.

Think about dreaming. Ever had a dream that was astoundingly rich and detailed, either visually or spatially? When you heard music or speech in your dream, or saw a complex thing like a person, do you think that the lowest level processors of your brain literally experienced the same input as if they had received it from outside? Clearly not. Your eyes are closed. Nothing in the environment is making those sounds. How would you see anything or hear?

There must be a kind of separation between the visual system which processes input from your eyes, and the system which constructs visual scenes, forms images, recognizes people, etc, because there simply is no raw input when you are asleep. So instead the scene-construction system and the audio classification system are working with "mock" data that is generated by other neural activity.

How can consciousness be "off" while there is still brain activity? In fact, during most of sleep there is no experience whatsoever, no memories created, even though brain activity continues. A global workspace doesn't explain how consciousness can be suspended.

> a mind isn't "like software"

The mind is like a state automaton. It has states and it transitions between them.

True, as far as it goes, but it tells us literally next to nothing about what a mind is and how it works. Unless the analogy is more specific than this, it does not advance our understanding much.

The thing is, though, that neural networks and similar advances haven't left us with any alternative analogies for the mind/brain relationship. Should we come up with a new one?

Then again, artificial neural network are written in programming languages and executed on computers. So you could turn this around and say that as AI becomes more sophisticated, software becomes more like the mind.

While most neural net experts have discarded the idea that neural nets simulate the brain, you may be correct in the long run.

The thing is that software in the absolute, arbitrary abstract can be like just about anything. But the arbitrariness means it stop being a human-comprehensible analogy - in contrast to the way some of the original AI schools thought of software as being a collection of rules.

>...Should we come up with a new one?

To me, the history of mind analogies, and of how they tend to be reflections of the then-current technology, suggests that making analogies is not particularly useful (and perhaps harmful, if doing so leads us to over-estimate our level of understanding.)

A fascinating subject. From a medium can come a new system of patterns. Like waves, really hard to figure their math, if you study water molecules. Plus waves are medium independent, and it gets really interesting as waves cross into a different medium (it bends).

Software is like waves, it needs a compute medium, but is independent from it. And software in the broad interpretation as information processing, indeed your brain seems like a compute medium, your mind is the information processing.

As for consciousness, the article mentions a follow up, anybody know if that exists somewhere? Personally I think consciousness is a pattern of information processing, a lot like reinforced learning[1]: keeping a history of observations, actions, and plans, and continuously introspecting and judging those plans.

So I wonder, if you run such an algorithm on a computer, would that feel like something to that program?

[1] https://en.wikipedia.org/wiki/Reinforcement_learning

The article is interesting but the historical overview seems dubious to me and quite biased to show that "originally" people already thought it was a physical thing.

What about ancient egypt for example? It has extensive philosophy about the afterlife and the "soul" yet is not mentioned at all.

The author states that:

  > Until recently there were really no very good analogies available

On the contrary, our theories of mind have mirrored our latest technology for thousands of years. First, a hydraulic model, with "humours" flowing throughout the body's plumbing; then a mechanical model with tiny machinery; then an electo-chemical model with charges and currents.

This article argues strongly that our present computer-based metaphor is likewise off-the-mark: https://aeon.co/essays/your-brain-does-not-process-informati...

> This article argues strongly that our present computer-based metaphor is likewise off-the-mark

Well, it is a metaphor, so, by definition, off the mark. The article only does well to hammer home the fact, as does the OP (at least, early on).

The metaphor eventually breaks down when you take it too far, too literally, otherwise it wouldn't be a metaphor. It will be the exact same thing.

The brain-as-hardware-mind-as-software is a useful model for dealing with the mind-body problem, and giving us a fair idea of how a virtual mind can exist in (on?) a physical brain.

Of course, you can't fault earlier ideas of thinking about the brain: it is only natural (and beneficial) to think of the mind in terms of the most complex man-made (or man-understood) technology currently available, without necessarily holding fast to it dogmatically when a better model is found.

