Der Artikel “Intelligenz braucht Gefühle” der Wiener Zeitung kann hier gelesen werden:

Link zum Artikel

und hier downgeloaded werden:

Artikel Wiener Zeitung

Was treibt einen Computer? Wie berechnet man Besetzung?

Freuds zweites topisches Modell, die Triebtheorie und deren Anwendungen in der Computertechnik

Der Flyer zum Workshop kann hier downgeloaded werden: flyer-workshop-20032010

Moderation: Dr. Gyuri Fodor
Samstag, 20. März 2010, 9 – 14 Uhr
Salzgries 16/3, 1010 – Wien

Die Psychoanalyse und der Paradigmenwechsel in der Künstlichen Intelligenz

Die von der Computertechnik zu bewältigenden Prozesse werden zunehmend kom-plexer. Aber bis jetzt können die Forscher viele der großen Erwartungen in die „künstliche Intelligenz” noch nicht zufriedenstellend erfüllen. Nicht zuletzt aus diesem Grund arbeitet ein Forschungsteam von Technikern und Psychoanalytikern unter der Leitung von o. Univ. Prof. Dr. Dietmar Dietrich am Institut für Computertechnik der Tech-nischen Universität Wien an einem Paradigmenwechsel in der Künstlichen Intelligenz: Aufbauend auf der Triebtheorie und dem zweiten topischen Modell von Freud soll „der psychische Apparat” für technische Zwecke modelliert und schließlich emuliert werden, um technische Probleme bei großen Datenmengen besser bewältigen zu können. Die speziellen Bedingungen und Erfordernisse der technischen Modellierung zu berück-ich-tigen und gleichzeitig wesentliche psychoanalytisch definierte Inhalte zu wahren, bedeu-tet eine große interdisziplinäre Herausforderung, bei der es nicht nur eine eindeutige „Lösung” geben kann. Als größte methodische Herausforderung hat sich dabei ergeben, dass die Axiomatik in der Psychoanalyse nicht ausreichend präzise formuliert (beschrie-ben) ist, weshalb es gilt, diese wissenschaftlich neu herauszuarbeiten – eine Jahrhundert-aufgabe.
Mit dem Workshop soll ein regelmäßiges Forum hergestellt werden, in dem die Psycho-analyse aus der Perspektive der Naturwissenschaft und das technisch orientierte Modell unter dem Blickwinkel der Psychoanalyse betrachtet und diskutiert werden.

Programm

09.00 Eröffnung (Fritz Lackinger)
09.05 Problemstellung (Dietmar Dietrich)
09.20 Das Modell im Überblick (Dietmar Bruckner)
09.40 Ausgewählte Problemstellungen in der  Modellentwicklung (Klaus Doblhammer)

10.00 Diskussion*
11.00 Pause

11.30 Diskussion*
12.30 Pause

13.00 Diskussion*
14.00 Ende

*je nach Teilnehmerzahl in Kleingruppen oder Fishbowl

In den Pausen: Stehbuffet mit Fingerfood zum Netzwerken

Potentielle Fragen für die Diskussionsrunden

• Wie können Psychoanalytiker und Computertechniker zusammenarbeiten?
• Welche Prinzipien aus der Psychoanalyse können in der Computertechnik Verwendung finden?
• Wie kann die Computertechnik die Psychoanalyse als Wissenschaft unterstützen?
• Kann ein Computer nach dem Lustprinzip arbeiten?
• Wie lassen sich die Gefühlsqualitäten der Aggression in einer Maschine simulieren?
• Welche Bedeutung können Affekte und Bewusstsein im Computer haben?
• Was treibt einen Computer?

… Version 4.0 …

1. Subjectivity and science: how does it comply?

(a) Psychoanalysis deals with subjectivity. However, there exist common, objective mechanisms behind subjective experiences. Hence, subjectivity arises through content, not through function.

(b) If natural scientists do not take into account subjectivity, they will never understand how the brain and its mental apparatus works.

2. What is the specific contribution that psychoanalysis in particular can make to computer engineering (as opposed to psychology in general)?

Psychoanalysis offers a number of contributions, the major to be mentioned are:

- A functional model of human mental processes

- A distinction between conscious and unconscious processing

- A motivational drive theory

- A theory about defense mechanisms

- A theory about decision making based on subjective evaluation

3. What contribution can computer engineers make to psychoanalysis?

Computer engineers can offer psychoanalysts knowledge in

- Creating functional models

- Creating abstract models

- Formalizing and refining theories

- Removing contradictions

4. Which mechanisms do we use to implement the psychoanalytic model in a computer? What is the difference and advantage to classical AI?

