Category: intelligence

chatGBT about Rationality: Emotions, Mystik, Unconscious, Conscious, …

eJournal: uffmm.org
ISSN 2567-6458, 15.January 2023 – 16.January 2023
Email: info@uffmm.org
Author: Gerd Doeben-Henisch
Email: gerd@doeben-henisch.de

CONTEXT

This texts reflects some ideas following the documented chat No.4.as part of the uffmm.org blog.

Remark: See for a follow-up reflection the text of my post “chatGPT – How drunk do you have to be …” from 15./16.
February 2023.

Chatbots as Interfaces to the Human Knowledge Cloud?

Already at the end of the documented chat No.4 I had the impression, that an interaction with the chatbot chatGBT is somehow different compared to what most people until know have reported or stated in some reflected way about chatGBT.

In a first — and still a little bit vague — conclusion I have stated: “Apropos ‘rational’: that is a point which did surprise me really: as a kind of a summary it came out “that human rationality is composed of emotions, spiritual experience as well as conscious as well as unconscious cognitive processes. This is clearly not what most philosophers today would say. But it follows from the ‘richness of the facts’  which came as a resonance out of this chat. Not that the chatbot would have given this summary in advance as an important characterization of rationality, but as a human counterpart I could summarize all this properties out of the different separated statements [of chatGBT].”

And, indeed, the millions, if not yet billions, of documents in the world wide web are reflecting fragments of knowledge generated by humans which as a whole form a ‘cloud of knowledge’. The reflected echo of the real world through the medium of human brains is distributed in libraries and in the digital space. No individual person as such can make use of it; it is by far too big to be comprehensible.

Meanwhile search-algorithms can help us to make fragments of this ‘human knowledge cloud’ visible, but the search-results so far are ‘transformed’ in a way which is only of very limited use.

My encounter with chatGBT revealed some new glimpses of a possible new paradigm which perhaps wasn’t intended by openai themselves, but which seems now to be in reach: the individual brain has a limited capacity for ‘many documents’, but it has still an incredible ability to ‘transform’ billions of events into complex abstract patterns, inducing complex networks of relations, complex models, even complex theories.

If one looks to a chatbot like chatGBT as an ‘interface’ between a single human person and the ‘human knowledge cloud’, and this by only using ‘everyday language’, then — depending from the quality of the chatbot — this individual human can only with some ideas and questions ‘trigger’ those documents in the huge ‘human knowledge cloud’ which somehow are ‘fitting’ the triggering words. Thus this individual human person is step-wise encountering those fragments of the ‘human knowledge cloud’ which are in ‘resonance’ with his own words. In a bad case this interaction resembles those ‘echo chambers’ of the internet and the real world where people only get back what they put in.

But in a creative case the input of the individual human person can trigger fragments which are somehow confirming, but somehow non-confirming. This can be a trigger onto the the individual human person to generate some new ideas to feed back again the chatbot interface. While the ‘human knowledge cloud’ at a certain moment of time as such is ‘given’ and will not create ‘something new’, an individual person as an ever running learning process (in principle) could generate new aspects and therefore transform the feedbacks out of the ‘human knowledge cloud’ into some ‘new configurations’.

This principle as such is not new. The human brains are working since millions of years with a mechanisms, where the billions of abstract-concrete cognitive entities derived from perceptions and thinking are ‘stored’ in a more or less ‘unconscious space’ called memory, and only if some actual perception or thinking is ‘touching’ some of these stored entities by some ‘kind of similarity’ these will be re-activated and will become part of the actual perceiving and actual thinking. Thus ‘actual new things’ can become merged with ‘past stored things’ forming some new-past construct. This inherent duality of ‘new-past’ is the basis for that important human capability to be able to transcend the almighty presence and escape its clasp.

If we switch to the level of the human population then the ‘Human knowledge cloud’ plays the role of the ‘unconscious memory system’ collecting ‘past things’, and some kinds of chatbots (if they would be good enough) could play the role of the ‘merger mechanism’ bringing together the ‘past’ (given in the ‘Human knowledge cloud’ ) and the ‘presence’ (given in the actual living human person) and it is then up to the human person to generate some useful new-past patterns of thoughts and possible actions.

