Introduction to the Proceedings of the Workshop" Natural Organisms

Introduction to the Proceedings of the Workshop “Natural Organisms, Artificial Organisms, and Their Brains” at the Zentrum für interdisziplinäre Forschung (ZiF) der Universität Bielefeld, March 8-12,1998 Hennig Stieve Institut für Biologie II, RWTH Aachen, D-52074 Aachen, Germany Z. Naturforsch. 53c, 439-444 (1998); received May 18, 1998

There have been many meetings in recent years bingen, Hans Flohr, Bremen, Hans-Joachim on “Brain and Computer” topics under a number Freund, Düsseldorf, Karl Georg Götz, Tübingen, of different aspects. Although it is not my own Klaus Hausen, Köln, Wolfgang Prinz, München, field of experimental work, I have been discussing Helge Ritter, Bielefeld, Jürgen Schnakenberg, this subject with brain- and computer scientists, ro Aachen, Werner von Seelen, Bochum, Jan-Dieter boticists, psychologists, and scholars of the human Spalink, Durham, Gerhard Vollmer, Braun ities for several years (Stieve, 1995). The confer schweig and Reto Weiler, Osnabrück. ences about mind and brain which I attended or At first I tried to organize a one year research read about were often characterized by a lack of group at the Zentrum für interdisziplinäre true discussion, i.e. disputation of differing argu Forschung (ZiF) in Bielefeld. But this seemed not ments, and did not seem to have an impact on on to work. Later I became convinced that an intense going research. People presented their data, find workshop would be a better frame for such a dis ings and opinions in monologues without seriously cussion. discussing their grounding and the reasons for the For such a workshop it seemed necessary to disagreements with others. This was especially so • Select topics which need and deserve interdisci in the those fields in which consciousness and feel plinary discussion and where such discussion ings are involved, e.g. the question whether a com promises to yield results, because there are puter can have feelings. In many cases convictions many very interesting topics in this field where were mixed with confirmed facts, and plausibility discussion does not promise results now. or wide acceptance was taken as proof. Compared • Choose a framework which ensures a successful to conferences in biophysics and biochemistry this discussion. I chose the Dahlem Workshop Con is quite a contrast. cept with which I already had some positive ex Therefore I felt there is now - after the first perience. wave of these “Brain and Computer” confer A conference of the Dahlem type had never ences - even more need for interdisciplinary dis been held at the ZiF in Bielefeld. This appeared cussion in certain parts of this field. I tried to as to be quite a challenging opportunity to demon semble a group of people from neurobiology, strate at the same time the efficiency of this con computer science, robotics and philosophy for a cept. On the other hand, the directors of the ZiF serious discussion. With this aim I pursued the were not easily convinced that this concept would following strategy: work. For scholars of humanities a conference I contacted more than 50 people from the vari which is so different from a sequence of mono ous fields concerned and received very useful ad logues and from the standard workshops of the vice. I only want to mention a few of them who ZiF may be difficult to accept. All the more, we were especially helpful. Peter Bieri, Berlin, Tobias wish to thank the ZiF which after some hesitation Bonhoeffer, München, Valentin Braitenberg, Tü- decided to house and support for the first time a Dahlem-type conference, our workshop. We also wish to thank the staff of the ZiF, especially Ma Reprint requests to Prof. H. Stieve. Fax: (0241) 8888133. rina Hoffmann, Andreas Lueking, and Daniela E-mail: [email protected] Mietz, and its managing director Dr. Gerhard

Sprenger, and last, but not least Bettina Halbe, a following seven: Niels Birbaumer, Tübingen student of the RWTH Aachen, for their sensitive, (clinical and physiological psychology); Tobias effective and friendly help in preparing and con Bonhoeffer, München (neurobiology, brain physi ducting our workshop. ology); Yadin Dudai, Rehovot (neurobiology, memory); Rolf Pfeifer, Zürich (computer science; The Concept of the Workshop artificial intelligence and simulations); Jan-Dieter Spalink, Durham, NC (electrical engineering, The concept of the Dahlem Conferences is a communication networks); Hennig Stieve, Aachen concept for an unusual - but to my mind very (neurobiology, photoreception); Gerhard Vollmer, efficient - type of workshop. This Concept has its Braunschweig (philosophy, epistemology). Three emphasis on the discussion of the meaning of new of them already had positive experience with Dah and old observations, models, and theories and lem conferences in Berlin. aims to encourage new cooperations and the de The Committee met in Bielefeld in the Zentrum sign of new critical experiments. für interdisziplinäre Forschung (ZiF) on April 14 This concept includes some “tricks”: and 15, 1996 - almost two years before the work • Small discussion groups which design their own shop - selected the topics for the four discussion