DRAFT
The initial title of the presentation was "Man-machine connection 2100". While the main emphasis of the presentation lies in the communication between human beings and machines I decided to change the word connection into symbiosis while the word connection gives more an idea of physically connected entities. The word symbiosis refers to situation in which both parties benefit from the existence of the relation. We human beings naturally would like to benefit from the existence of the machines but one could ask what is the status of the other direction. In my thoughts it is not meant to refer to the rights of the machines but could merely mean that the machines may exist if they have positive influences in our lives.
"Gordon was very much concerned with the role that computers and the new information technologies can play in making positive contributions to our lives. He foresaw most of today's new developments decades ago. It is becoming increasingly acknowledged that his ideas and vision were well ahead of their time. His book Microman gives an accessible account of many of them (London: Century, 1982, co-author, Susan Curran). Gordon looked to the day when all human knowledge would be located in self-organising, interactive, multimedia archives, with intelligent agents to support learning and access. He referred to such support systems as 'vehicles for driving through knowledge'. He talked generally of an era of 'man-machine symbiosis'. He speculated about new forms of immortality, in which not only cultural artefacts are preserved but also, in some way, individual minds and personalities."
One can ask what kind of developments the first century of the new millennium will bring us. Do they follow mainly Pask's positive visions or are we facing an era during which oppression and top-down control are enhanced using the tools of the latest technology? Are we becoming more and more dependent on the information technology that forces us to adapt to the requirements of the systems rather than vice versa?
In the following, I will outline two areas which may be important in the development towards a positive vision.
One way of enhancing the representation is to take into account the unclear boundaries between different concepts. Many names have been used to refer to this phenomenon such as gradience, fuzziness, impreciseness, vagueness, or fluidity of concepts. (Honkela 1997)
When one considers epistemological questions (questions of knowing) in the framework of classical logic problems in such a view seem to be unsolvable: either you believe in something or you believe in its opposite. However, when one adopts the background assumptions of the classical logic one also tends to forget the rather self-evident (IMHO) nature of the world which is not a collection of entities and their relationships in the level of modeling the domain and propositions that refer to the domain. Thus, propositions like "snow of white" or "the pine grows in my yard" refer to highly complex states of the world with possibly millions of details (some of which are relevant in evaluting the propositions and some not). Moreover, there are large number of more or less different instances of "snow", "white things", "pines", "yards", etc. There is a one-to-many mapping the words to the world. Ambiguities make this mapping even more complicated. Because one "state of the world" can be described in several ways the mapping becomes many-to-many.
The contents of propositions such as shown above should not be conceptualised straightforwardly as is being done in the model theoretical approach (and related approches) in logic or in the knowledge representation formalisms of artificial intelligence or cognitive science. These simplifications may be often useful but if they are always applied without considering the complicated relationship between the perceptions and the languages, and the conceptualization of the whole domain, problems arise.
What does all the above tell us about the development of better man-machine symbiosis? The discussion points out the problems of many present-day natural language processing systems and expert systems (and even information systems in general). In order to develop systems that would be able to communicate with human beings in much more fine-grained and contextually tuned manner, one has to develop the ways how machines can experience the world (e.g., through images), learn from the experience, and learn to associate language with the perceptions, or more accurately, with the perceptional clusters, and finally create conceptual systems with a reference to the languages and conceptual systems of human beings (ranging from languages to individuals).
At the www page http://www.mlab.uiah.fi/~timo/emotions/affective.html, one can find a summary of topics related to affective computing including a list of (potential) applications.
Affective computing considers mostly how the digital applications could be "aware" of the human emotions. Also computational modeling of emotional phenomena may approve to be useful. By modeling the development of, for instance, anxiety, depression and mania (Hyvärinen & Honkela 1999), the avoidance and treatment of such states may become more efficient. One interesting systemic state is paranoia which seems to be very deconstructive and developments in avoiding unwarranted paranoia would most likely lead into positive effects in the society.
When modeling of emotions is considered, the objective should not be in providing expectations related to individual cases or human networks. This means that an emotion modeling system should not be used to determine the fate of a person by providing a prognosis if the person is a co-operatively behaving person or not. The main aim would rather be in the understanding of the complex phenomena and especially their emergent qualities. This understanding would provide means to stabilising systems and self-help methods for finding a way to deal with the circumstances in a way that each individual finds most suitable for him or her. In concrete terms, one could have a therapeutic system that helps in avoiding, for instance, anxiety on his or her own will, or in developing the communicational skills. I hope that the theory of learning system provides the mankind a balancing mindset as opposed to the domination of genetic determination. In the cognitive and emotional realm, human beings appear to have a lot of "neural plasticity" which means that development is surely possible. It remains to be seen if computational tools can be used in successful modeling of the phenomena and relationship between cultural evolution and individual development, and guiding the choices into useful directions.
