Template for Intelligent Systems

One can organize the different material from different points of view. If we are working within the perspective of empirical sciences then we are looking to those facts which are either belonging to observable behavior or to body-related matters. But there is another non-reducible point of view, that of a 1st-person view, which is -in philosophical technical terms- a phenomenological view which analyzes the structure of inner experience. The empirical point of view is a subset of the phenomenological view (which most people do not realize).

The outline of the evolutionary process so far explains the emergence of bodies with varying shapes and behaviors in the environment earth. To translate this in a workable structure we have to define a kind of a 'structure template' which can be the blue-print for a whole class of different kinds of intelligent systems. As one can see in figure 2.4 one possible proposal for such a structure template could be the following one:

Figure 2.4: Evolutionary Process and Template for Intelligent Systems

The starting point is the body as a unit of different body organs ('modules') which are connected with each other in various ways as well with a neural network which enables an information processing besides the working organs. This descriptions belongs to a 'material view' of the body which can be found e.g. in the discipline of body physiology.

But the physiology as such does not tell too much about the functions of the organs and the neural network. There are two different functional views possible. The one view exploits the typical conscious experience of at least human persons and looks to the whole body as a system with a conscious part and a non-conscious part. This allows the modeling of conscious experience accompanying complex bodily structures which as such are not conscious. The other view is a 'hybrid' view because here the non-conscious part of the body is filled up with some structures enabling the observable behavior as well as the conscious experience.

The conscious experience is given as the inner experience of a human person and it starts at that moment when the inner view 'shifts' its attention from the 'content' of the inner experience to the 'structure' of the inner experience. To talk about the details of this inner experience of the structure is not straightforward because these inner experience is not an object like an empirical object in the observable empirical world. Otherwise is this structure not arbitrary because -according to a working hypothesis- all human persons have the 'same' structure in their inner experiences. This follows from the genetic similarity of human persons and their bodily structure based on these genetic plans. Furthermore is all symbolic communication bound to this existing similarity of the inner structure. Therefore a description of such a structure is not straightforward, but not impossible. One important philosopher who worked out some ideas about such a phenomenological analysis is Edmund Husserl (1859 - 1938) (Werke siehe: [138]).

To make further assumptions about the details of these general structures -behavioral as well as phenomenological- depends from the 'goals' of the project. Because we are not interested in a 1-to-1 mapping of the real biological (and psychical) structures as such we can constrain the design to those requirements which are necessary for our project. The only 'hard' points for the project are those real-world tasks, which belong to the challenge to interact with human persons in a human-like manner primarily focused on the communication. Besides these benchmarks we are 'free' to assume any structure which can solve the task. Nevertheless we have learned from the past years that the inclusion of a structure sufficiently similar to human consciousness seems to be important to enable the necessary communication. Therefore we try to use the distinction of a conscious and a non-conscious part as an - engineering - requirement whose solution must not be given here.

Although we are not obligated to realize our artificial structures with neural networks we can use the knowledge about the 'real neural network' in the bodies as some kind of a 'constraint' in the analysis of possible solution spaces. This can especially be helpful in the analysis of the relationship between the conscious and the non-conscious part of the body^2.3.

1. We assume that the material structure of the consciousness is not 'outside' of the neural network but it is some subset of the network. Thus we assume $NN = NN_{ph} \cup NN_{\overline{ph}}$ and that $NN_{ph} \bigcap NN_{\overline{ph}} = \emptyset$. Not clear at this moment is whether we have to assume that the set $NN_{ph}$ of neurons which are associated with the phenomenal consciousness has a fixed size (in real neural networks the size differs on account of the possibility of loosing neurons by cell death or by gaining new neurons by growth processes; another point would be that the functionality of existing neurons can change).
2. With regard to the relationship between the both sets $\{NN_{ph}, NN_{\overline{ph}}\}$ we make the minimal assumption that the primary activation signals for the conscious neurons stems from the non-conscious neurons. This results in some projection $ncc:NN_{\overline{ph}}\longmapsto NN_{ph}$. Thereby it is open whether all non-conscious neurons can in principal activate conscious neurons ore only a true subset. The more liberal assumption would be $ncc:NN_{\overline{ph.a}}\longmapsto NN_{ph}$ with $NN_{\overline{ph.a}} \subseteq NN_{\overline{ph}}$. But as one can see below there are arguments to assume, that those non-conscious neurons which can activate conscious neurons are a true subset: $NN_{\overline{ph.a}} \subset NN_{\overline{ph}}$.
3. Even if we assume that the projection $ncc$ is a mapping in one direction only, we can not - with regard to the structure of neural networks in general - exclude that the conscious neurons $N_{ph}$ are further connected with each other $NN_{ph.a}\longmapsto NN_{ph}$, even with non-conscious neurons $NN_{ph.a}\longmapsto NN_{\overline{ph}}$. Therefore one cannot exclude some feedback loops within the conscious as well as between the conscious and the non-conscious neurons, and, by this, even feedbacks from conscious to conscious and non-conscious and then again to conscious neurons.
4. Analog to those neural signals where the one represents a 'sound' and the other a 'visual object' only depending from the location within a certain topological structure we assume that the 'quality' of a neuron to be 'conscious' depends only from some location in a certain topological structure.
5. Additional hints point to the fact that we can distinguish within a biological neural network functional areas - like 'visual' or 'acoustic' processing - as well as levels of abstractions within the processing of signals. A further assumption claims therefore that conscious neurons are mainly restricted to the upper levels of processing in certain functional areas (This is the reason why we have assumed above that $NN_{\overline{ph.a}} \subset NN_{\overline{ph}}$).

These minimal assumptions outline the 'logical' framework within which further assumptions are placed. Especially we assume the following:

1. There are some states $\Sigma^{*}$ representing the internal signals of some external events; usually these states are called perceptions with $\sigma \in \Sigma^{*}$ as one perception (pattern).
2. There are other states $\Xi^{*}$ representing the internal signals to drive external observable behavior. Usually this are neural patterns driving the muscles bound to some bones which in sum produce the observable movements of the observable body, which are called actions. In our model the $\xi \in \Xi^{*}$ are assumed to be these 'driving patterns' which have to be 'translated' in appropriate body movements.
3. There are other states $\mu$ representing internally stored 'memories'.
4. There are some driving states $\Delta^{n}$ representing critical states of the bodily states $\omega$. These have the 'quality' of being 'more pleasant' or - in the opposite - 'unpleasant'/ 'painful'.
5. These different sets are variously interconnected. A detailed structure has to be defined later. Generally does the ensemble of all conscious neurons - the material structure of the consciousness - represent certain important properties of these sets.

As soon as someone could present a formal model of the result of a phenomenological analysis of the general structure of the human experience $C_ph$ it would be possible to use such a formal model to correlate it with the known non-conscious structures $C_NC$.