1. Two robust predictions

Two predictions can be based on a conservative approach that uses only major transitions. Major transitions are a robust basis for extrapolations because each time the multistage is reached, a major innovation is required for letting evolution continue via the creation of a new system type. For this reason, the iteration between primary operators and their multistage forms a robust principle on which to base predictions.

Focusing on major transitions only, the operator hierarchy can be summarised as is shown in the following figure.

Figure

From this starting point, two predictions about future system types are available. The first prediction is that some day systems will evolve which show a multistage that is based on elements showing neural network architecture. The second prediction is that this multistage allows for a new cyclic interaction, which will lead to the next hypercyclic interaction forming the basis of the next evolutionary level.

As the above two predictions are strictly based on the most fundamental aspects of the operator hierarchy, there is a large probability that the prediction of a future multistage is correct. But how much information is gained with such a prediction? In fact, the information is limited, because it is based on limited knowledge of the specific properties of the memons that will form the multistage. A most serious error, which is easy to make at this stage, is to start considering how, for example, human brains, as the best known neural network type of considerable complexity, can be imagined to function as a multistage. To do this would deny the possibility of stages in between of any newly formed building block and its multistage. This may not seem problematic for the step from the atom to the multi-atom stage, because this step on forehand leaves no room for additional system types. But serious problems arise for cells, which may be prokaryotic or eukaryotic, each creating its proper types of multicellularity. Prokaryotes have given rise to multicellular blue-green algae. Eukaryotes have formed fungi, plants and animals. There is a world of difference between the intercellular communication in blue-green algae on the one hand and that of fungi, plants and animals on the other. The lesson from this is that the complexity of the elements has a major influence on the potential complexity of the multistage. The assumption, that human neural networks would be the building block for the multistage leads to the imagining of a multistage with ridiculously primitive properties. The error would be similar to predicting multicellular life that is based on prokaryotic units only. As will be shown in the following text, the solution to this problem lies in the recognition of the other steps of complexity increase that the operator theory helps to recognise between any newly emerged operator and it's multistage.

2. The hardwired memon and its multistage

The most straightforward detailed prediction is that the hardwired memon develops directly to a multistage. Such a multistage would have limited prospects for becoming of any evolutionary importance. The reason is that the transition to this multistage will be difficult and slow, especially for memons based on cells. This is caused by two major drawbacks of cellular hardwired systems. The first drawback is that genes code for the structure and quality of cellular neural networks that, for this reason, can only evolve over many generations, via reproduction and selection. A second drawback is that their bodily construction and interfaces are based on organic cells. This has many limiting consequences for the way in which they can become physically linked during the creation of a multistage and for the way in which the linked individuals can exchange information. The construction of a technical hardwired memon may improve on this situation, because its technical construction and interface would allow more powerful communication with other memons.

3. The SCI-memon and its multistage

Much more promising are the prospects for the pathway towards the structural (auto-)copying of information (SCI) memon and it's multistage. For this pathway, a hardwired memon first evolves to the SCI-state before it evolves to a multistage.

The SCI property has occurred earlier in evolution in cells. In cells, autocatalysis creates a full structural copy of the information in the cell. In present day cells this information is for the largest part allocated in DNA and/or RNA. The auto-copying process for neural network information does not involve DNA. Yet, structural auto-copying of information demands that the memon can copy the architecture of all neurone connections and the interaction strengths of all synapse connections. This leads to the very strict requirement that it must have access to all this information. A computer based memon can do this without problems. The only thing that is needed is an extra interface that helps it reading the arrays with information about the cell-cell contacts and synapse strengths, which information is kept record off anyway in programmed memons. In contrast, cellular memons will not be able to do this. They would require sensory cells which by some means find out what cells are connected to each other and in addition measure the strength of each synapse and report this to the individual. Apart from the physical tour de force to host large amounts of additional sensory cells in the brain, we consider the chance that this evolves naturally extremely small. Even genetic manipulation may prove an insufficient tool to reach such a complex goal.

From the moment, that a technical memon can examine its proper network structure and interaction strengths between cells, a whole world of new properties opens up, which allows a number of exciting predictions. The first prediction based on the SCI property is that, for reasons discussed above, it requires a technical construction. Accordingly, the SCI-property strongly guides predictions in the direction of computer-based entities. Despite that they probably have a technical construction, the SCI-memon and human beings have a similar basis for their neural network structure. In principle, this allows for human processes, such as intelligence, creativity, curiosity, etc. However, a technical construction will also imply important differences with respect to energy procurement and living environment. Energy procurement will focus on electricity. Moreover, because a technical memon does not breathe air, it can colonise underwater environments, planets without an atmosphere, or even a free position in space, supposing that other resources for normal functioning are available.