Classical AI solves sub problems. There exists no serious, comprehensive modeling of the human mind. What has been normally done is applying different specific behavior phenomenon from psychology (even often from different schools without analyzing their interoperability) and trying to implement it in a computer with the experience gained through self-observation. Based on scientific premises several mistakes have been made as it can be seen today:

(a) Engineers cannot conclude from behavior to the psychic model, because they do not have the domain knowledge required for that field. It would be presuming doing this in regards of neurologists, psychologists and psychoanalysts.

(b) One should not mix up the knowledge from different schools without asking for their interoperability.

(c) One should not mix up the terms projection, behavior and model. A doll eventually may seem human-like but still consists of dead material. Moreover, a doll has no human characteristics entailed. Humanoid robots are not humanoid, as long as they are not equipped with consciousness, unconsciousness, and sexuality. Scientists should take care that their wishes do not get confused with reality, even if a lot of money can be won.

5. How can we unify the specifications of basic emotions and drives?

Psychology and especially psychoanalysis offer lots of different and sometimes rather contradicting specifications about emotions, affects, drives, and feelings. In order to get a starting point for a comprehensive, non-contradicting model, we use Freud’s definitions and subsequent research findings that do not contradict but enhance them. See also answers to questions 25ff.

6. To what extent are human control loops understood? (emotions, hormones, …)

The process “human being” is controlled by myriads of control loops on physiological as well as mental level. Here, we do not investigate the physiologic part, the body. On the mental side we follow Luria’s definition of three layers: projection, association and comprehension. We see that parts of the lowest layer are understood (e.g. Förster found 6 hierarchical layers of neurons in the eye). Also large portions of the highest layer can be modeled in principal, because there exists theory (this is what we do). However, in-between there are large white spaces. We welcome any contributions in this area! What we will come up with in future is an interface description from both sides as a starting point for defining the required functionality of the “white gap” in-between.

7. How far is learning, adaptation, plasticity integrated into the model and into the current implementation?

There is no comprehensive model of the human mind without learning. However, human learning is more than simply storing something. Knowledge needs to be put into context. Currently, we want to avoid this problem as far as possible. However, we found out to require some basic learning in form of memorizing scenarios, memorizing thing and word presentations and assessing memories with emotions and feelings.

Never-the-less one has to take into account, that learning entails all levels of the mental apparatus and cannot be only reduced to conscious learning. Including this knowledge from the beginning would be way too complex for the project. However learning is foreseen for the future like other processes, such as sexuality, unconsciousness and consciousness.

8. How “memory” is functionally organized?

See also the answer to question 7.

Mental content consists of four basic types: affects, thing and word presentations, and associations. Thing presentations are processed unconscious (and in levels even below), whereas a thing presentation with attached word presentation can be processed conscious. This ability is called pre-conscious. Both types can be of arbitrary complexity, ranging from not more than mere sensor values up to the concept of a well known person. All (thing and word) presentations are mutually associated with other (thing and word) presentations and evaluated with affects. This allows a flexible concept of memory.

9. Should we model a brain or should we model a mind? Can we model the brain without considering its physiological basis?

We should model a mind. We need a detailed hierarchical functional model that ranges from the highest abstraction layer (i.e. Id, Ego, Super-Ego) down to mathematical equations. The latter can be performed on any hardware, be it neurons or transistors or something else. Information theory can be implemented on different types of hardware. Functionality shouldn’t be affected, whereas a performance deficit must be accepted.

10. How can we model cathexis – the drive to invest emotional energy in a person, object or idea?

Cathexis is incorporated into the model as emerging property – as it happens in humans. A person, object, or idea is assessed with emotions and feelings all the time it walks through the model (is of current interest). If over time a particular (or varying) assessment always increases, a positive feedback loop starts that leads to more and more concern with it.

11. How can we model inspiration, creativity, or being a positive force and an inspirational force for others?

In our thinking and our model inspiration and creativity happen through associating different mental contents and bringing them into some common context. How far this happens in an instance of our model will be parameter-dependent, since it is an intrinsic feature.

