Future Prospects of Human Interaction with Artificial Autonomous Systems

Abstract

The growing complexity of intelligent systems and technologies raises questions concerning their interaction with human intelligence. The loss of an ability to control artificial intelligent and autonomous decision systems (AADS) due to their high level of sophistication exceeding human analytic capabilities may be referred to as one aspect of ‘singularity’. The latter term is often used to describe potential threats to the mankind coming from the development of AADS that may outperform human intelligence in its most relevant aspects. This paper presents some results of a recent foresight project SCETIST which shed a new light on the above ‘singularity’ dilemma. The project aimed at building the scenarios and trends of selected advanced information technologies. Based on a classification of AI enabling technologies, two basic scenarios concerning the development of AADS have been built. The first one points out that the newly emerging global expert systems (GES) coupled with the Brain-Computer Interfaces (BCI) will allow the human societies to explore in full all data streams and knowledge repositories available. The global knowledge from all sources will be further processed by GES so that rational human decisions will not be outperformed by those made by AADS. In the second scenario, the AADS enabling technologies will develop faster than BCI and GES, so that autonomous decision systems can dominate. The third scenario, ranked as least probable, indicates a possibility of slowing down the development of ICT and AI, so that the singularity problem is deferred. The time horizon of the above scenarios was 2030. Finally, we will present the recommendations arising from SCETIST from the point of view of shaping the R&D policies.