Consciousness is a recurrent neural network
Let us begin with an example of seeing a dog: when you see a dog, photons from the dog come to your eyes and converted to neural signal, then, a neural network will recognize that what you're seeing is a dog.
How do you know that you're seeing a dog?
I believe we can explain this in the same way as seeing a dog. But in this case, what you see is not photons but signals from within brain itself. In other words, brain takes its current states (signals) as its input (this is possibly no different from brain takes signals created by your eyes as its input.) And as you may already known, this is similar to recurrent neural network.
In this way, consciousness is a concept learned by brain about the inner processes of brain itself which is similar to: dog is a concept we learned when seeing many dogs!
Stream of consciousness is the result when brain trying to model itself and continuously receiving signals from itself!
We can put this further and ask: is any brain of any kind (dog, cat, whale) conscious?
My idea is that there is a threshold when brain becomes complicated enough to model its own signal!
49 comments
[ 3.1 ms ] story [ 110 ms ] threadHow do you define conciousness? Like the voice in your head? Are you saying that voice is essentially the predicted copy of you I.e. what you predict you would do in the current situation? I think this could explain why we have a concious and unconscious mind, the unconscious is the actual brain, the conscious is just our prediction of what we would do in the current situation.
if this is not a "stoner philosophy" on consciousness then what is?
And this theory fails to explain how we can feel pleasure or pain, and, it fails to predict whether an artificial neural network can feel pleasure or pain.
Pleasure and pain are related to reward signalling in reinforcement learning. Humans have inborn reward signals which have been selected by evolution. Artificial agents have reward signals too, but they are designed.
So I think the path to consciousness is to create artificial agents that learn to survive and adapt. When such a complex behavioral system is designed, it naturally learns about its relation to the world. It has consciousness by extension of being capable to protect its life.
This is what I'm interested in, but what the theory presented here fails to explain.
It would also need to be embodied and responsible for its own survival, more generally, otherwise it's just a computer running software, in a lack of external context.
some things to remember:
- neurons are not linear.
- neurons are not time-invariant.
- neurons are not causal.
- neurons and actual anatomical connectivity do not map well to electronics-inspired wiring diagrams. (see: dendritic arbor)
- anatomical structure does not imply functional connectivity.
- functional connectivity does not imply anatomy!
- synaptic junctions respond more or less well to different neurotransmitters, all of which are continuously present, at different times. We don't know why.
- So far, what we have observed about anatomy and physiology of the brain does not look like a RNN.
- The relationship between the meat architecture and the phenomena it hosts may not be as 1:1 as you'd like. Opinions vary.
My favorite analogy about relating brains to consciousness is this: "if you believe that brains are like computers, (which you shouldn't, but just for the sake of argument, let's), then even if you really and truly produce a full map of the brain, what you've _got_ is the spec sheet for an x86 processor. What you _want_ is the user manual for Mac OS."
What does that mean?
- repeatedly presenting the same input to a neuron has a response that depends on neurotransmitter reuptake rates. If the synapse isn't "recovered", you get a different spike rate out.
- neurons fire stochastically in the absence of stimulus. Responses to small stimuli are indistinguishable from noise.
- The above two phenomena propagate through connected series of neurons in a nonlinear way.
I think "the kernel's source code" would be a better analogy. For all I know the Mac OS manual is probably silly and explains how to turn it on and off, and how to go to appstore and itunes.
Why not? Isn't the brain a biological machine?
I highly recommend to read Godel Escher Bach(if you dont have time - just read the introduction to get the general idea). In this book author explains how the meaning arises when things(like language, or math equations, or neurons in the brain) "mirror" things in the real world, when their structure can be "mapped" onto some other structure(so called "isomorphism").
He says that brain "mirrors" the world around it as it builds a world model. But the brain itself is a part of the world, so it builds the model of itself as well. Neurons recognizing/observing/experiencing other neurons.
Just like you can see a dog you can "see" your own brain state.
I've also heard a cool quote somewhere - "Consciousness is simply just what it feels like to have a brain". You can close your eyes and feel the position of your body, you can feel your stomach being full, and you can feel your brain thinking.
I don't think you'll get any real sense of what GEB is about just by reading the introduction.
But you're right, the whole book does contain more information than it's first few pages. No shit.
There's no reason for a brain of a dog or a mouse to not be able to observe/experience itself. It's model is probably simpler, but I bet it exists.