We have to take into account that the mental apparatus continuously stores optical, acoustic etc. images, scenarios, and acts for a life-time. These will be continuously linked to each other. Inspiration and creativity is generated, if the association of these linked scenarios and images is not suppressed or hindered by the function of the Super-Ego, for example. Strictly religious and other ideological rules and laws reduce the ability of making associations in the mental apparatus, which can be seen in the developed model.

12. Would a machine with its own drives, emotions and feelings be willing to work for man?

Well, if we design a personality with exactly the parameters of a human that is grown up in our time and world, it may not want to solely work for others. However, that’s not the goal. If we tune parameters in a way that assess working for others very positive it may love humans for allowing it to do useful things like animals.

13. How does a human being make decisions?

Decisions are first based on emotions and feelings that assess memorized images and scenarios. Second, they are based on acting-as-if with expected, again, emotionally assessed, effect.

14. What are the origins of decision?

See also the answer to question 13.

Nature has invented higher level entities, such as consciousness and unconsciousness in order to give the higher developed creatures individual capabilities to survive better. The origin for decisions should be linked with unconsciousness and consciousness that are connected with past experiences that have been attributed with emotions and feelings. Decisions are generated through the wishes, psychological requirements and external influences, etc.

15. Does intelligence require embodiment?

See also the answer to question 14.

Intelligence was evolutions answer to allow a species with such a complex body as ours to survive. Hence, there could not be something as our intelligence without a body and its needs. However, the principles have evolved and there is no reason why it should not be possible to copy them to another domain. On the other hand such adaption only makes sense in case the same kind of problem class has to be overcome: to mediate and draw conclusions between contradicting requirements.

16. Does our model require embodiment?

See also the answer to question 14 + 15.

The model requires something that generates needs and interactions with the environment. In this sense the need for performing some task could be seen as the interface to some artificial body. To what extent this can be simulated is also subject of research.

17. Where is free will in our model? What are the mental functions behind it?

Even when a lot of people try to ignore it, neuroscience has narrowed it down: there is no free will or at least much reduced: Neither in our model nor in humans. The concept of something extra in our mind next to the mind itself that can act totally free is rather new in philosophy; it arose during the Enlightenment. In earlier times people were happy with their place in the world. Free will is a modern social concept that cannot retain deeper analysis. It is similar than with Darwin’s observations: We have to learn to acknowledge scientific facts even if they opposite our convictions, especially from a religious and judicial point of view.

18. Would a human want its robot to show free will?

Well, we guess: yes, to some extent. N.B.: See also the answer to Q17 that describes our view on “free will” as a whole. However, we guess that a robot can have a bit of its own “will” in the sense that it could need the ability to convince a human of not doing something harming or else way completely wrong (think of patients suffering from dementia). However, this is a question of the far future. In other words: principals of the human mental apparatus should be adopted in the model, given that free will eventually exists, it should be included. Unfortunately we don’t know, whether free will exists actually. Thus, it will be shown within the next few years.

19. Which are the basic terms that can be transposed from psychoanalysis to computer science?

Siehe unsere Excel sheet (link angeben)

20. Why did we choose the second topical model?

The second topical model consists of three instances: Id, Ego and Super-Ego within the function blocks. These function blocks support the functional development within the technical modeling.

21. Is there a relation between early Freud works (aphasia-studies, “Entwurf”) and the approaches of the ARS-project?

Only up-to-date findings will be adopted, bearing in mind that they do not contradict with the classic Freudian school. It should be emphasized that we do not want to reject one of the other schools, but we have to focus on one school in order to achieve a homogeneous model and not a mixture. Hence, interoperability has to be preserved. Of course, other departments can take advantage of other schools. We have to take in mind that we are financially limited and therefore cannot create another project.

22. How is “psyche” described in ARS-project?

The term for “psyche” for engineers represents partly the model and partly the behavior, which has to be strictly distinguished when developing the model. The methods can be reviewed in various dissertations.

23. Is computer-science influenced by psychoanalysis?

In order to develop intelligent systems, computer and cognitive science have investigated the creation of an artificial intelligence framework. With the psychoanalytical approach, we see the first good chances to get a controller and a decision unit model that operates on the principles of the consciousness and the unconscious.

24. Which epistemological concept of psychoanalysis is applied in computer science?

Given that we can look at nature from a monistic point of view and the mental apparatus is going to be ever-more transparent like physiology, we will gain more understanding from a nature science point of view about consciousness and the unconscious function. In medieval times it was forbidden to sect corpses. Even today some people believe that the mental apparatus is not accessible for nature scientists. However, we are not sharing this opinion.

