A Primer on “Self-assembling Data Networks”

These data structures provide true “learning” without programming.

Imaginary data networks can be grown like data crystals or data trees in electronic memories without programming or supervision. This can produce true mathematical learning in electronic “autosopher” using similar rules and principles found in self-assembling physical structures like crystals, living trees, or societies. The basic algorithms were first disclosed by Klaus Holtz in 1975 (Patent 4,366,551) and presented in 1977 in a paper: “Here comes the brain-like, self-learning, no-programming, computer of the future.” There are seven known self-learning data structure, but only the “serial” networks will be explained here as an example.

The following concepts will be explained:

1 Autosophy “learning” is like growing imaginary data structures in electronic memories, without programming or supervision.

2 The mathematical process is like defining points in Omni Dimensional Hyperspace using simple algorithms.

3 Small data items are combined into larger and larger data items in a multi-level progression with the same methods of assembly on every level in the progression.

4 Complex data items contain imaginary “Information Entities” which may determine our behavior, and which are the true subjects in evolution.

5 Self-growing data networks can greatly improve data communication and replace the programmed data processing computer with brain-like self-learning archives.

The “Serial” Omni Dimensional Network

This type of network is used to compress and store serial data sequences such as: text, still images or video.

“Learning” is a process of defining “knowledge” as nodes stored in a space with an unlimited number of dimensions. The network in Figure 1 shows how four data sequences: 1-3-2, 3-2, 5, and 1-5-2 are stored as nodes in 1, 2, or 3 dimensional space. This is all that can be illustrated on paper. The network starts growing from a reference point, node 0, which is also known as the “seed” node. The numbers or symbols in the input data sequence are used as vector length to define a location in the next dimension. Input number “1” in the example is used as length 1 to locate a first location in the first dimension, the next number “3” is used as vector length in the second dimension, while number “2” is used as vector length in the third dimension. Each newly defined location is marked with a new “node address”. The node address numbers are assigned in sequence, such as node 1, 2, and 3 in the example. The input sequence 1-3-2 is therefore learned by three nodes 1,2,3 in a three dimensional space. The final node address “3” then represents the input data sequence 1-3-2. The input sequence 1-3-2 can be retrieved from node address “3” by following the nodes backwards through the dimensions towards the seed node 0. An infinite number of data sequences can thus be stored in the network, as node addresses, where each node represents a quantum of “knowledge” as a node in Omni Dimensional Space.

Figure 1. “Knowledge” storage as nodes in Omni Dimensional Hyperspace

A quantum of knowledge, in Figure 1, therefore consists of an “Origin”, also known as “POINTER” to define the previous node address, a “Vector Length”, also known as “GATE”, stored together in a “Node Address”. The more nodes are defined in the network the larger the amount of stored “knowledge”. Every location in the network can be defined only once by a new node address. This is because one cannot learn what one already knows.

The method shown in Figure 1 allows us, for the first time, to accurately define and measure “knowledge” and “learning” in a mathematical way. “Knowledge” can thus be measured as “nodes” or “engrams” (a suggestion was made to measure knowledge in units of “Holtz”, the discoverer of the knowledge quantum). The network example only shows nine nodes. Practical hyperspace libraries used for communication may contain tens of thousands of nodes, while very large archives may contain many Billions of nodes. This is far beyond human intelligence to program or repair. Locating and repairing individual hyperspace nodes in a self-learning memory device is just as impossible as locating and repairing defective neurons in the human brain. Fortunately for us, these networks will grow by themselves, using simple algorithms, which once set up will require no human programming or outside supervision of the internal functioning. The special memory devices, known as CAROM or DECAM Memories are self-organizing, self-checking, self-repairing, self-healing, and even self-cloning. The memory devices will not fail or provide wrong information even after suffering severe physical damage. This would allow us to replace the Programmed Data Processing Computer with a next generation of self-learning, brain-like, failure-proof “Autosopher”.