In case of the chat No.4 the human actor triggered lots of questions and suggestions to get ‘out of the human knowledge cloud’ those fragments which are somehow corresponding to the triggering words. And after about 8000 signs in the chat the chat ended up in these paragraphs:

Human: OK, this all sums up to the conclusion that human rationality is composed of emotions, spiritual experience as well as conscious as well as unconscious cognitive processes. Do I miss something important?

Machine (citing human documents): No, you have accurately summed up the components of human rationality. Additionally, human rationality is also composed of social and cultural influences, which can shape and inform our understanding of the world and our decisions. Additionally, human rationality is often impacted by our values and beliefs, which can provide a moral foundation for our behavior and decision making. Finally, human rationality is also impacted by our environment, which can provide a source of information and insight.

By itself the chatbot would not have been able to start with such a characterization. He can ‘reach out’ in the ‘Human knowledge cloud’, but the bot algorithm as such has not — as far as research can see at the moment — any kind of ‘creative rational power’ to transform the ‘given knowledge’ into some new ‘helpful’ knowledge. But at the other side, the human persons would not be able too to make use of ‘all the available knowledge’.

In the course of interaction the human questions could ‘drive’ the bot to collect ‘interesting facts’which could then become ‘accepted’ by the bot because they haven become ‘part of the chat’. Thus at the end of the chat the bot could accept that human rationality is composed of emotions, spiritual experience as well as conscious as well as unconscious cognitive processes. A human person ‘helped him’ to state this. This bot algorithm as such does not know anything and he cannot understand anything. Because chatbots — until now — do not possess real emotions, no real mystical experience, no unconscious or conscious human-like cognitive processes, they have no intelligence in the human format.

It is an open question what kind of ‘intelligence’ they have at all. Until know there is great number of ‘definitions’ around. No one is accepted as ‘that’ definition, especially the relationship between the ‘collection of machine intelligence definitions’ and the possible — also not really existing — collection of ‘human intelligence definitions’ is more or less unclear. Thus we are somehow ‘dreaming’ of intelligence, but nobody can really explain what it is …. We could seriously try, if we want …. but who wants it?

SIMULATION AND GAMING

eJournal: uffmm.org,
ISSN 2567-6458, 3.May 2019
Email: info@uffmm.org
Author: Gerd Doeben-Henisch
Email: gerd@doeben-henisch.de

CONTEXT

Within the AAI paradigm the following steps will usually be distinguished:

  1. A given problem and a wanted solution.
  2. An assumed context and intended executing and assisting actors.
  3. Assumed non-functional requirements (NFRs).
  4. An actor story (AS) describing at least one task including all the functional requirements.
  5. An usability test, often enhanced with passive or interactive simulations.
  6. An evaluation of the test.
  7. Some repeated improvements.

With these elements one can analyze and design the behavior surface of an  assistive actor which can approach the requirements of the stakeholder.

SIMULATION AND GAMING

Comparing these elements with a (computer) game then one can detect immediately that  a game characteristically allows to win or to lose. The possible win-states or lose-states stop a game. Often the winning state includes additionally  some measures how ‘strong’ or how ‘big’ someone has won or lost a game.

Thus in a game one has besides the rules of the game R which determine what is allowed to do in a game some set of value lables V which indicate some property, some object, some state as associated with some value v,  optionally associated further with some numbers to quantify the value between a maximum and a minimum.

In most board games you will reach an end state where you are the winner or the loser independent of some value. In other games one plays as often as necessary to reach some accumulated value which gives a measure who is better than someone else.

Doing AAI analysis as part of engineering it is usually sufficient to develop an assistive actor with a behavior surface  which satisfies all requirements and some basic needs of the executive actors (the classical users).

But this newly developed product (the assistive actor for certain tasks) will be part of a social situation with different human actors. In these social situations there are often more requirements, demands, emotions around than only the original  design criteria for the technical product.

For some people the aesthetic properties of a technical product can be important or some other cultural code which associates the technical product with these cultural codes making it precious or not.

Furthermore there can be whole processes within which a product can be used or not, making it precious or not.

COLLECTIVE INTELLIGENCE AND AUTOPOIETIC GAMES

In the case of simulations one has already from the beginning a special context given by the experience and the knowledge of the executive actors.  In some cases this knowledge is assumed to be stable or closed. Therefore there is no need to think of the assistive actor as a tool which has not only to support the fulfilling of a defined task but additionally to support the development of the knowledge and experience of the executive actor further. But there are social situations in a city, in an institution, in learning in general where the assumed knowledge and experience is neither complete nor stable. On the contrary in these situations there is a strong need to develop the assumed knowledge further and do this as a joined effort to improve the collective intelligence collectively.