agendas. This allows effective discussions in groups, chose a name for the workshop, and sug which each member of the group can partici gested the participants to be invited and their roles pate. The jointly designed agenda allows to find in the workshop. room for the topics which the members find The date of workshop, 8-12 March, 1998 was worth discussing and brings the responsibility to chosen to fit to the possibilities of the Zentrum für the participants. interdisziplinäre Forschung (ZiF) in Bielefeld and • No lectures at the workshop, but background to be outside the teaching periods of German uni papers distributed in advance. Lectures tend to versities. However, this is the time of many other be monologoues and could kill the path of the meetings and in addition it is during the teaching discussion. Background papers distributed in periods of universities in some other countries. advance allow the members to prepare them Our time window happened to include the 70th selves to the discussion. birthday of the designer and founder of the Dah • Joint group reports which are published right lem Conferences, the late Silke Bernhard, to after the workshop with all the members of a whose memory the conference is dedicated. discussion group as the authors. This is an essen With our program we closely followed the well- tial requirement because it forces the group to tested time schedule of the Dahlem-workshops, reach conclusions during the limited time avail with one exception: we made it one day shorter. It able for discussion, and (since every member is turned out that it probably would have been better a co-author) every member of the group has to not to reduce it. agree on what is written down. The task to finish The participants: As the number of participants a group report implies that everybody tries to of the workshop should not exceed about 40, only stay focused during the group discussion. a small number of the persons interested and com • A “Program Advisory Committee” consisting of petent in the concerned fields could be invited. We representatives of the various disciplines rele tried to assemble a variety of participants includ vant for the theme of the workshop. It proposes ing younger and older, who were prepared to co the topics and the participants of the workshop operate, could supply interesting contributions, and their roles as moderators, rapporteurs and were able to listen with an open mind, and would authors of background papers. commit themselves for the entire duration. Not all who were invited could participate, but those who The Preparation of the Workshop did come made up for these losses by working The Program Advisory Committee: For the Pro hard, inventively, and cooperatively. gram Advisory Committee I won the cooperation The multidisciplinarity of the workshop partici of six persons from different backgrounds. So the pants may be demonstrated by the following. Nor Program Advisory Committee consisted of the mally, a participant personally knows about seventy H. Stieve • Introduction to the Proceedings of the Workshop 441 or even eighty percent of the participants already 4. Emergent properties of natural and artificial before coming to a workshop. In ours, most partici systems. pants had not met at least two third of the others. The term “emergence” has become fashionable. Our aim, then, was to have an efficient discussion It is used with quite different meanings by brain between workers in neurobiology, robotics, infor researchers, mathematicians, and philosophers. matics (computer science), and philosophy (episte- It seemed to be useful to understand the dif mology) in comparing certain properties of brains ferent definitions in which the term is used, try and computers of organisms and robots. Whereas to find a definition of emergence which is useful the interdisciplinary exchange of views between for our purposes, describe supposed emergent computer people and brain researchers has already phenomena in brains and computers and to dis begun a few years ago, the inclusion of epistemolo- cuss how far we today understand their origin. gists seems an innovation. To make such different Suggested topics were non-linear mechanisms in people talk to each other in an understanding way brains and computers, brains as generators of (patiently listening and exchanging arguments) is emerging faculties, and the so-called binding no easy venture. We hoped that the Dahlem con phenomenon, i.e. the adequate handling of dis cept would make this possible. tributed brain activities which correspond to dif The topics: The suggested topics for the four dis ferent properties of the same object. cussion groups of our workshop were chosen be It was suggested by the Program Advisory Com- cause the Program Advisory Committee thought ittee that the discussions at the workshop should that they deserved an open-minded interdiscipli focus on the comparison of properties of brains nary discussion and were likely to lead to interest and computers with an emphasis on the functional ing results: dependence of brains on their natural “bodies” 1. Representation of the environment in natural and of computers in robots on their connected