There are many definitions for intelligence as well as types of intelligence. Most definitions include lists of activities inherent to humans, such as reasoning, planning, solving problems, thinking abstractly, comprehending complex ideas, learning from experience, and adapting effectively to the environment. When attempting to develop intelligent machines, researchers try to make them perform a subset of activities inherent to humans with as high a level of plausibility as possible. However, from the standpoint of knowledge formalization and for a more precise formulation of research goals, it would be useful to point out the smallest subset of the activities to establish that an animate being or inanimate machine performing them is intelligent enough.
First, it is necessary to mention that intelligence implies a goal (in the “Handbook of Human Intelligence” Sternberg and Salter defined intelligence as “goal-directed adaptive behavior”). In the case of animate being, the goal could be surviving, and in the case of a machine created by humans, the goal could be solving tasks assigned by humans. Animals show their basic intelligence, for example, when they adapt to the environment, seek for food, and build nests. One can say today’s machines show their intelligence by effectively performing tasks people created the machines to perform. This is not a too distorted interpretation of intelligence if we review the book “Symbolic Logic and Intelligent Machines” of Edmund C. Berkeley published in 1959, where the author called intelligent machines electromechanical arrangements capable of solving problems that involved logic.
As animals are capable of satisfying their basic needs for living, let us assume they possess a basic level of intelligence. To better satisfy needs for living, an advanced level of intelligence would be the production and use of work tools. As it turned out, the production and use of work tools is specific not only to humans but also to chimpanzees and to a number of other animals. For example, chimpanzees are able to find stones of appropriate weights and sizes and crack nuts using them. Chimpanzees are also able to find long and thin sticks and use them to kill small rodents living in trunks of a certain species of trees and extract killed rodents for food. It is important to mention that chimpanzees not only find ready-to-use work tools in nature but can also produce them. For example, chimpanzees can construct arrangements consisting of small tree branches and put them into termite mounds to eat termites crawled on them. Another example is an experiment conducted on a pygmy chimpanzee when experimenters successfully trained him to manufacture a sharp stone tool for cutting a string to open the door of a chamber with sweets.
In a digital world supported by computers, work tools are computer programs helping solve various tasks. In a computer environment, the aforementioned advanced level of intelligence is the capability to synthesize and use computer programs and algorithms. A machine with an advanced level of intelligence should be capable of automated synthesis of algorithms in some form or fashion. Such an algorithm might have a form of an ordinary computer program, possibly containing a set of subroutines.
There does exist a higher level of intelligence—we shall call it an expert level of intelligence,—specific only to humans and not to chimpanzees. Indeed, what is an essential difference between humans’ and chimpanzees’ intelligence if they both produce and use work tools?
In the book “The Complete Idiot’s Guide to Human Prehistory,” its author Robert J. Meier briefly noted an important fact about the production and use of burins1. Such specially sharpened stones archaeologists find during excavations corresponding to the time when, tentatively speaking, we can call creatures that have been living on Earth humans. Burins are tools created for manufacturing other tools. Chimpanzees do not make tools for making other tools. That is, an expert level of intelligence peculiar to humans is the ability to organize processing chains where one kind of work tools takes part in the production of other kinds of work tools. The established concept for this is production of the means of production.
For the computer environment, programmers develop such tools as compilers, various operating, execution, and development environments to simplify developing other computer programs. A machine with an expert level of intelligence would be capable of the automated synthesis of algorithms that the machine would execute to synthesize other algorithms to solve problems assigned by humans more efficiently. The creation of such a machine is a quite specific goal researchers could try to achieve.
As the author of this manual read a translation of that book, its original text in English might use a word other than burin.