Figure 2. Self-assembling “serial” hyperspace libraries

Figure 2 shows the same network nodes as in Figure 1 but rearranged in a more convenient way. The vector length is replaced by a GATE value, which is usually a Hyperspace ADDRESS. The dimensions are no longer needed, where the network may grow into an unlimited number of dimensions. The Knowledge Quanta are stored in individual nodes, where each node contains a GATE (the new information), a POINTER (the ADDRESS where the node originated), and an ADDRESS where the Knowledge Node is stored. Each input data string finds or generates its own trail through the Network Nodes. The process can be imagined like the growing of data crystals or data trees in an electronic memory. The final version of the network is shown on the right, where all the network nodes are stored in a memory device. New Network Nodes are added in a “Next Empty” location, (Node 9) where the network may grow into Billions of nodes.

The precise steps are show in Figure 3 in the patented Network Algorithms. Input data, such as text from a book, are fed into the encoding algorithm to generate a very large network tree. Every text word is learned only once, to be reused in the following text. The entire library memory must be searched before a new node may be stored in the memory. This would require a Content Addressable Memory (CAM) for very fast data storage. The best memory devices would be the Autosophy native CAROM – DECAM mass memories, which are both Random Addressable (RAM) or Content Addressable (CAM) and which may provide enormous storage capacities in memory devices that will not fail even after suffering physical damage.

Figure 3 also shows how the text words are retrieved from the network, in reverse order, from the hyperspace ADDRESSES. The algorithms are very simple and may be implemented in both software-only or in hardware chipsets. There is no need for embedded microprocessors or micro-program memory devices.

Figure 3 The serial network generation and retrieval algorithms

Text data is stored in the memory as a serial progression, shown in Figure 4, in which smaller data symbols (ASCII text character) are converted into Hyperspace ADDRESS codes, called “Tip”, which are the node ADDRESSES at the final tip of the tree branches. Each tip code may represent any large amount of text data. Lower level symbols (text character) are converted into higher and higher level ADDRESS codes representing sentences, paragraphs, sections, and finally the text in a whole book. Each level in the progression follows the same algorithms show in Figure 3. The output tip ADDRESS codes from lower levels are fed as input codes to the next higher level in the progression. Each string of ADDRESS codes is terminated by a special “End of Sequence” code, which is different for each level. The text in a whole book may thus be converted into a single last ADDRESS code from the highest level in the progression. The book ADDRESS code may then be used to retrieve the entire text from the book. All levels in the progression may share the same library or mass memory device by assigning a separate “SEED” address for each level in the lower part of the library.

Figure 4 Text as a serial progression

The text contained in a book is, however, not merely a progression of codes. It contains imaginary “Information Entities” or stories, which are the true purpose of books, and which are the true subject of evolution. Reading a book causes the Information Entities to copy themselves into the human mind, where they mingle with all the already existing Information Entities in the mind, to change human behavior towards political ideologies, religious believes, or provide computer expertise. All physical self-assembling structures contain such imaginary Information Structures. The purpose of Autosophy research is to understand these processes to eventually build intelligent, self-aware, conscious, machines or robots.

Figure 5 Data storage in Hyperspace libraries for archiving and communication

Autosophy data communication and storage is entirely different from conventional (Shannon) bit and byte communication and storage. The Shannon form of communication and storage can be regarded as an example of extreme inefficiency because it does not include a knowledge library.

As shown in Figure 5, Shannon data communication and storage is linear. A doubling of the data volume would require a doubling of the transmitted bits and bytes or the doubling of the storage capacity in memory devices. Autosophy data storage, in contrast, is hyperspace saturating, in which the more data is already stored the less additional storage space is required to store additional data. This is because every data item in the text progression is stored only once and reused in further data storage. No text word, sentence, or paragraph would be stored twice. In text communication, likewise, the larger the hyperspace library becomes the fewer codes need to be transmitted. The hyperspace libraries also provide an unbreakable encryption method for communication and storage security.

Autosophy methods can provide enormous commercial opportunities both for Applications ready for use and for Research opportunities, for applications not yet ready for the market. This may improve our entire communications infrastructure by orders of magnitude. It may also replace the programmed data processing computer with brain like archiving systems, leading to self-learning intelligent robots, and eventually to true Artificial Intelligence.