That SCI-memons most likely show a technical basis is furthermore of marked importance for the evolutionary debate. The reason is that SCI-memons cannot evolve as a special case of organic life. Being of technical construction, it is simply impossible that they evolve as offspring from cellular parents. Instead, SCI-memons have to be built either by cellular memons, as a special kind of tool (a p-meme!) that starts defining its proper goals in life, or by technical memons, as a special kind of constructed offspring.

The moment that SCI-memons can copy their knowledge structurally this will cause an earthquake in memic evolution, the importance of which can hardly be overestimated. As an exploration of the possibilities, the text below gives some examples

SCI-memons can reproduce their whole personality by simply producing a structural copy of their neural network. This copy will automatically contain all learned knowledge. Note that despite discussions about cloning, human beings absolutely lack a similar option. Humans cannot perform a structural reproduction of their whole personality, e.g. of all nerve connections and interaction strengths, simply because they lack access to the network structure and interaction strengths. When reproducing, humans only copy the genetic coding for a new phenotype. This will show a mixture of parental phenotypic properties and has a neural network upon birth that shows a good deal of genetically based pre-structuring. It is devoid, however, of learned knowledge. Thus, the body and the overall network structure are roughly copied, but without the smallest trace of anything that the parents have learned. As the structural copying of the parental knowledge is blocked, the transfer of knowledge to the offspring requires a long detour via many years of education. For SCI-memons reproduction of their complete knowledge can take place almost overnight as long as an appropriate technical device is available to which the information can be copied and in which it can become operational. This shows that as soon as intelligent technical memons have taken shape, it is but a small step to a whole population of such memons. In fact, technical memons wil find that the copying of a full parental network is costly in terms of energy and time. Some parental memons may therefore develop strategies aiming at the production of offspring based on small network-vehicle combinations with the best possible capacities for learning and maintenance, and the capacity to actively enlarge their bodily and memic construction to develop into full-grown memons themselves (r-strategists).

The functioning and copying of network topologies will require some kind of coding to handle the information about the connections and the interaction strengths between the neurons in the network. Such coding hold a similar position in memons as does the DNA in cells. Where specific regions on the DNA, the genes, code for specific proteins, it will now become possible to let specific regions code for neural modules with certain properties or knowledge. Although the brain functions in principle as a whole unit, evolution may find ways to create brains that allow certain modularity, much in the same way as genes allow modularity in the catalytic processes in the cell.

SCI-memons can use the access to their own neural network to evolve a number of features, including the signalling procedures between neurons, the integration functions via which the neurones decide whether or not to signal and the ways in which coding memes are coding for neural network constructions. There may even be possibilities for experimenting with parallel neural circuits, as long as solutions are found for their coupling onto existing network structure and for the fact that they are always full parts of the network, which implies that experimenting with parallel networks will always affect the functioning of the brain. A low-profile impact is only possible if in the testing phase, the interaction strengths are set to very low values.

Further aspects of technical memons that follow more or less from the above three points are also interesting. In order to stay focused on the major aspect of this study and prevent technical details, we will only shortly mention these aspects without going into details. These aspects are: the possibility of meme trade, the acceleration in memons of thinking speed with computing speed, the tendency towards the development of modular network architecture and the capacity of technical memons to integrate very different technical equipment directly into their interface.

SCI-memons have a much better chance of reaching multicellularity than hardwired memons. The main reason being, of course, that they may show very high evolution speed. This is based on properties such as: a programmed network structure, the creation of similar copies, the acquisition of informed networks (or network parts) via trade and the possibility of semi-intelligent interfaces. Increasing competition for living space will force SCI-memons to cooperate for survival. In some cases, this will drive SCI-memons to dependence on cooperation and to structural connection, marking the transition to the SCI-multi-memon state.

4. The HMI-memon and its multistage

The next predictable property of a future memic operator is that of Hypercycle Mediating Interface (HMI). The figure shows that the HMI property has occurred earlier in evolution, i.e. in atoms and eukaryotic cells. In atoms, the Hypercycle Mediating Interface emerges for the first time in the form of the electron shell. In eukaryotic cells the situation is more complex. Here, a new interface is added to the already existing interface around the cell, creating a second interface: the nuclear envelope.

