Inhalt
 
ARCHITECTURE OF BIOLOGICAL AND ARTIFICIAL INTELLIGENCE
A Short Comparative Outline
 
Rainer K. Liedtke, MD, Munich Germany
PHARMED INSTITUTE OF CYBERNETICS
 
PIC Res. Comm . 4/07
 
 
Introduction
Intelligent information processes of the human brain, in particular its strategies and goals, are significantly different to those in a computer. In addition, there doesn't yet exist a generally accepted definition to characterize the "intelligence" of a technical system. Principal differences between biological and artificial intelligence appear primarily based on two major aspects: the strategies and goals to utilize information and the structure of the hardware. In order to characterize some principal properties of these systems this shall be compared within a brief descriptive outline.
A comparative view on the structure of biological and artificial intelligence
In everyday life human intelligence is subsumed under the concept that a person "knows what’s going on". Synonyms are "cognitive capacity" or "thinking ability". Human intelligence is therefore usually equated with a capability for planned and rational thinking. It should be noted at the outset that this reduces human intelligence to conceptualizing thinking as processes of so-called "higher" (abstract) brain functions and more or less fails to consider the objectives of "why" and "for what purpose" we as living beings developed and need intelligence at all.
From a biological and evolutionary viewpoint intelligence, however, primarily derives from a necessity to survive therefore constitutes an essential individual advantage used in efforts to succeed in a competitive environment. In real-life reality, therefore biological intelligence is an individual characteristic which must prove its worth under the constellation of individually different conditions in which people live.
Consequently biological intelligence also includes interactive (social) aspects and "subordinate" processes like those of the emotional and motor informational processes. Human intelligence therefore turns out to be not only an ability to abstractly analyze and to prepare routes for decisions, but also to implement them in a suitable way. Accordingly, biological intelligence has various components of different origins which every person has, though in individually different proportions. These various components complement and influence each other. Technically these interactions appear based on the use and influence of different memory types.
In contrast to a biological intelligence the artificial "intelligence" of a computer, or a machine, can be summarized as a reduction of processing abstract data only. This already refers to the existence of different hardware architectures. Compared with biological brain systems show technical systems a topology that exhibits a smaller number of interactive levels and also different signaling pathways. In addition, the brain's structural complexity includes a dynamic variability (e.g. repairing and adapting mechanisms) whereas a computer has a static monolithic structure.
 
Fig 1. A simplified and schematic comparison of basic components in the informational architecture of a brain (left) and a computer (right). The brain exhibits a dynamic multi-level control structure whereas a computer has a static monolithic structure. The red lines indicate the interactions via bus systems. It should be noted that a "bus" in a biological system does not only work via neural connections. Biological systems dispose of a diversity of signaling pathways which address various targets in different sequences. They build cascades of both topical (extra- and intracellular) and systemic feed-backs which again are controlled by other cascades of signals. A large number of these signals can currently not sufficiently reproduced by usual electronic circuits. In addition, biological systems have compared with computers also a toolbox of different commands. (Abbreviations: CPU: central processing unit, MEM: Memory (internal), I/O: Input/ Output ports).
The "fixed" architecture of the computer hardware determines an over-expressed singular capacity with a limitation to parametric processes. Consequently computers do not exhibit human properties such as e.g. creativity, anticipation, or individuality. In principle they produce a "continuous planning process without sufficient implementation". From a biological view such property is similar to persons called autistic savant. For a biological individual, however, any planning without implementation is an unreasonable action which has no practical value. The same is also the case in the opposite direction of "permanent actions without any underlying planning" (something we know in the every day life from persons who show a "hyperactivity"). The technical attempts to implement human properties, such as e.g. "emotions", in machines are more or less based, and also limited, on programs that trigger conditioned responses (biologically in the sense of a so-called conditioned reflex).
Further problems of artificial intelligence are various aspects of redundancy. This appears due to a lack of creative (associative) anticipation and which produces an imprecision in conclusions: Asking wrong questions means generally searching in a wrong direction. This problem is well known from internet search machines, where only a small percentage of the information offered by the computer has some practical value for its user. Similar problems represent, on a next level, those processes where computer intend to produce projections. Computers often fail in complex cases, since a prerequisite of solving such problems does not only require a fast access to large data bases (where they are often more efficient than human beings), but also a property of contextual associative processes.
Conclusion
The brain performance as a whole mirrors more than activities of a neural network that only transports and processes abstract data. The current hardware architecture of computers hasn't yet reached such technical level which can compete with the complex and dynamic structure of the brain. In addition, the current software doesn't yet employ the options of "shortcut strategies" as they are used in biological systems. Consequently a transformation of the information processes of a human brain into an efficient technical solutions requires a basically different logical approach which also implements a new kind of contextual elements.
 
© 2010 Rainer K. Liedtke