Proto-Intelligence
If the phenomenon of intelligence is to be viewed as a spectrum of phenomena, what is the underlying mechanism which created such a phenomenon in the first place? What lies at the base of the phenomenon of intelligence? All living systems including the subsystems of advanced biosystems exhibit some level of intelligence as defined in the present work. The question then must be asked: May inorganic, non-living systems exhibit intelligence?
Leaving aside “machine intelligence”, which will preoccupy us in later chapters, let us consider the case of a crystal of manganese dioxide dropped into a solution of potassium permanganate. The crystal, instead of dissolving, becomes the focus of an autocatalytic reaction which converts the solution of potassium permanganate into manganese dioxide.
The ability of a crystal of manganese dioxide to convert its external environment into more of itself demonstrates that an inorganic system may reproduce. As such the crustal has fulfilled one criterion for ascertaining intelligent behavior – the enhancement of reproducibility.
A seed crystal of salt dropped into a supersaturated solution of salt or a seed crystal of silicon dropped into a cooling mass of molten silicon will also trigger off reactions which result in the precipitation of salt crystals, or the growth of the seed crystal into a large crystal of silicon. Again, the criterion of reproducibility has been fulfilled. Does this mean that crystals possess intelligence? We are now at the borderline of our spectrum. We need to make a decision as to how we delimit our spectrum. When we define the “visible” spectrum of light, we define it in terms of light being visible to the human eye. We exclude portions of the ultra-violet – which happen to be visible to bees, for example.
Similarly, our definition of intelligence will exclude individual crystals. Instead, we will consider them to be proto-intelligent. That is, crystals, being highly organized entities which possess considerable information and engage in substantial information processing, lack true intelligence: Under usual circumstances, they fail to duplicate themselves. When they do reproduce, they are engaging in intelligent behavior. However, once the reaction is completed, the crystals return to their wholly inanimate state, totally subjected to the vagaries of fate which their environment may impose on them. Nevertheless, if only briefly, crystals may exhibit “flashes of intelligent behavior”. For this reason we invoke the concept of proto-intelligence.
The concept of proto-intelligence is important: It is important in its own right. It is important to any analysis of intelligence. And it is a prerequisite for the analysis of machine intelligence.
Proto-intelligence may be defined as phenomena which involve aspects of intelligent behavior but exist only temporarily, or if permanently, only partially satisfy the criteria for intelligence. For example, memory is an integral part of the learning process. Many non-living systems exhibit the property of memory: “Memory metals” and other materials which “remember” previous shapes and processes, a disturbed pendulum returning to its resonant frequency, many computer systems – both hardware and software – all of these exhibit various levels of memory and as such, constitute a form of proto-intelligence.
Still more primitive in organization involving fewer components, are molecules, atoms and subatomic particles. As Haefner (1991) has pointed out, these entities manage to maintain their identity – a proton behaves as a proton, an electron as an electron and a hydrogen atom as a hydrogen atom. It is the ability of these entities to maintain their physical integrity, to engage both in information exchanges and in information processing, which implies that they are stable information systems which exhibit aspects of intelligence. In the present work, their inability to reproduce themselves, unlike biological systems, puts these inanimate entities into the lower forms of intelligence – that is, proto-intelligent systems.
Reproduction in biological systems is based on the transmission of information across generations via a stable genetic material (DNA or RNA). The selection of new characteristics (mutations) to adapt the system better to the environment involves a learning process. That is, just as a rat learns its way through a maze by trial and error, then remembers the successful moves, so does a species learn by the trial and error of random mutations, then remembers the successful ones by incorporating this beneficial information into its genome (DNA or RNA) for future use.
Therefore, the ability of a system to reproduce itself, or be reproduced externally, is a vital component of intelligence because without reproduction, the system has virtually no chance to evolve. All evolution involves a learning process, and the capacity to learn may be as useful a guide as any for ascertaining whether a system exhibits true (rather than proto-) intelligence.
The concept of proto-intelligence must be an integral part of the concept of the evolution of intelligence. The evolution of the lung in animals living on dry land was preceded and derived from the evolution of the swim bladder in fish. The evolution of bones, important for survival on land, was preceded by the evolution of bony fish. Fish don’t need hard bones to function well in water. Sharks, which possess only relatively soft cartilage, have fared very well for hundreds of millions of years swimming around in the oceans. Fish moving upstream into fresh water, however, faced an uncertain supply of vital calcium. What more logical solution than to create a calcium “bank” by depositing calcium compounds among the cartilage? Thus the cartilaginous mechanical structures may be considered as “proto-bones”; similarly, a swim bladder can be considered to be a “proto-lung”, and certain fins of fishes as “proto-legs”. The move from an aquatic to a terrestrial environment is a most remarkable step upward in the evolution of life on this planet, involving not only the vertebrates but the invertebrates (insects were probably the first land animals), plants and the earliest of all invaders, micro-organisms.