ma and artificial systems. chineries. The significance of this interdependence How is the information about the properties of for the behavior of an organism has only gradually the environment represented in the brain to en become clear during the past few years. Compar sure the generation of appropriate behavioral ing those properties of brains and computers actions? Examples: The representation of space, which affect the behavior of organisms and robots motor action and faces in brains and in the anal helps to understand both better. Most of today’s ogous control “organs” of robots. robots are still relatively rigidly programmed to 2. Functional advantages of organisms with brains fulfill certain well-defined tasks in a straightfor in evolution. ward “if - then” manner. There is, however, a de The functional consequences of certain brain velopment of more advanced robots which can properties for the behavioral fitness of organ make more autonomous decisions to be used e.g. isms. Examples: brains of birds and insects as in emergencies or on Mars. These may allow even instruments to cope with ecological pressures, more interesting comparisons to living organisms. the evolution of robots’ capacity to cope with We know of relevant properties of computers ecological pressures, and the significance of the which cannot be achieved by brains (e.g. speed of brain as a learning machine for success in evolu processing). Are there relevant properties of tion. brains (accessible to natural science) which cannot 3. The behavior of natural and artificial systems: be achieved by computers, now or ever? solutions to functional demands. Limitations: Natural science is limited to objec Comparing performances of systems with func tively observable phenomena, in other words, to tional demands, including questions like behav consider only phenomena which are accessible to ioral decisions in conflicting situations. Sug independent observers is a conditio sine qua non gested examples: How molluscs decide on taste for “clean” natural science (Stieve, 1998: this issue, preferences, recognition of optical objects in pp. 445-454). This confinement has certain prag natural and artificial systems, resource alloca matical advantages: The scientific experiment is tion in distributed technical networks, and walk the acid test which decides whether an assumption ing robots. or a theory can survive or has to be discarded. 442 H. Stieve • Introduction to the Proceedings of the Workshop

Humanities do not have this independent judge. derlying structures and may play a part in brain Therefore many old theories cannot conclusively function (see Stephan, 1998: this issue, pp. 639-656 be discarded (e.g. those of the ancient Greeks). and Walter, 1998: this issue, pp. 723-737). How They may be recycled over and over, according to ever, since today there is no clue to determine the “Zeitgeist”. By contrast, in natural science whether such a problem is unsolvable (e.g. whether many theories (some of them brilliant) have been a phenomenon is strongly emergent), the only suc falsified and discarded by the force of critical ex cessful strategy for brain research is to try to solve periments. the riddles at hand until we know more about them. This limitation sets boundaries to pure natural Our Program Advisory Committee suggested scientific brain research. It excludes consciousness that topics which today are unlikely to lead to and feelings as possible objects of research since much progress (e.g. the mind-body problem) must these are only subjectively accessible and alterna not occupy much of the group’s activities, but the tives cannot be decided objectively. If beliefs and arguments for the current stalemate situation convictions come into play, this makes a discussion should be stated in the discussion group report. very difficult, if not impossible. Whether comput It seemed, however, desirable to come to a joint ers can have feelings, be creative or have intuitions statement on what the participants agree and like a human mathematician are questions of be where and why they disagree. lief; they can not be scientifically decided. Other brain problems which can not be solved How Was Our Workshop? today, may be solved in the future. Our present conjectures or assumptions may be right or wrong. Our groups worked in a self-organizing manner An example is the historical controversy about the after the moderator and rapporteur and main signal conduction in nerves: In the twenties and topic had been determined by the Program Advi thirties of this century there was a vigorous con sory Comitee. They found their own agenda and troversy whether signal conduction in nerves is way to handle their tasks. They acted somewhat “electrical” or “chemical” (see e.g. Muralt, 1946). like organisms. Sadly, our workshop was hit by un Both sides had powerful advocates. Finally, this foreseeable losses. Apart from the usual drop-outs, controversy was solved experimentally in the late an infectious flu deprived us of several important forties and fifties (see Hodgkin, 1951). Those participants. As they had been carefully chosen to 1920s scientists whose theories emerged as true be part of our small