On Definition of Artificial Intelligence

Summary

The paper shows the efficiency of interpreting the behaviour of humans, animals, and robots as manifesting the relationship between two texts-afferent and efferent. The transformation of one text into the other is treated as quasi-linguistic translation. Intelligence of behaviour is defined in terms of exactness and brevity of translation.

A model is presented, as developed at the Polytechnical Institute (Leningrad), which affords the recording of the efferent text as a function of the informational structure of the afferent text. The read-out is executed by reproduction of the afferent text; if the latter consists of insignificantly modified fragments of the earlier text, fragments of the efferent text are reproduced at the output in a correspondingly modified sequence. The intelligence of the behaviour of this system is analyzed, and it is found to be higher than in the case of “word-for-word translation” since the model has capacity of detection and extrapolation, and also since there is no preliminary segmentation of texts into “words”.

The linguistic approach suggests that it is not justified to assimilate the brain to the computer.