Klaus Doblhammer

Examination of foundational AGI-agents in Artificial-Life simulations

The examination of AGI agents is an interdisciplinary challenge. This is particularly the case for their foundations, assumedly (in compliance with Damasio) drives and emotions. We demonstrate how these foundations of a human-like control system are tested using simplified exemplary cases as test-specifications. After showing how we model the required functions of the agents decision unit, we provide details of the incremental steps of our examination. As a first step, we use calibration-specifications, which we concretize with the help of psychoanalysts and neuroscientists and show how a flexible parameterization of the model generates expected behavior. This step enables a deeper model analysis in the second step of our examination, which corresponds to model exploration and provides further information for adaptions of the parameters, model, or assumptions. Overall, this examination methodology allows for a stepwise model development and examination and provides the ground for comparing the simulation data with empirical data, which we plan as a next step.

The current state of psychoanalytically-inspired AI

Complexity of technology is constantly increasing; for the field of automation this means that economic considerations dictate a need for corresponding measures. Artificial intelligence boasts noteworthy successes in this area; however, its achievements appear modest when compared to the faculties of human intelligence. This paper will demonstrate that a new modeling approach is required via possibilities offered by the mindset and tools of computer technology, thereby demonstrating why a psychoanalytic approach seems sensible and necessary. The paramount goal of the research introduced in the following is a formal approach to describing the human psyche based on the neuro- symbolic and neuronal functional layers. Questions associated with this approach include: To what extent can the psychoanalytic model be confirmed by computer technology and the possibility of simulation? To what extent is an axiomatically developed terminology in psychoanalysis a prerequisite for the better integration of the natural-scientific way of thinking in psychoanalysis? The paper is based on the results of several fundamental research projects within the framework of ARS (Artificial Recognition System) which were funded in part nationally and in part by the EU.

Action planning for autonomous agents based on neuropsychoanalytical concepts

Action planning methods for autonomous agents are necessary to successfully perform tasks and purposely achieve goals within the agents environment. Based on already existing knowledge about the environment, action sequences are built. Development of actions is important for autonomous systems because the possibility of action generation determines the degree of freedom to act within an environment. This article introduces an action planning system for autonomous agents based on neuropsychoanalytical concepts of the human psyche. Based on a technically implementable model of the human mind, an action planning system is modeled and implemented to create a planning system for technical applications. The result is a flexible system that is able to create action sequences purposely achieving goals within the agents environment. The system can be compared to the human planning process and is the first homogenous work in the field of Artificial Intelligence completely based on consistent and unambiguous concepts about the human mind. To test the developed model, a simulation environment is used where agents controlled by an implementation of the model have to fulfill use cases.

Transformation of perceptions as filter mechanisms in artificial intelligence

When using the human mind as a template for artificial intelligence systems, principles of human thinking have to be modeled. One such very important principle is defense. Hence, in this article we describe a formalism to implement psychoanalytic defense mechanisms in artificial intelligence systems. They can be seen as transformations. A tuple1 consisting of perceptions or drives (of a software agent in our case) combined with a value indicating their strength is transformed into a different tuple, if a potential interpretation is considered inappropriate by the defense. Some types of defense mechanisms can easily be transformed. Other more difficult ones and abstract types of defense mechanisms are transformed by the use of transformation tables. The transformations can alter, suppress or pass each of the components of a tuple or the whole tuple. In this article we show the transformation tables, we give a categorization of defense mechanisms, we show the general form of the transformations and we show some examples of transformations and ways to implement them in artificial intelligence.

Perceptual prediction for bionically inspired autonomous agents

To manage the increasing volume of data per time unit, achievements in information processing and artificial intelligence were made. But still the complex processes of human perception and scenario recognition are not fully understood and still far from implementation in technical applications. The contribution of this article to the field of cognitive automation is the concept of prediction for perceptual- and scenario-recognition frameworks. It is a model where prediction originates from neuro-psychoanalytical theories. Inspired by experience-based planning, which is used by the psychoanalytical decision unit, the prediction of possible outcomes from scenarios can be used for proactive acting. It results in a higher detection rate and a faster performance for recognition-units. This first implementation shows the possibilities of the concept and gives an outlook of the performance as soon as the system is fully integrated in the decision-unit.

On the way to bridging the gap between the mental apparatus and the neurobiological layer

The mental apparatus – mind, psyche, soul – expresses itself in thought, play, love, creating, sex, planning, in fact in all human activities. Its foundation is the nervous system, as it is the central information system of the human being. For those interested in the connection between the neuronal level and the mental level, how is it possible to comprehend and model the connection between the neurobiological and the psychological? What does a natural scientific model of the mental apparatus look like? These questions led to the SiMA project (Simulation of the Mental Apparatus & Application). This article surveys the foundations of information theory and computer technology that underlie the SiMA project, and explains how the view of the neuropsychoanalytic world can be mapped into a model that can then be simulated and tested. The first results of several simulated case studies are described, suggesting that SiMA provides a way to validate theories of neuropsychoanalysis.

Funktionale Abbildung der menschlichen Psyche für die Automatisierungs- und Regelungstechnik

ZusammenfassungDas Projekt SiMA (bisherige Bezeichnung ARS) des Instituts für Computertechnik der Technischen Universität Wien hat zwei wesentliche Ziele: Zum einen soll nachgewiesen werden, dass die Psyche auf der Basis des psychoanalytischen Funktionsmodells nach naturwissenschaftlichen Prinzipien wie andere Prozesse der Natur simuliert werden kann. Das konnte inzwischen in verschiedenen wissenschaftlichen Veröffentlichungen nachgewiesen werden. Zum anderen soll untersucht werden, ob die damit gewonnene Architektur für regelungstechnische Modelle, aber auch für jene der Wirtschaft, sinnvoll eingesetzt werden kann. Dieser Aspekt soll im vorliegenden Beitrag behandelt werden. Es soll der Fokus auf die Motivation und die Notwendigkeit gerichtet werden, diesen Schritt zu gehen, für den es nach Ansicht der Autoren keine Alternative gibt, um komplexe Prozesse steuern und regeln zu können. Eingegangen werden soll auch auf die entscheidenden Vorteile solch einer bionischen Architektur auf der Basis eines Mehrschichtenmodells mit inkludiertem Primär- und Sekundärprozess in der dritten Schicht. Die Natur hat diesen Weg erfolgreich eingeschlagen, und wir sollten diesen Weg verstehen.

Comparison of technical filter mechanisms and defense mechanisms of the human mind

Looking for new paradigms in artificial intelligence, we are investigating functionalities of the human thinking process to manipulate information and filter perceptions. In this paper we introduce defense mechanisms of the human mind to be applied in artificial intelligence. We compare functionalities of defense mechanisms of the human mind with nowadays used filter mechanisms in artificial intelligence and explain reasons why defense mechanisms of the human mind open a broad new spectrum of possibilities and opportunities for artificial intelligence. In particular are these the defense mechanisms repression, deferral, sublimation, projection, disavowal, isolation, separation, depreciation and idealization. These defense mechanisms were chosen and devised with a team of psychoanalysts. We compare state-of-the-art artificial intelligence with psychoanalytic notions in our ongoing ARS project and explain why psychoanalysis is important for future developments in artificial intelligence. Finally, we give examples of similar projects.