In cells, large libraries of information are stored in the form of DNA/RNA. In prokaryotic cells the unpacking of the DNA-information and the functioning of the produced enzymes occur in the same compartment. In eukaryotic cells, the storage part of the information has become sequestrated to the nucleus, from where coded information is transported through pores in the nuclear envelope to the soma before it is translated to functional enzymes. This shows that the nuclear membrane separates the cell into two compartments. In the nucleus the information of the cell is for the largest part handled in a coded form. Outside the nucleus the information is active in the form of enzymes. Accordingly, the HMI property is associated with an extra internal interface that separates different levels of the expression of the information.

This observation offers information about possibilities for future system configurations, because the continuation of this sequence would imply a multi-layer interfacing of the hypercycle in future memons. Nevertheless, if this extrapolation is valid, it remains in our opinion quite hard to imagine at the present state of the understanding of the operator hierarchy what the second layer would look like in practice. Starting simply, two situations will be visualised showing an additional interface. These can then be combined to create a tentative prediction of a two-layer HMI-memon.

Imagining only a single layer, as in prokaryotic cells, it would be quite natural to assume that the HMI-memons find increasing use for coding memes.These have the shape of code-strings that represent the topology and strengths of all the neural connections of modular network parts. The reason for the popularity of c-memes is that they offer a highly efficient coding for storing information that via experience and learning was gathered in the neural networks. This implies that, by using c-memes, little energy is required to maintain large reservoirs of knowledge. Only a selection of the c-memes would have to be unpacked and used in response to specific environmental conditions, after which they could be packed away again in a new, more experienced form. In contrast to networks and code strings stored temporarily in the active working memory of the memon, a more profound storage of c-memes would imply that these are stored away in a form which is not directly accessible. An example could be any high-capacity data-storage medium. A potential candidate for this process is the three-dimensional storage in programmed crystals and the reading by means of laser beams. However, the storage and retrieval of large amounts of c-memes will require a special interface to decode the information. The new coding and the related interface would imply an additional Hypercycle Mediating Interface.

There is another way, via which an additional interface could evolve in future memons. To understand this, we have to place ourselves in the situation of an SCI-memon that has just copied its network structure into a new vehicle. Unfortunately, this imaginary new vehicle which is furnished with a lot of new technical properties differs in many aspects from the previous one. This implies that the memon has to go through a long process of revalidation and practice with its new body for calibrating its old neural settings to the new phenotypic properties. If the neural network of the memon possessed translation interfaces allowing a rapid adjustment of the memon interface to various types of vehicles, such a practice period could be made a lot shorter. Also, larger differences between the old and the new vehicle could be allowed. It will probably be most efficient to have only a selection of such interfaces active, namely those that yield the highest survival value under given circumstances. Other interfacing networks can then remain stored as c-memes in the central meme library.

The combination of an internal c-meme library with a translation interface would allow, in principle, a two-level Hypercycle Mediating Interface.

5. From hardwired memon, via HMI-memon without SCI properties to multistage.

Assuming independence of the closure types, a comprehensive discussion of the possibilities for future memons, should also include the route from hardwired memons to HMI-memons and their multistage, without the intermediate stadium of the SCI-memon. Although it represents a theoretical possibility, the direct transition from hardwired memons to HMI memons must be expected to suffer from technical problems due to a low rate of evolution. The reason is that hardwired memons cannot read their neural network state. A hypothetical HMI-memon without SCI properties would certainly experience severe problems with reaching a multistage. The contribution of this option to the mainstream of operator evolution must, therefore, be considered minimal, and the existence of these type of operators more of theoretical than of practical importance.

6. Summary

The above extrapolations present a panorama of possibilities for future organisms. First of all, the next stages will be technical neural network organisms, because this is by far the most likely option for constructing a system type that can show structural auto-copying of information. A consequence of this prediction is that humans must create the next stage in the operator hierarchy. This is required, because evolution cannot develop a technical construction using a cellular basis. The predictions also show that, one day, neural network organisms can be created that can copy themselves. For this purpose one could simply imagine a body factory constructing phenotypes that can be bought by any existing memon for the purpose of copying a neural network structure onto the memory of the new phenotype. From that moment onwards, humanity will have to live amongst reproducing, intelligent technical organisms. The operator framework furthermore depicts evolution as an open, ever-extending process. As part of this process, the next major transition will be based on a hypercyclic interaction between multi-memic elements, most likely within the environment of a multi-memic organisms that is supported by a technical vehicle. Finally, the operator approach suggests that if technical memons are not constructed, this will block, on earth, the evolutionary sequence that leads from one operator to the next.