Like before you have learned how to code, your brain dedicated only a tiny amount of neurons to represent the concept of "programming", or "building a website". But as you gain experience, the model of this stuff becomes more complex. I imagine that brains that have less neurons can still recognize their own signals("be coonscious") , but their map of their brains would have a "smaller resolution".
----
As I grow older and learn things, I notice that I get better at "meta" thinking, that is having a map/model of my own thought processes. When I'm trying to figure something out, I'm also paying attention to the mental operations my brain goes through as it tries to figure something out.
The perfect example of it is writing fiction. When you begin learning to write, and are just trying to type a certain amount of words per day(to practice), you are doing it without any "self awareness", your brain generates words/ideas, and that's all there is to it, you don't know how.
But as you get better, and pay attention to what your brain does as you write, you develop a neural network that represents the process of writing, the operations your brain needs to go through to craft a story. Because in the case of writing fiction this model is complex, it is really easy to notice. Like I can feel the neural network in my brain that repressnts the programming concepts that I've learned, I can also feel the neural network that represents mental operations I need to go through to generate a story.
Good writers have a more complex and "correct" map of mental processes they need, so they can competently craft stories any time they want. But you can also learn to write well without "meta" thinking. Then your brain just strengthens the correct mental pathways as you practice(without "meta" thinking), so you can be good at writing without understanding how.
I think that this is a really cool example, where you can directly experience a brain building a map of itself and then using it.
I think there is a general trend towards thinking of consciousness as an emergent phenomenon which exists when a system is complex enough and has certain attributes, such as memory, information processing capability, etc.
But none of these definitions enlighten me as to when or why such a system might become self-aware.
Perhaps there is no "why" of consciousness. There doesn't need to be a reason for it. If we take evolution as a given, properties of organisms arise either because having that property makes you more likely to survive in some environment, or as a side effect of other properties.
We like to think that consciousness is special, because it "feels" special. But it's likely just some consequence of complexity.
As for when it occurs, i think there's a bit too much black and white thinking going on there. For one, most people probably wrongly assume that humans are fully self aware. If you spend a minute thinking about it, you'll realise how little we're actually aware of in ourselves. Therefore we're on some spectrum of self awareness. "Less conscious" beings would feel a little less self aware, and "more conscious" beings would be more self aware. It's unlikely there's a defined point where you "become" self aware.
To illustrate the point. When would you say you are "drunk"? What's the right definition of drunk? Is it when you start to feel any effect at all? Is it when you have the first sip? Or is it when you can no longer walk? Most people usually think of such words as "tipsy", which would come some time before drunk. Or maybe even "blackout" which is associated with drunk but is usually somehow more so.
What this makes clear is that there's less a problem with working out "what is consciousness and when does it happen", and more a problem of how we have evolved to discuss concepts.
This is why we're only just coming to terms with the non-binary nature of many concepts. It's not innate in us to recognise gradations. It's innate to draw lines between things and categorise them. But that doesn't mean that such classification applies so well to the universe.
To me, to be 'conscious' is to be able to make decisions. Presented with all our sensory data, memories, knowledge etc the brain creates possible models of the future every millisecond, and passes them to the conscious component which is where we choose how to proceed.
After a time, it doesn't even have to ask us anymore as it just replays past decisions. For example walking, driving a car, reading words etc
To give the simplest example: you can model a color in many different encodings, 'red', 'hsl(0, 50%, 50%)', 'rgb(255, 0, 0)', etc.
Any "thing" can be encoded in an infinitely different number of ways. The same applies to the state of your mind. Why should one encoding give rise to consciousness?
Arguably, the manner in which water is flowing within a sewage pipe network could be interpreted to be encoding some kind of information. Would you argue that some particular arrangement of water flow within a pipe network could give rise to consciousness?
I think there's a difference between a system modelling itself and becoming conscious of itself.
Any system can model itself. We can do it easily with computers. Arguably the linux kernel has a model of itself, its hardware, its inputs, etc. That doesn't make it conscious.
Sure, some computer programs model specific things quite well. But these models are very limited, fixed, and useless without human interpretation.
What model _isn't_ useless without human interpretation?
Even the models of the world in our own brains are not fully known to us, and we largely make use of them without interpretation.
Basically, any model that wasn't created by humans does not need human interpretation.
Those who think there's not a "hard problem of consciousness" or hand-wave it away with purely materialist explanations probably aren't conscious.
Folks. We have p-zombies in our midst...
(Sips tea.)
+ You have forward-inverse models e.g. by Wolpert.
+ You have a sequential winner take all process, e.g. see Baars.
+ You have honing in on the on/off switch. Search for claustrum and Francis Crick.