number, this was a hard blow. had not had the better arguments, but better intu But this is the beauty of organisms, they can com ition and possibly more luck. Guessing the right pensate for losses. If a shore crab ( Carcinus) loses answer is not in the end sufficient for the progress one or two legs it can adapt its way of walking in of our scientific understanding, but designing and a way that beautifully compensates for the missing making the critical experiments is, and even a limbs. And our discussion groups did exactly the wrong theory may lead to it. Of course, to look at same thing. They compensated for the missing the available data in the right way, which involves members synergetically and did a complete job by sensitive guesswork, intuition, and even aesthetics, modifying the suggested topics according to the is very important for designing good theories and expertise present. Actually the groups discussed important experiments. the following: For our understanding of brains and computers there may be problems involved which are insolu Group 1: Representation in natural and artificial ble in principle. Gödel has shown that there are cer systems: tain mathematical problems which can not be solved: “All consistent axiomatic formulations of The concept of representation was in the focus number theory include undecidable propositions” of the discussion in this group. Examples can be (Gödel, K. (1931), quoted after Hofstadter (1979)). found in descriptions of the visual system, the mo Similarly, the so-called ‘strong emergence’ is de tor system, as well as in central systems like the fined as the occurrence of a novel phenomenon hippocampus. Although the description of neuro which cannot be explained on the basis of the un nal activity is often phrased in terms of representa H. Stieve • Introduction to the Proceedings of the Workshop 443 tions, the concept is less well understood and has Group 2: Influence of brain and computer design several implications which have to be considered. on the performance of natural and artificial Firstly, the relevant aspect of neuronal activity organisms: establishing a representation is subject to an in The discussions were focused on the relations tense debate - some researchers favor the mean between structures (the materialistic implementa activity of neurons as the relevant signal conveyed tion at the level of molecules, subcellular organ to other neurons, while others emphasize the pre elles, neurones, and networks) and functions de cise temporal structure of neuronal activity. Thus, fined teleologically for brains and computers (on estimates of what is represented in a particular a variety of different levels of explanation). Exam structure are dependent on the assumptions made ples of evolutionary achievements in natural or on the relevant variables and might be different ganisms were compared with the advancements of for an outside observer and other brain structures. computer and robot design that improve their per Secondly, the concept of neuronal representa formance. Pivotal points of the discussions were tions is often used in a passive sense. The emphasis the role of algorithms for the functioning of brains is placed on recreating properties of the external and computers, their embodyment in the internal environment as faithfully as possible. Thus, the in architecture of natural brains and artifacts, the im teraction with neuronal structures upstream and plementation of plasticity, and some possible downstream have traditionally been assumed to limitations for the performance of computers, e.g. occur in a feed-forward manner. in the field of creativity. Thirdly, representations are not static entities but have to form during development and learn ing. This poses the questions of maintenance of Group 3: The behavior of natural and artificial representations, of the interaction of new experi systems: solutions to functional demands: ences with previous ones already stored, and poses the problem of grounding a representation in the Natural organisms have different strategies for real world. the adaptation to their environment. For example For these three aspects this group considered it there is a great variety of eye types in natural or timely to investigate the concept of neuronal rep ganisms adapted to a certain environment and dif resentations within several fields of neuroscience ferent tasks. The discussion group tried to focus research. Exploiting the broad range of expertise on the similarities between natural and artificial of their members, the group undertook to investi systems and the benefits for the different fields of gate the concept of neuronal representations from research. At the moment it is common sense that as many aspects as possible. In the group report a both fields could profit from each other, but the series of examples is presented in a comparative major problem is to give concrete examples for fashion. For each example an attempt is made to successful interactions and to raise new and inter identify the relevant variables, the nature of the esting questions. The examples discussed ranged interaction with upstream and downstream struc from sensory information