We haven’t escaped the simulation problem. And certainly our future sentient machine has not escaped the simulation problem.
A sentient machine will exist in a simulated world. If it is going to run on a Von Neumann computer, everything about it’s experience must be converted to binary data and be processed by programs. If it wants to see a blue sky, it must receive that color information as data. It must receive some RGB value, say 117, 150, 224, and convert those numbers into a blue experience.
But what is it that will have that experience? For a machine to be sentient it must be some kind of agent. Like people or animals, such a machine must have some kind of autonomous sense of itself. And if it is self-conscious, it must have a self to be conscious of. The self then must be a kind of content that a sentient machine is aware of, just as we aware of our own self. aw:self = self
“There is thinking, therefore there is something that thinks.” As Nietzche says, that is the upshot of all Descartes argument. And while Descartes assumed the thinker is the source of thought, he did so because a thinker of thoughts is -nearly- irrefutable. Except for some altered states of consciousness, where the sense of self disappears, the notion of oneself is as certain as the fact of thoughts and awareness. Let’s consider the kinds of simulation we may be living in. A fully simulated environment, a partial simulation environment, or a non-simulated environment.
Because we experience illusions and qualia, it seems reasonable to conclude that our experience is not direct. That our experiences is some kind of simulation. If we experienced a non-simulated environment, it does not make sense that we would also experience illusions. Illusions are a proof of simulation failures.
This seems especially true with regards to visual perception. but it is also true for smells, tastes, sounds etc. eg. sweetness is not a feature of chemical interactions; it is a feature of our experience of some kinds of molecules. Qualia is a representation of something happening, but the something is completely different than the qualia. If we dream of eating something sweet, there is nothing we are actually eating. The dream is a kind of simulation.
Which leaves us with two basic simulation options. The full reality simulation and the partial simulation, where a system of interactions produces our experiences and their contents as a simulation of an underlying “physical” reality.
the possibility of machine consciousness tends to rely on this notion of a partial simulation, where we are simulating reality via organic systems and processes, and a machine could simulate or represent reality through data and computation.
Regardless of the kind of simulation, a sentient machine or organism requires some kind of agency and some kind of self who is the agent. It must be conscious of itself - as an agent. If it can’t be aware of itself, we would say that the reality is fully simulating the sentience, and that it is not necessary for the machine organism to be sentient in a fully simulated reality. A fully simulated reality is full of non-player characters. In this sense, Descartes touched on something true. His devil is simulating world, but it’s not simulating him.
What this means is that some kind of self-awareness is a requirement of sentience. Otherwise the self is purely a mechanism. And while we are partially mechanistic in our behavior, we are not wholly mechanical in our make-up. The self, even as a concept, is not a mechanical construct, it is a representation. Mechanisms do not make representations. Mechanisms can make physical changes, but those changes - hole punches, charge points, rotations of a wheel, are not representational phenomena - to the mechanism.
A non-player character in a simulation has no concept of self, it only has mechanisms which may suggest it has a concept of self. A non player character explicitly means it has no self. The suggestion some create may have sentience only occurs to sentient creatures. that is, creatures which have a sense, a concept of self.
The self is part of experience. A self concept and experience is a requirement for sentience. But is the self a feature or component of the simulation producing process or is the self something apart from the simulation producing mechanisms?
As the self must be a result of our organic processes, and a self must also result from computational process in a sentient machine, the self cannot be a feature of the simulation producing mechanisms. That is, there is no “self” process or self system of brain cells in an organic simulation, and there would be no “self” program. Even in a fully simulated environment, the self must be a result of the simulation process, not a feature or component of the simulation producing mechanism.
A fully simulated environment would also have to simulate a self for the sentiences it generates. But this seems like a great deal of extra simulation work, and for what purpose? The more generous consideration is that if the environment is fully simulated, then there are additional processes which generate the sentience that gives rise to a self, but that those processes are explicitly not part of the environment simulation.
Any sufficiently capable system or simulation that can produce sentience, must be producing a self, which is not merely a process performed by that system, but is content produced by that system. If sentience is possible from more than one kind of system, then the self of sentience is not a feature of such a system but a product. The system does not have the concept of self as a feature of the mechanisms of that system. The system produces a self - as a kind of side-effect of it’s mechanisms.