Similar understanding of the complex phenomena of bodily functions in general and their relationship to external conditions could provide tools for healthier life. In 2100, one may carry a watch that can tell: "Your liver function is lowered by 40% mostly because of the previous meal you had. I recommend that you have a 6 hour pause in eating and next meal would contain ...". One can, of course, argue whether this is a positive vision or not.
As a final remark, I may remind you of the passage in the obituary for Gordon Pask: "He speculated about new forms of immortality, in which not only cultural artefacts are preserved but also, in some way, individual minds and personalities." I also foresee a time in which our minds can be preserved to a certain extent. This does not mean preserving ourselves as feeling and acting individuals - for such an immortality I cannot see any more than anecdotal means. However, we may be able to developed during the following century highly personal assistants that are given the same perceptions that we gain and receive information about our acts and choices. Then, later, this kind of systems could answer questions such as "what would have Timo said about this topic". On the other hand, the existence of such systems would be a considerable risk. These systems should be under the total control of each "host" individual, and, finally, the (hu)mankind should have reached a level in which the will to harm the others (esp. those who are different) should be at much lower level so that we would be able to use the developments of technology for the best of all. Whether there could be a feedback or feedforward connection from the technology to the human system that helps in achieving such goals remains to be seen.
Honkela Timo (1997): Self-Organizing Maps in Natural Language Processing. Thesis for doctor of philosophy, Helsinki University of Technology, (http://www.cis.hut.fi/~tho/thesis/")
Hyvärinen Aapo and Honkela Timo (1999): Emotional Disorders in Autonomous Agents? Advances in Artificial Life, Proceedings of ECAL'99, European Conference on Artificial Life, Springer, 1999, pp. 350-354.
Bernard Scott: Obituary for Professor Gordon Pask, http://www.venus.co.uk/gordonpask/gpaskobit.htm, 4.1.2000.
What kind of modules would be needed in a system that would be able give intelligent answers to questions like "find all the pieces I need to repair this car", "find in the database all the pictures that show autumnly lake views", or "play me some pieces of music from the late 19th century or early 20th century that seem best to characterise Russian music at that time". Assume that neither of the databases (pictures, music) do not contain textual metadata or classifications. Which modules or tasks seem to be the most difficult?
1.2.
If you sketch an object-oriented model (semantic network) of the knowledge related to the questions above, which issues or subtasks appear to be the most difficult?
1.3.
What kind of possibilities there might be to link automatically the symbolic descriptions with the images or audio samples?
What kind of applications of Self-Organizing Maps and Independent Component Analysis you can find in your areas of interest? (cf. the web pages http://http://www.cis.hut.fi/research/som-bibl/) and http://http://www.cis.hut.fi/projects/ica/)
What kind of uses can you find for intelligent computing technologies combined with traditional technologies, for instance, cars. Could autonomous agents help in driving the cars or even replace human drivers? Would people accept this development if the number of casualties were lowered considerably? What kind of combinations of affective computing could there be in the field of transportation?
What kind of long-term developments of information technology you would find most beneficial?
Honkela has conducted research on information retrieval and data mining methods based on Kohonen's self-organizing maps in the Websom project in the Neural Networks Research Centre of Helsinki University of Technology (1995-1998). He has served at VTT Information Technology as a project manager in Glossasoft project (EU Telematics, 1993-95) that developed methods and tools for dealing with multilinguality and cultural diversity in software development. In the 1980s Honkela had a research position in Kielikone project funded by Sitra Foundation. The project was developing a large-scale natural language interface for Finnish. Honkela has published a large number of scientific articles in the areas of natural language processing, information retrieval, neural networks and cognitive modeling. He is also a former long-term chairman of the Finnish Artificial Intelligence Society.
Professor Honkela's most recent research interests include development of methods for dealing with vocabulary problems in collaborative environments and information retrieval, computational pragmatics supporting situated and contextual information processing, and use of novel information representation methods and adaptation for information systems as well as development and scenarios for their use in the information society.