Think of it as necessary stepping stones to the destination.

If anything, it is naive to think that the brain/mind works exactly like our current computing architecture. It is also naive and premature to completely banish the idea that the brain is some sort of computing mechanism.

Whenever this topic comes up in HN, a sometimes quite large fraction of the comments do make, or are predicated on, what seem to me to be unjustifiably specific analogies to current software technology (I guess that's no surprise.)

I imagine it's because they're implemented in PHP.

That's totally flawed thinking.

The brain is of course written in JavaScript.

Ah yes, I remember that article from a different HN discussion a few months ago. The notes about historical analogies are interesting, but I can't agree with any of the conclusions Epstein makes.

Here is a counter-article that I think sums up the problems pretty well: https://sergiograziosi.wordpress.com/2016/05/22/robert-epste...

As for the analogies, I think it's good to know the historical ones have existed and put the "brains are computers" metaphor in the right context. But Epstein seems to conclude "there already were lots of different analogies in the past, therfore all analogies are useless". I don't see how that sentiment gets us anywhere.

Wouldn't it be more useful to accept that we're always trying to explain the world in metaphors and find tools to separate the good metaphors from the bad ones?

> Wouldn't it be more useful to accept that we're always trying to explain the world in metaphors and find tools to separate the good metaphors from the bad ones?

Absolutely.

I completely disagree with Epstein's conclusion as it reads from the text. I'll give him the benefit of the doubt and interpret him as trying to steer clear of the oversimplification a satisfying analogy can encourage. In fact, he hammers on some of the ridiculous consequences of overenthusiastic embracing of this metaphor.

I wouldn't want to believe he doesn't understand, as you put it, that "we're always trying to explain the world in metaphors and find tools to separate the good metaphors from the bad ones".

I think it also highlight that different people - even experts from different fields - can have vastly different definitions for things that seem common - e.g. what a computer is or even what "information processing" entails.

So I could kind of understand him if he wanted to defend against people who heard "the brain is like a computer" and then jumpted to conclusions "oh, so the brain has a hard drive and a processor and can be programmed to do stuff"...

Exactly. This happens when formal definitions of these "common things" are absent, or ignored in the argument. He's genuinely right in calling out naive conclusions about the brain based on current computing.

I think this should have a [1992] tag on the title

Good point. It sounds very similar to the mind/brain stuff in Snow Crash.

I doubt that somebody's else mind (the program) would run well on my brain (the hw) and viceversa. Mind/brain are probably more like the first computers, programmable by manually reconfiguring their hardware. So, more like FPGA than general purpose CPUs. Interestingly we're moving in the direction of blurring the difference between hw and sw. Example: all the stuff that Apple is moving in custom made processors on their phones that once was run by the main CPU.

I'm curious as to why you think this

Because even if our brains share a common general design they grow and become different. Synapses connections are created or destroyed according to our personal history, the experiences we had etc. To upload your mind into my brain would mean to rewire my brain at least down to the synapses. There could be some excess neurons left or not enough neurons for you to be there. Not to talk about possible preexisting damages that would prevent recreating the exact structures that let you be you. Plus, all the connections with the rest of the body and the influence that my other systems (endocrine, hormones, etc) could have on the way your mind is allowed to perform. In the best case we should have to find an alternate configuration of my brain that would produce your mind. That configuration would be different from the configuration of your brain. I concede that's not impossible, but it's not how we're doing it with computers and I wonder how we're going to test that. Make the two copies talk together for a while until they agree they're the same person?

Compare that with computers which are built from standard parts. Every processor of the same model is the same down to the nanometer scale, every RAM, bus, HDD or SSD are the same. Reboot with a different OS or image and you get a totally different system. Boot two equal systems (because they can be equal) with the same OS and you get two exact copies. After all we've built general purpose computers to be general purpose, didn't we?

I'm curious as to why you think this

Reply on news.ycombinator.com