If one sees the models and rules underlying a simulation as a kind of a representation of the assumed collective knowledge then  a simulation can help to visualize this knowledge, make it an experience, explore its different consequences.  And as far as the executive actors are writing the rules of change by themselves, they understand the simulation and they can change the rules to understand better, what can improve the process and possible goal states. This kind of collective development of models and rules as well as testing can be called autopoietic because the executing actors have two roles:(i)  following some rules (which they have defined by themselves) they explore what will happen, when one adheres to these rules; (ii) changing the rules to change the possible outcomes.

This procedure establishes some kind of collective learning within an autopoietic process.

If one enriches this setting with explicit goal states, states of assumed advantages, then one can look at this collective learning as a serious pattern of collective learning by autopoietic games.

For many context like cities, educational institutions, and even companies  this kind of collective learning by autopoietic games can be a very productive way to develop the collective intelligence of many people at the same time gaining knowledge by having some exciting fun.

Autopoietic gaming as support for collective learning processes
Autopoietic gaming as support for collective learning processes

 

 

THE BIG PICTURE: HCI – HMI – AAI in History – Engineering – Society – Philosophy

eJournal: uffmm.org,
ISSN 2567-6458, 20.April 2019
Email: info@uffmm.org
Author: Gerd Doeben-Henisch
Email: gerd@doeben-henisch.de

A first draft version …

CONTEXT

The context for this text is the whole block dedicated to the AAI (Actor-Actor Interaction)  paradigm. The aim of this text is to give the big picture of all dimensions and components of this subject as it shows up during April 2019.

The first dimension introduced is the historical dimension, because this allows a first orientation in the course of events which lead  to the actual situation. It starts with the early days of real computers in the thirties and forties of the 20 century.

The second dimension is the engineering dimension which describes the special view within which we are looking onto the overall topic of interactions between human persons and computers (or machines or technology or society). We are interested how to transform a given problem into a valuable solution in a methodological sound way called engineering.

The third dimension is the whole of society because engineering happens always as some process within a society.  Society provides the resources which can be used and spends the preferences (values) what is understood as ‘valuable’, as ‘good’.

The fourth dimension is Philosophy as that kind of thinking which takes everything into account which can be thought and within thinking Philosophy clarifies conditions of thinking, possible tools of thinking and has to clarify when some symbolic expression becomes true.

HISTORY

In history we are looking back in the course of events. And this looking back is in a first step guided by the  concepts of HCI (Human-Computer Interface) and  HMI (Human-Machine Interaction).

It is an interesting phenomenon how the original focus of the interface between human persons and the early computers shifted to  the more general picture of interaction because the computer as machine developed rapidly on account of the rapid development of the enabling hardware (HW)  the enabling software (SW).

Within the general framework of hardware and software the so-called artificial intelligence (AI) developed first as a sub-topic on its own. Since the last 10 – 20 years it became in a way productive that it now  seems to become a normal part of every kind of software. Software and smart software seem to be   interchangeable. Thus the  new wording of augmented or collective intelligence is emerging intending to bridge the possible gap between humans with their human intelligence and machine intelligence. There is some motivation from the side of society not to allow the impression that the smart (intelligent) machines will replace some day the humans. Instead one is propagating the vision of a new collective shape of intelligence where human and machine intelligence allows a symbiosis where each side gives hist best and receives a maximum in a win-win situation.

What is revealing about the actual situation is the fact that the mainstream is always talking about intelligence but not seriously about learning! Intelligence is by its roots a static concept representing some capabilities at a certain point of time, while learning is the more general dynamic concept that a system can change its behavior depending from actual external stimuli as well as internal states. And such a change includes real changes of some of its internal states. Intelligence does not communicate this dynamics! The most demanding aspect of learning is the need for preferences. Without preferences learning is impossible. Today machine learning is a very weak example of learning because the question of preferences is not a real topic there. One assumes that some reward is available, but one does not really investigate this topic. The rare research trying to do this job is stating that there is not the faintest idea around how a general continuous learning could happen. Human society is of no help for this problem while human societies have a clash of many, often opposite, values, and they have no commonly accepted view how to improve this situation.