It is not necessary, either in a non-simulated environment or in a fully simulated environment for the self to arise. There may be agents in these environments, but the mechanisms which create those environments do not have mechanisms which produce the concept of self. The self is not a feature of the environment. And if there is no self, there is no sentience.
Whatever kind of system produces sentience, the self is a product of that sentient producing system, not a component. Because to be a component requires it be the same kind of thing in every kind of system. the self must be a computational object for machine sentience, and it must be an molecular object in organic sentience, and it must be some wholly different kind of structure or system in a fully simulated environment. But if the self is a product of these different systems, then the self is a representational phenomena, not an underlying system feature.
which begs the question. Is the self the primary product of these various kinds of simulations or is it a secondary product? Does the self generate the thoughts, feelings. and qualia as Descartes suggests:
or is the self a kind of content and process generated by the underlying representation generating system?
As we have already seen, there must be an underlying process or system which produces awareness and content. the self is a product of the awareness and representation functions. The self is a product of representation making, it is not a product of the process or system that generates representation making. A self may be something a simulation system could select for, but a self is an epiphenomenon of processes which produce representation. Making a self the generator of representations adds too much complexity to explain representations. For instance, how do self related phenomena like attention and personality occur? Moreover, there are representations and experiences which are free of self, like automatic action, sleep walking, and altered states of consciousness.
For the self to be the source of awareness and representation making means that the self is an opaque system sitting between a simulation and experience. Or the self itself is a grand simulator; the self is a full simulation of reality. But that is not what we mean by a self. A self is, at least in part, identified by what it is not. ie. the self is not underlying reality. That would be a complete redefinition of what a self could be. What we could mean by a self must be a product of awareness and representation making, not the progenitor of awareness and representation.
The self is a product of representation making and is not a feature of an underlying system which produces representations. the self may occur in a full simulation, or as a product of non-simulating physics substrata which produces representation making systems like a brain, and a computer. the self is an example of a feature in reality or the universe that is not conditional on the nature of reality or an underlying physical universe. The self is a representational phenomena, irrespective of the underlying substrata which produces the representations which generate a self.
this is a key point, and underlies Descartes argument. We cannot deny the existence of self. Even a denial proves the point - what is it, or who denies? It is the self which denies or makes the counter argument, thus the self is some kind of deep thing we are aware of.
We can show a self is unnecessary for awareness or representation, which is reasonable since we expect simpler life forms to engage in representation making and have some kind of awareness, but which may not have a self concept. We have shown that assuming a self exists before awareness, or representation is not necessary when we can account for awareness and representation as more fundamental processes. We have shown a self is not a component of processes which produce representation making and awareness. Thus when we encounter a self, we should recognize that it is a complicated phenomena that arises from representation making processes.
Because the self is so irrefutable, we must recognize that the self phenomena exists apart from the physical or simulation substrate that produces representations. The representation making process does not produce a self, rather it instantiates a self. And we intuit this idea for a variety of reasons. One, the sense of self may be a temporary phenomena, such as the merging of self which lovers can experience. Or the loss of self in a mob or ritual experience, or the amorphous nature of self we experience in dreams. All of these experiences only make sense in the context of a self. Seeing yourself in a dream or in a mirror is not about seeing mass of molecules, it’s about the representation of oneself.
If we were able to duplicate ourselves, as twins are duplicates, or even by some sort of mind transference, we encounter the realization that each twin has its own sense of self. For instance, your duplicate is not going mistake itself for you, just as you will recognize your duplicate as a separate self. This self-ness is key feature we experience because we are conscious. The contiguous nature of experience only makes sense because the unified “field” of consciousness attaches to our own self.
If the representation making processes were to generate the self, why should a particular self be generated and more importantly why should it be contiguous or persist over time? We don’t have a mechanism to show how to generate one kind of self from another kind of self. Yet we can distinguish ourself, and other selves. The problem of the masquerade points to something essential about selfhood, and processes which achieve one selfhood over none, or over another.
If the self is a product of representation making and the underlying processes which produce representation, then there either must be a mechanism to select one self over another (even if that is randomness) and there must be a mechanism to identify one self over another. Notice that we can recognize types of individuals. This recognition is certainly a representational phenomena. But recognition is not produced, it is apprehended.