We recognize that there is a difference between animals that spend their life on land, and those that spend their life in water. However, there are numerous amphibians which represent an intermediate state. In a like manner, the evolution of information systems into intelligent systems probably involves intermediate stages. It is here that the concept of proto-intelligence becomes useful. We look for phenomena that, in themselves, do not satisfy our criteria for intelligence, but which are related. We would not look to the eye of a fish to evolve into a lung, nor its blood vessels to become bones; instead we look to its swim bladder and its cartilage. Similarly, we look to information systems exhibiting limited aspects of intelligence as constituting the phenomena of proto-intelligence form which intelligent systems evolved.
The above discussion has identified several phenomena which may be classed as proto-intelligent:
1. Survivability, as demonstrated by the stability of atomic particles.
2. Reproducibility, as exemplified by the growth of a crystal.
3. Memory.
Other phenomena will be identified as we delve more deeply into the subject. However, one form of proto-intelligence which is so basic as to make impossible any informed discussion of intelligence without it, is feedback. All intelligent systems process information along at least one major feedback loop.
Feedback loops have their antecedents in the form of regular cycles which, under certain circumstances may represent a form of proto-intelligence. Such cycles may be observed in the organization of an atom, a swinging pendulum, a resonating electronic system, a planet swinging around a star, the stability of Benard cells, cyclic chemical reactions – that is, any cyclic phenomenon with a regular periodicity involving two counteractive forces: centrifugal/centripetal, electrostatic/electromagnetic, oxidative/reductive, etc. Like a gyroscope spinning, regular oscillations tend to maintain the stability of the system even when the environment changes. It represents a major mechanism in the survivability of a system. In some proto-intelligent systems, the oscillations dampen down in time and finally disappear. The system collapses. In others, such as atoms, the organization is maintained indefinitely by a complex interaction of internal forces. In contrast, in biosystems, internal rhythms are maintained by the controlled inputs of an external source of energy. Plants do it by utilizing sunshine, animals by eating plants or other animals. Similarly, mechanical or electronic systems function by having available a source of energy.
A grandfather clock is a prime example of a piece of machinery which constitutes a proto-intelligent system. It is a goal-oriented object. Its goal (imposed by its human designer) is to move the hands of the clock in small, even exact, and continuous steps. It processes information in that the time it takes for a weight to descend is converted into the movement of the hands of the clock. It achieves this goal by regulating the input of energy by means of mechanical gears and levers and a steadily swinging pendulum, so as to achieve a constant output. However, the clock contains no system which compensates for changes in the environment. The steady, gravitational pull of the weight, moves the steady, ticking machinery. If the weight becomes insufficient, as it does when it reaches the end of the chain, the clock stops. If the weight becomes too heavy, the clock tries to run faster and may break. An increase in temperature causes the pendulum to become longer, and the clock slows down. The reverse happens if the outside temperature cools.
The grandfather clock exhibits aspects of intelligence: It is goal-oriented. It processes information. It regulates (in limited fashion) the throughput of energy. It converts energy into information. And its pendulum exhibits memory (in so far as it will return to its basic frequency of oscillation if disturbed). However, it is totally dependent on the right combination of externally imposed factors in order to achieve its goal. And it cannot learn. For this reason one may class a grandfather clock as a proto-intelligent device.
The grandfather clock exemplifies a system which has many of the attributes of intelligence, yet it should be classed as a proto-intelligent system. When we examine machine intelligence in later chapters, we will see that the divide becomes increasingly blurred. The problem is a familiar one to biologists trying to define taxa: “When is a variant a new species?” – a problem which has its counterpart at all levels of classification.
For example, in vertebrate taxonomy, the class “mammals” differs from the class “reptiles” from which it evolved, in that mammals in general, possess hair, give rise to live young and are warm-blooded, while reptiles have scales, lay eggs and are cold-blooded. However, the armadillo, a mammal, has scales; the duck-billed platypus, another mammal, lays eggs, while certain snakes and other reptiles give birth to live young; and certain dinosaurs are believed to have been warm-blooded. The reason for this overlap is that mammals evolved not once, but independently several times from the reptiles, and the products of various lines of evolution exhibit an overlap of characters.
In like fashion, the evolution of intelligent systems from proto-intelligent systems must have occurred on numerous occasions, involving quite different kinds of advanced information systems – resulting in the insoluble taxonomic problem of defining unequivocally what is a proto-intelligent system, and what is an intelligent one. We must recognize, therefore, that the bottom end of the intelligence spectrum is blurred: It becomes impossible to create a clear demarcation between advanced information-processing systems which exhibit proto-intelligence and those which could be classed as truly intelligent.