The challenge of course is to know what the brain knows and learns about itself. Oscillations at an alpha, beta, gamma level have as far as I have seen no place in current networks. I find it suspicious that we don't reproduce this behaviour. Are we sure that it is nonfunctional?
I'd like to posit a progression of animal awareness. (In the full knowledge that there is no "tree" or "hierarchy" of evolution; the progression is merely a convenient way of presenting some data.)
1. Single-celled animals, such as amoebae and /Paramecium/. Many of these display simple taxic responses: they move towards light, away from heat, and towards or away from certain chemicals - they pursue concentration gradients. In other words, a single cell can display what could be called "voluntary" movement; it does not follow programmed paths but responds to its environment. You can watch a Paramecium in a microscope, swimming through a world of bits of plant and mineral matter in water. If they bumble into something, they recoil, and set off in another direction. If they catch a scent of something that might be food, they change direction and set off in pursuit of it. It's much like watching a much bigger animal, like a mouse, explore an unfamiliar environment. Surprisingly like.
Similar behaviours can be observed in all sorts of small animals, like collembolans and nematodes.
Small animals - even single-celled ones - interact with their environment, responding to stimuli in ways that are more than a simple, determinate pattern. They are not like a clockwork mouse or toy that always follows the same path.
Woodlice have enough "brain" to at some level form a model of their surroundings. They can learn a very simple map.
3.Many small invertebrates form social colonies. In land-living ones, like ants and termites, animals leave a chemical trail as they move around, depositing markers indicating desirable discoveries. This is how an ant colony "finds" sugar in our kitchens, say. One randomly-exploring ant finds sugar, and retraces its steps home leaving a marker indicating that it's found food. Other ants follow the trail, reinforcing it, until a path is marked for the colony to move large amounts of food back to the nest.
In other words, ants make and read signals for other ants.
In other words, bees have symbolic communication. Their signals are not direct, one-to-one, follow this to the food; they are abstract and require interpretation. Obviously, this is innate, it is not learned behaviour, but individual insects learn the way to food and communicate this symbolically to other individuals.
5. Small mammals can be taught complex behaviours in the lab. Rats can be taught to press levers to obtain treats; mice can be taught to run mazes; "experiments" with wild squirrels have shown that the animals will perform an amazing array of actions to obtain a food reward. Many of these patterns of behaviour show that the animal is able to learn that unrelated actions can elicit a reward, showing some kind of understanding of cause and effect.
6. Non-mammalian vertebrates perform all sorts of ritualistic behaviour that is not directly related to finding food, a mate, predator avoidance and so on. Many birds have mating rituals; male bower birds construct huge and elaborate structures to tempt females into mating, which are presumably derived from nest-building behaviour, but actually the structures serve no purpose other than to be on some level "pleasing" to the females. Many migratory birds such as albatrosses and swans form long-term pairs which persist over years, sometimes lifelong. They can recognise their mate and when they rendezvous at breeding time indulge in long rituals which appear to stimulate and reinforce the pair bond. Obviously mate recognition does not involve some simple biochemical cue of kinship which might be used in a mother finding her offspring in a large, mobile breeding colony independent of environmental cues, such as penguins. Maybe penguins can "smell" or "taste" their own offspring by some form of genetic resemblance; I'm not aware of any research into this. However, even if they can, and don't somehow just learn what their offspring look or sound like, the same cannot apply to individuals recognising their mates, who are, by choice, usually not directly related.
In other words, these birds "know" their non-related partners, can recognise them and distinguish them from strangers, and behave in entirely different manners around their partners to around other individuals. When encountering their partners for the first time after a protracted absence, they indulge in displays and other complex behaviours which are not directly survival-related. It is hard to watch such a pair being reunited without thinking that they "feel happy" to see one another. It is also a reasonably common experience to see one which is isolated and has lost its partner, witness its lack of animation, listless behaviour and so on, and the human reaction is to interpret this as the animal "feeling sad".