25. How do psychoanalysts and computer engineers cooperate in practical research?

It has taken a long time until we have found a practicable way: both fields have to be ready to engage within each other’s mental worlds. One has to respect the scientific field of the other. This may sound trivial, but it is on the contrary. All the time, we hear technicians utter: “We do not believe that.”, although, it is psychoanalytical expert knowledge. On the other hand, psychologists will argue that the brain is not a computer without knowing what the definition of computer entails. Furthermore, others argue that consciousness will never belong to the computer as a part of it. Taking note of such unscientific statements one has to ask, whether we have something located in the brain that cannot be explained with natural scientific investigation? Are we still thinking in such medieval perspectives? Eventually, the more questions will rise based on the more questions we can answer. However, this is not related to the fact that we gain a better understanding on consciousness and the unconscious and finally having the ability to simulate it.

26. Are the different schools of psychoanalysis represented at the ARS-project?

See also the answers to question 4 and 20.

27. What is the position of Freud in “simulating the mind”?

He is founder of the theoretical concept. No more, no less. He has made the grounds for the way we investigate the mental apparatus from a nature science point of view.

28. Why are psychoanalytic concepts used at the ARS-project to simulate the mind?

See also the answers to question 4 and 20.

29. Why is it necessary to integrate death drive in the model?

It is necessary to integrate Freud’s later drive theory in order to satisfy the phenomena of the fusion of drives motivating libidinous and aggressive behavior. Freud’s motivation for a libidinal theory has to be represented in the technical model, at least on a very basic level. It is technically considered with the death drive.

30. In which way affects are modeled and how do they emerge in the model?

According to Freud’s early drive theory, affects will be modeled as a first representation of the drive besides thing representations. In the model a technically necessary cause-and-effect chain of functions will be implemented of processes that cannot be thought separated from each other in psychoanalysis. Based on this, drive tension develops and then thing representations and finally affects will be generated, which represent pleasure and aversion at these earlier stages. More precise affects will be defined via the inner perception (Solms), which is a requirement for a secondary-like processing of affects. This processing is achieved in the model by means of defense mechanisms in the pre-conscious/conscious part of the ID.

31. Psychoanalysis has a very differentiated psychology of human development. How does the model consider this fact?

This particular aspect has not been covered yet in the model, as the behavior that has been generated in the model does not rely on the content of an individual developmental experience. Decisions in various function modules are based on hypothetical constructs that seem meaningful and from a psychoanalytic point theoretically comprehendible.

32. Psychoanalysis deals with dynamic phenomena. That corresponds to Freud´s lifelong development of theories, which sometimes he has refused for arguing another concept. What does it mean for modeling?

Technical modeling is based on principals that require from now and then making some “cuts” within the Freudian theory in order to decide for one or the other concept and basic structures, which describe the theory.

33. What does it mean, if technicians argue, that their science helps psychoanalysts to define their terms more exactly?

For engineers the usage of ambiguous expressions presents a challenge, since these are additionally used in a contradictory context. Technical modeling requires a clear definition of terms that do not contradict themselves. If this is not the case, this reveals a weakness of the theory from an engineering point of view. By questioning psychoanalytical terms from various sides we want to be able to define these terms better. This is one of the main aspects the ARS project concentrates on regarding philosophy of science.

34. What are the possible applications for a system/robot/agent, based on the psychoanalytical theory?

The main targeted applications are control for complex automation systems (e.g. in factory automation, transportation systems, vehicles, aviation …), additionally, we think of applications in AI and robotics.

The table can be downloaded here: Psychoanalytical Terms and Definitions (Version 12 – 17.11.2009)

In the future our two research groups will collaborate on language,
especially its cognitive and emotional aspects. To this end, several
projects are being developped right now.

Enjoy running through this exciting day again and watch the presentations of Mark Solms, Aaron Sloman or Jaak Panksepp. Just click on “The Videos” on the top navigation on this site.

Dietmar Dietrich, Brit Müller, Tobias Deutsch und Roland Lang

Der Artikel zum Vortrag kann hier downgeloaded werden: .pdf.
Die Folien zum Vortrag finden Sie hier: .pdf.