processing to motor be tures and the influence of training and develop havior, and some speakers gave nice examples of ment. By this careful analysis of neuronal repre artificial systems incorporating biologically moti sentations in different contexts a more precise and vated strategies in sensory-motor integration. But consistent use of this concept might be possible. these examples showed clearly that the develop The most outstanding result was the common ment of a theory of the motivational system is a alty in a number of principles of representation demanding task with a lot of unresolved problems that were common to all of these examples. The both in natural and artificial systems. importance of the interaction of organisms with The role of models, their usefulness and limita their environments was observed in each of the tions, for brain and computer research was also systems studied. discussed in this group. 444 H. Stieve ■ Introduction to the Proceedings of the Workshop

Group 4: Emergent properties of natural and Conclusions artificial systems: The aims of our workshop were to exchange The discussion of this group focused on the phe ideas and arguments and to evaluate them jointly, nomenon “emergence”. Due to the complexity of and to initiate and plan new experiments and co natural systems and the limitations of the related operation. theoretical approaches to their properties, the Here we present the outcome of our efforts. As term emergence originally was mainly used to la expected our workshop was hard work. But all bel certain properties as somehow appearing from participants seemed to enjoy to work so hard for some underlying structures. As it became more satisfactory results. Many members, especially the and more technical, the need increased to relate rapporteurs, often worked until very late at night this term to other theoretical approaches and to (sometimes until 4 a.m.). The longer the workshop discuss it in terms which can be reduced to observ proceeded, the more we all felt the time pressure able phenomena. Thus, suggesting that all observ to produce results jointly before the workshop able phenomena can be reduced to a consistent set ended. However, with high pressure you can make of underlying scientific laws, at least in principle, small diamonds from coal! would imply that emergence has no intrinsically Our group reports are of course not the ultimate methodological significance as a technical term. truth but represent the present state of under On the other hand, the term emergence would standing as determined by the joint expertise of play a significant role under the assumption that the members of the group; they state the there exist properties that in principle are not re- agreements and disagreements and reasons for ducable. The related question about the princi both. pally reducibility of systems phenomena is truely I hope that for all the tough work we all enjoyed at the very border of science and appears to be this exercise in discussion culture, the intellectual impossible to answer. pleasure of exchanging competing arguments. As a practicable avenue, the term emergence was discussed in the frame of pattern formation, especially in the context of neurobiology. It turned out that pattern formation may provide a metho- logical and conceptual frame to discuss empirical My thanks go to those who read the manuscript findings in the fields of neurobiology and behav and made helpful suggestions and contributions , ioral sciences on a rigid mathematical ground namely: Bettina Halbe, Peter König, Hubert while opening clearly shaped ways to different but Preissl, Stefan Reimann, Rupert Schmidt, Tilman related fields. Stieve and Gerhard Vollmer. In this group it took a major effort on the se The workshop was supported by the Deutsche cond day to resolve some initial misunderstand Forschungsgemeinschaft, Ministerium für Wis ings between brain and computer scientists on the senschaft und Forschung des Landes Nordrhein- one hand, and philosophers on the other. But this Westfalen, Zentrum für interdisziplinäre For paved the ground to an interdisciplinary discussion schung (ZiF) der Universität Bielefeld, Daimler which apparently was satisfactory to all the mem Benz AG, Georg Thieme Verlag, Schering AG, bers of the group. Siemens AG, and SmithKline Beecham Stiftung.

Gödel K. (1931), Über formal unentscheidbare Sätze der Stephan A. (1998), Varieties of emergence in artificial Prinzipia Mathematica und verwandter Systeme. and natural systems. Z. Naturforsch. 53c, 639-656. I. Monatsh. Math. Phys. 38, 173-198. Stieve H. (1995), Limits of natural science: Brain re Hodgkin A. L. (1951), The ionic basis of electrical activ search and computers. Z. Naturforsch. 50c, 317-336. ity in nerve and muscle. Biol. Rev. 26, 339-409. Stieve H. (1998), Comparing properties of brains and Hofstadter D. R. (1979), Gödel, Escher, Bach: An Eter computers. Z. Naturforsch. 53c, 445-454. nal Golden Braid. Harvester Press Ltd., Hassox, Sus Walter H. (1998), Emergence and the cognitive neuro sex, GB. science approach to psychiatry. Z. Naturforsch. 53c, Muralt A. v. (1946), Die Signalübermittlung im Nerven. 723-737. Birkhäuser Verlag, Basel.