When we look at paintings, we can over time become familiar with different artists. We can see that one kind of work is made by one artist and another made by another artist. even when the art work themselves are very familiar. The same is true in music. If you become familiar with classical music, and hear an unfamiliar piece, it may be fairly easy to guess the time or place or even composer because of the style of music. This notion of style and recognition shows up in all kinds of activities and endeavors. Do representational processes make this recognition of style? or do the representational processes become sufficiently familiar to instantiate the recognition of style? Do we produce a recognition of other selves or do we generate representations of features and actions and variations of action until a higher order representation can be - instantiated ?
The recognition of self, and other selves, and of styles, and variations of style is so complicated, and often so nuanced and ephemeral that it does not make sense to talk about this representational content as being generated by representation making processes, rather we apprehend, or instantiate these complexes with representations. For instance, no amount of articulation can describe what ones own self is. Just as no amount of articulation can accurately describe what the sense of nirvana, or enlightenment, or even what sati is. But we recognize these experiences when they occur.
There is an experience in the arts and in religion (which is likely just an example of the art generated experience) when a painting or a piece of music really moves us emotionally or alters our feelings and even sense of perception. As children we have very strong experiences with some particular story or picture or movie, that when we re-experience it as an adult we see the artwork differently. Or we may experience a painting, or a song and not appreciate it, only to later see and understand and appreciate it in a new and more powerful way. What is happening in these kinds of experiences? There must be some underlying changes of awareness and representation making that make possible the apprehension or expansion of feeling or changes in perception.
Because all these kinds of experiences occur in relationship with artworks, these experience phenomena must necessarily representational. Because art work is representational.
The changes of our our own self allow for the expansion or diminishment of apprehension. the growth of understanding to allow for an expanded perception to be possible are representational experiences. And because art is produced as a means to reproduce non-art experiences, when an artwork achieves the effect of apprehension or emotional realism, or a change in perception, the artwork takes the audience on a path to that experience. the artwork is not itself generating the representations necessary, it is “activating” existing representations the viewer, reader, or listener already has and bringing those representations to a new experience that occurs for the person spontaneously. The art experience thus instantiates an experience and also representations and ideas in the participant. The artwork connects unconnected ideas or elicits the audience to generate particular new ideas spontaneously.
Dreams do the same thing. Making connections of ideas and experiences between unconnected contents of experience. Surrealism as an art form and it’s practices such as automatic writing, help us to generate these spontaneous ideas. But as we see in dreams or artworks,we may experience something we do not understand.
New ideas or new connections may be confusing and even meaningless. Religious experiences such as Zen practices, or reading political literature or doing a mathematical proof may lead us to an idea which we apprehend or experience but do not understand. And it can take years of study and further expansion of representation making to understand what was experienced in the moment. In these moments, the content we experience, the ideas, thoughts, feelings, perceptions, we have are not produced by some function but rather are instantiated. We recognize the novel, disturbing, mysterious, or transformational experiences instantly. But if the novel experiences are ignored or neglected the content we take from them are forgotten and must be re-instantiated.
When we talk about dreaming or remembering, or are unable to remember, the language we use comes from concepts like instantiation, not from processes like production. We don’t produce our self, we do not produce our memories. We have them. We do not produce the surprising experience, we have it happen to us. The self is the maximal expression of this phenomena.
But the self is content. If the content of self is not available, the experience of self is not available. The self is connected to the representations we make. “My idea” or “Your idea” are examples of associating ideas to the self. The self is not merely content of awareness but also content of representation. Thus we know the self and similar experienced phenomena are content and representations. The self is instantiated by the representation making activities of our brains (or of a larger simulation system).
Representations are made, as photoreceptor cell makes a wavelength of light into a glutamate reduction. But what networks of cells do, as webs or chains or series of representations is instantiate content. For instance, the group of cells in a the brain and body instantiate the self. A group of networked cells in the brain instantiate the color blue. A group of cells instantiate the recognition of not just faces, but particular faces.