Moderation: Wolfgang Groysbeck
Donnerstag, 19. März 2009, 20 Uhr, c.t.
Salzgries 16/3a, 1010 – Wien

Ist das menschliche Gehirn auf dem Computer simulierbar? Mark Solms, der sich klar zu den Monisten zählt, schreibt: Das Gehirn ist kein Computer. Werden Millionen von Euro trotzdem für eine Utopie ausgegeben? Oder geht Sigmund Freuds Wunsch in Erfüllung, dass seine Modellierungen in der Naturwissenschaft nicht nur eine Basis für weitergehende Forschungsarbeiten finden, sondern seine Ergebnisse größtenteils belegt werden können?

Bisher hat die Technik, im speziellen die künstliche Intelligenz (heute nennt sich eine der abgespaltenen Strömungen: {technische} kognitive Wissenschaft), keine Notiz von der Psychoanalyse genommen, noch mehr, sie hat sie mit der Unterstützung vieler Psychologen ignoriert, in der Wissen-schaft eigentlich ein Novum. Das hätte nie passieren dürfen, ist doch der Part “Related Work” ein Muss jeder wissenschaftlichen Arbeit, und darin hätte die Psychoanalyse immer wieder angesprochen werden müssen. Aber nein, es ist einfach so gut wie nicht geschehen, und dies nun seit über 60 Jahren, ohne dass dies jemand ernsthaft moniert hat.

Dietmar Dietrich, Ordinarius für Computertechnik an der Technischen Universität Wien, hat vor ca. 10 Jahren auf einer internationalen Konferenz erstmals auf dieses Problem hingewiesen. Die darauf folgenden Diskussionen und Dissertationen machten deutlich, dass die Integration der Psychoanalyse in die Technik eine gewaltige Chance für beide Fachrichtungen beinhaltet – zumindest von der wissenschaftlichen Seite aus betrachtet. Er gründete deshalb ein Team von Technikern  (Elektrotechniker und Informatiker) und Psychoanalytikern. In der Zwischenzeit wurden viele Diplomarbeiten und Dissertationen geschrieben, die ersten theoretischen Ergebnisse liegen vor, im Sommer sollen die ersten praktischen Simulationsergebnisse veröffentlicht werden.

Der hier angekündigte Vortrag hat zum Ziel, vor allem Nichttechnikern das Projekt vorzustellen und die Machbarkeit vor Augen zu führen, ist doch die Psychoanalyse der letzte Bereich, in dem der Computer bisher keine Bedeutung hat. Dietrich ist davon überzeugt, dass die Modellierungsversuche von Sigmund Freud nicht nur der Technik zu einem enormen Aufschwung verhelfen werden, sondern auch die Psychoanalyse wird davon langfristig einen merklichen Nutzen haben. Auf all das wird im Vortrag eingegangen.

Mitarbeiter des ARS-Teams sind: Dietmar Bruckner, Dietmar Dietrich, Tobias Deutsch, Edgar Holleis, Roland Lang, Clemens Muchitsch, Brit Müller, Nauman Qadeer, Anna Tmej, Heimo Zeilinger, Tehseen Zia, Gerhard Zucker
Berater des ARS-Teams sind: Elisabeth Brainin, Dorothee Dietrich, Georg Fodor, Samy Teicher

Vist: http://www.simulatingthemind.info/the-book/pressetext/

or http://www.simulatingthemind.info/the-book/press_release/

Simulating the Mind

Simulating the Mind

Dietmar Dietrich, Georg Fodor, Gerhard Zucker, and Dietmar Bruckner, editors. Simulating the Mind – A Technical Neuropsychoanalytical Approach. Springer, Wien, 1 edition, 2009, ISBN: 978-3-211-09450-1.

Can psychoanalysis offer a new computer model? Can computer designers help psychoanalysts to understand their theory better? In contemporary publications human psyche is often related to neural networks. Why? The wiring in computers can also be related to application software. But does this really make sense?

Artificial Intelligence has tried to implement functions of human psyche. The reached achievements are remarkable; however, the goal to get a functional model of the mental apparatus was not reached. Was the selected direction incorrect?The editors are convinced: yes, and they try to give answers here. If one accepts that the brain is an information processing system, then one also has to accept that computer theories can be applied to the brain’s functions, the human mental apparatus.

The contributors of this book – Solms, Panksepp, Sloman and many others who are all experts in computer design, psychoanalysis and neurology are united in one goal: finding synergy in their interdisciplinary fields.

More information can be found here …