ENGINEERING

Engineering is the art and the science to transform a given problem into a valuable and working solution. What is valuable decides the surrounding enabling society and this judgment can change during the course of time.  Whether some solution is judged to be working can change during the course of time too but the criteria used for this judgment are more stable because of their adherence to concrete capabilities of technical solutions.

While engineering was and is  always  a kind of an art and needs such aspects like creativity, innovation, intuition etc. it is also and as far as possible a procedure driven by defined methods how to do things, and these methods are as far as possible backed up by scientific theories. The real engineer therefore synthesizes art, technology and science in a unique way which can not completely be learned in the schools.

In the past as well as in the present engineering has to happen in teams of many, often many thousands or even more, people which coordinate their brains by communication which enables in the individual brains some kind of understanding, of emerging world pictures,  which in turn guide the perception, the decisions, and the concrete behavior of everybody. And these cognitive processes are embedded — in every individual team member — in mixtures of desires, emotions, as well as motivations, which can support the cognitive processes or obstruct them. Therefore an optimal result can only be reached if the communication serves all necessary cognitive processes and the interactions between the team members enable the necessary constructive desires, emotions, and motivations.

If an engineering process is done by a small group of dedicated experts  — usually triggered by the given problem of an individual stakeholder — this can work well for many situations. It has the flavor of a so-called top-down approach. If the engineering deals with states of affairs where different kinds of people, citizens of some town etc. are affected by the results of such a process, the restriction to  a small group of experts  can become highly counterproductive. In those cases of a widespread interest it seems promising to include representatives of all the involved persons into the executing team to recognize their experiences and their kinds of preferences. This has to be done in a way which is understandable and appreciative, showing esteem for the others. This manner of extending the team of usual experts by situative experts can be termed bottom-up approach. In this usage of the term bottom-up this is not the opposite to top-down but  is reflecting the extend in which members of a society are included insofar they are affected by the results of a process.

SOCIETY

Societies in the past and the present occur in a great variety of value systems, organizational structures, systems of power etc.  Engineering processes within a society  are depending completely on the available resources of a society and of its value systems.

The population dynamics, the needs and wishes of the people, the real territories, the climate, housing, traffic, and many different things are constantly producing demands to be solved if life shall be able and continue during the course of time.

The self-understanding and the self-management of societies is crucial for their ability to used engineering to improve life. This deserves communication and education to a sufficient extend, appropriate public rules of management, otherwise the necessary understanding and the freedom to act is lacking to use engineering  in the right way.

PHILOSOPHY

Without communication no common constructive process can happen. Communication happens according to many  implicit rules compressed in the formula who when can speak how about what with whom etc. Communication enables cognitive processes of for instance  understanding, explanations, lines of arguments.  Especially important for survival is the ability to make true descriptions and the ability to decide whether a statement is true or not. Without this basic ability communication will break down, coordination will break down, life will break down.

The basic discipline to clarify the rules and conditions of true communication, of cognition in general, is called Philosophy. All the more modern empirical disciplines are specializations of the general scope of Philosophy and it is Philosophy which integrates all the special disciplines in one, coherent framework (this is the ideal; actually we are far from this ideal).

Thus to describe the process of engineering driven by different kinds of actors which are coordinating themselves by communication is primarily the task of philosophy with all their sub-disciplines.

Thus some of the topics of Philosophy are language, text, theory, verification of a  theory, functions within theories as algorithms, computation in general, inferences of true statements from given theories, and the like.

In this text I apply Philosophy as far as necessary. Especially I am introducing a new process model extending the classical systems engineering approach by including the driving actors explicitly in the formal representation of the process. Learning machines are included as standard tools to improve human thinking and communication. You can name this Augmented Social Learning Systems (ASLS). Compared to the wording Augmented Intelligence (AI) (as used for instance by the IBM marketing) the ASLS concept stresses that the primary point of reference are the biological systems which created and create machine intelligence as a new tool to enhance biological intelligence as part of biological learning systems. Compared to the wording Collective Intelligence (CI) (as propagated by the MIT, especially by Thomas W.Malone and colleagues) the spirit of the CI concept seems to be   similar, but perhaps only a weak similarity.