Abstract: Ahmed Abdel-Fattah and Kai-Uwe Kuehnberger

Remarks on the Feasibility and the Ethical Challenges of a Next Milestone in AGI

Ahmed Abdel-Fattah & Kai-Uwe Kuehnberger

The success of a project like IBM’s Watson or the collaborative research endeavor of the European Union to develop a Large Knowledge Collider (LarKC) for web applications paves the way for the integration of enormous amounts of (not necessarily homogeneous) knowledge for AGI applications. Speculating about the next milestones in the development of general intelligence, we hypothesize that a large, distributed, massively knowledge-based learning system will depict new aspects of general intelligence that cannot be easily simulated with classical (real-world) agents or isolated systems. Just to make such a system slightly more transparent, we think that the integration of knowledge coded in video streams of privately attached cameras and mobile devices could be coupled with information available online in private or public databases, general-purpose knowledge bases, or domain-specific ontologies in order to allow completely new forms of inferences, predictions, and control. Such a system could proactively ask you to close the window in your office before leaving, if the outside camera tells it that your window is open and the weather forecast predicts a thunder storm in the evening. All such types of information mentioned above are more or less freely available and do not presuppose access to governmental sources of information (like surveillance devices, satellite images, profiles of persons generated by police or intelligence services etc.).

In our presentation, we will try to give informally some reasons why such a system is conceivable in the nearer future. We will base our claim exemplarily first, on possibilities to (semi-)automatically analyze and label images and videos using machine learning techniques and second, on techniques to integrate heterogeneous types of knowledge in distributed systems. Furthermore, the combination of different reasoning services will allow the generation of new, non-trivial types of knowledge. In the second part of the presentation, we will raise some questions about certain ethical issues imposed by personal rights of citizens.