8. Outside of the remit of pet animals in human company, though, mammals which live in social groups form a complex of often hierarchical relationships: dominant and submissive members of the group and so on. They perform patterned behaviours where different members have different roles; hunting animals such as female lions and wolves, for instance, cooperate, so that hunting parties contain scouts, flushers, chasers and so on. Many group predators display patterns of behaviour such as setting up ambushes. These roles are not fixed but are interchangeable between members of the pack. This demonstrates that the animals not only know of each others' existence, but of their relationships, since for instance in felines often only relatives cooperate. Elephants, wolf packs and so on may comprise unrelated individuals, though. Particular members of the group have expectations of the ways that others will act; cooperation must be learned, and members that do not cooperate may be "punished" by biting or by withdrawal of food. They also perform planned activities; ambushes or driving prey towards a pre-placed fellow pack-member indicates some form of awareness of the future. One cannot plan if one does not remember past events and strives to re-create things that have worked before; this indicates an awareness of time and of modelling the behaviours of other animals, so that there are expected, desired behaviours and unexpected, undesired ones, both in fellow pack members and in the prey animals. This implies a considerable degree of ability to form and maintain mental models of the behaviours of unrelated individuals.
9. Many animals have been shown to make and use tools. Not only chimpanzees in the wild, or finches in Gibraltar which use thorns as probes and levers to get at otherwise-inaccessible food items. Crows have been experimentally demonstrated to be able to improvise tools from available objects to get at food items. In other words, tool use is not always inherited behaviour or mimicry of others; the invention of novel tools has been demonstrated, outside of the mammals.
10. Again in the Aves, recently, a well-known experimental African Grey parrot named Alex died. Alex had been taught human speech; he was able to identify a wide range of objects and colours by name in English, to count up to five or more, to understand simple questions in English and formulate novel answers in the same spoken language. He was able to spell simple words out phonetically - "nuh, uh, tuh, NUT". He was able to express desires in spoken words: "wanna nut, now", even when this did not form part of the experimental dialogue. Many instances have show that his utterances were not simply mimicry of those of his trainers; he was able to construct sentences of his own, as well as parse those of his trainers. This shows considerable proficiency in manipulating an alien to him form of symbolic communication, very considerably exceeding the abilities of chimps and gorillas in the use of sign language. Video clips are available of conversations with Alex; they compel most observers to complete re-assess the presumed levels of intelligence of a bird with a brain the size of a walnut, orders of magnitude smaller than a human's brain.
12. As previously discussed, chimps have been observed to lie, meaning that chimps are not only able to model the behaviours of other chimps in their troupe, they also model the mental state of those others. This is not to say that a butterfly with eyespots on its wings is consciously "lying" to predators, but when a more complex animal such as a chimp gives false information to other chimps, I think that what it's doing is certainly trying to manipulate another's mind, implying that it knows it has a mind.
What I'm trying to demonstrate here is that there is a fairly simple, steady, observable and demonstrable increase in the sophistication of animal awareness of the world. Few aspects of human cognition are unique to humans; just about everything we do except writing - a recent human innovation, not an evolutionary one - various animals do too. Animals can be shown to possess and perform just about every mental trick that we do, from symbolic manipulation to abstract thought. Cognition is not a uniquely human behaviour and neither is self-awareness. We're just better at it. It's a difference of degree, not of kind.
Now, this being so - and I think it is unarguable, but I welcome attempts - and the basic aspects of stimulus/response being readily demonstrable right down to single cells, what I want to ask is this:
Where is the step from simple reflex action to perception/thought/response?
Even in humans, functional NMRI has shown that the cerebral impulses governing physical actions arise before the conscious mind is aware of them. Whereas we do undoubtedly reason things out and act on them, in much of the basic action of the human brain, the conscious mind is merely a spectator, watching what's going on "beneath" it and then rationalising after the event that it "decided" to do that.
Thinking is not, I submit, some special event in the brain. It's merely a slightly more sophisticated version of the very simple environmental modelling that even small crustaceans like woodlice do. Right down at the level of animals that have no brain, merely a small loop of nerve tissue around the mouth with more ganglia than elsewhere, animals take a step back from simple direct-wired stimulus->response, filter the incoming signals, form a model of what's going on, and act upon it. This, I submit, is the simplest kind of "mind", and the difference between it and us is that we have an awful lot more neurons and much more complex neural networks in between "in" and "out". It is a difference of degree, not of kind. Purely quantitative, not qualitative.
A woodlouse "sees" in exactly the same way as we do. There's no deep difference. Many insects and birds and fish see colour better than we primates; they can see more colours, more differences over a greater range. The bigger the brain, the more complex the pattern-analysis; the bigger the patterns that can be identified. What happens, though, is still the same: a sensor detects a stimulus, sends an action potential down an axon to a ganglion, where it triggers a cascade of other action potentials that propagate across a network of neurons until they either elicit a response or not.
The difference is that in humans, the cascades are bigger than they are in other animals, except whales, dolphins, elephants and the like. In at least some of the great apes - chimps and orangs - some of the impulses originate ...