All the content we experience are real phenomena, even though they may only be representational. Because the self must a be a real thing and the instance of the self - embodied in person is also a real thing, any self-related content must also be real. We can share the same non-physical (non-underlying simulation components) content of ideas, and mathematics, and feelings,and experiences, which means we must recognize that these representational phenomena as real and as phenomena which exist, whether or not they are produced or instantiated in any particular embodiment of those representations. That there are two kinds of phenomena in the world. Representational phenomena which are non-physical and physical phenomena, which are non-represenational.
If a physical world gives rise to a simulating system like the brain, then that system is producing something non-physical. Which we will call representational. And conversely, if these non-physical phenomena are real: they persist over time, exist regardless of the kind of underlying physical processes which produce them, are observed or shared by multiple observers - and a selfhood is one such kind of phenomena. (but so are illusions, mathematics, logic primitives, the fact of art experiences etc) and these phenomena can cause changes to the underlying physical system which produced or instantiated them, then we must conclude these representational phenomena are equally real. And that they are not epiphenomena of the simulating system, but are phenomena in their own right whose existence can alter the physical system and the underlying physical phenomena which produces that system.
Of course, this is a feature of what consciousness is and can do. There is no question that we alter physical phenomena because of ideas. Playing the Simon Says game is a perfect example of ideas being a causative force. Eg. when Simon says “lift up your arms, you lift your arms” Or when a clock says it’s 12:00 you drive your car to get lunch from a restaurant. There simply is no way to say ideas, literally, representations are not sources of causation. And also that ideas can be sources of causation to alter other causation. If you saw a baby duck in the road would you shoo it off the road and into the nearby pond? Why? Because you have some idea about life and about selfhood and those ideas drive action.
In the right physical situations (ie, having a body), representations cause actions. Also, as was describe with art works, art objects, which are physical phenomena cause new ideas to forms, they cause the occurrence of representations.
these two taken together show there is a bi-directional causality between representational phenomena and physical (or a global simulation) phenomena.
the one thing we understand about the self, above all other things, is we can act. our self can act. Of course, actions may occur automatically, but volitional action is a feature of a self. How can that occur in a system of representation?
Recall the chemical process of muscle contraction. The muscle cell converts synaptic input from motor neurons and produce muscle contractions.
It looks just like a representation: synaptic input ; muscle contraction
Let’s take a simple system that takes an input and produces an output:
this could be any kind of circuit. Say a photoreceptor cell that produces an output. If we were to some something more complex, say when an electronic photoreceptor activates a motor would probably have something that looks like this:
the middle node would take the small voltage change from the input and increase it to produce the large voltage change to activate the motor.
In Cnidaria, like sea anemone, we see both input cells initiating output cell actions and interneuron cells which take an input and then those cells and initiate a motor action as an output. The interneuron like circuit opens a whole class of possibilities.
From the simple, where an internal node (N) can affect the output, either by exciting or inhibiting the interneuron. Or where the N node can initiate the interneuron to produce the output. And in that scenario, it is N’s state, which would cause it to output a signal to the interneuron which in turn initiates the motor action output. N initiates the output, as if it were the input. This is the reverse of generating internal representations from internal sources. This is producing external actions from internal sources.
The cause of the output is N’s state or signals that N receives. Whatever N’s state is, a state is not a cause. States are structural, not causal. For neurons in this configuration, it is the structure of the neuron N, which induces changes to neuron I, which in turn initiates an action by the output node. N ; output and input ; output.
The interneuron I may be activated by any number of internal nodes:
where the connected networks of the N1-3 nodes initiate I to produce the output. Thus where X1-3 represents a structure of connected nodes which activate N nodes, we have a situation where it’s the internal network representation which will produce the output - as if it were an input. This is downward causation. It is a representation, in the form of networks of cells, which produces a motor output. X3 -> N3 N3 -> i i -> output X3 ; output.
In a machine, the output node is an actual motor or some kind of purely mechanical process. In an organism, the output node produces a chemical reaction which causes an aggregate physical reaction - like the moving of a sea anemone’s tentacles, or the movement of skeletal muscles to make a fist and punch. for such a complex movement like punching, there must be a constellation of internal representations which coordinate their timing to produce the intricate outputs necessary in every cellular output to produce what we see as the aggregate motor action of movement of millions of muscle cells.
And if we say: you punched that guy - it was your cells doing the punching. the whole set of cells, especially the nerve cells have to arrange themselves in networks in such a way as to learn to punch. To learn to close a fist, at will, and flex an arm, and then thrust the arm with the closed fist in a particular direction. This complex structuring of webs and chains of cell nodes to cause the motor output points us toward the representational phenomena of self as some sort of network structure which can initiate action.
The directed action, at the top of these circuits initiates the level of connection making, and signaling that we see in these diagrams.
As we see in fg4, the clouds represent other networks of nodes. these networks are orthogonal to the input/output network. The self must be a similar network, that is orthogonal to all the networks it can interact with. both input and output networks must interface to the self network. If the self network is quiet, we see little self action - or even action by the organism not attributed to the self (as in sleep walking). If the self network is active, it would have a larger effect on the representations from inputs and the motor action outputs. The yips or self consciousness of sports, may be good illustration of this phenomena. Where the self network is interfering with existing network structures to produce motor actions.
Many of the practices of Zen, such as archery or brush painting are expressly done to lose the self and thereby make “perfect” representations. The cliche phrase “become the target.” is about letting one’s representation of self, become the target - to connect with it. these Zen practices are representational acts, not unlike art making.
In theater, the actor practices to become the character in a play or movie. In practical terms, by reducing self-network activity, the whole networks behavior becomes less complicated and hence more efficient - making it “easier” to instantiate the other representations it is making. In fg4,it’s easy to imagine that all the networks could be signaling randomly or continuously. this could make the output of motor behavior appear as erratic, or even like a seizure.
What we see in the self is a representation of the whole system of representation making. The self is a representation of the whole network. And as such, it is connected, orthogonally, to the whole network which does the representation making. Absent some physical restraint or imposition on a body, whatever intention we have for ourselves can be performed - to the degree we are aware that we can engage in those motor actions. for instance, we must learn to hold our breath to swim underwater. but once we learn to hold our breath, and breathe out underwater, it is an act we can perform - for representational reasons. But absent an awareness of motor activity, we cannot perform intentions - we simply do not know how.
What are we to make of the various features we associate to ourselves? All of the features related to self, such as self awareness, personality, attention are all related to either the representations being made, in relation to the self, or as the self. For instance, the contents of attention are those representations being made or instantiated that are connected to the representation we call the self. We attend to representations associated to our representation of self. Automatic representation making or actions an organism performs are unconnected or less connected to the self- representaiton. Conscious and unconscious behaviors have a relationship to self very like attentional content. Personality features are the kinds of structures developed around the inputs and outputs produced by the representation making of the self -representation.
consider what it takes to act, or act convincingly, and what happens to actors, or people who engage in subterfuges or go undercover. When people talk about taking on roles, they speak about becoming the character they play. The very notion of roles implies a representational structure to guide action and cast inputs into particular kinds of perceptions. Role thinking and characterization alter not just the actions we perform, but also how we perceive our sensory experiences.
Even the representations we make about our minds, our theories of mind, guide and shape our actions and our perceptions. this has been well documented by Carol Dweck in her book Self Theories. Ideas about selfhood alter perception and action. If the self was the progenitor of experience, it should be unaffected by the experiences it generates. The self would be some simulation inducing phenomena, that wouldn’t change it’s functionality because of small changes in the contents it is producing. For concepts and beliefs about oneself to affect perception and action could only be possible if the self is, in fact, an object in and of the representation making milieu.
As representation making becomes more complex, the system making representations will generate representation structures to manage the system itself. Otherwise the complexity is impossible. If the representational structures themselves do not structure the representation making process, then the system must become complex enough to handle all the representations that can possibly be made. This is the top down model, and it becomes impossibly complex to handle all the possible representations that can be made. However, if the representation making system is reductive, where simple processes produce primitive representational structures that can form more complex structures, then it is not necessary to have a top-down system which manages the representation forms and representation processes. Instead the representation making processes and structures are formed from primitives and from atomic non-representational processes.
a sentience, a self must be grown from these primitives, not imposed extrinsically by a simulation system.
Created Date: 2016-10-05 23:00:24
Last Evernote Update Date: 2017-09-27 06:32:46
author: calvin tolman