Home Page

Course Lecturers

Christian Gibson, Tiberiu Lupascu

E-mail addresses

c.h.s.gibson@hva.nl  t.d.lupascu@hva.nl

Assignment 1 - The meaning of 'artificial intelligence'

Assignment 1 - Feedback

Assignment 1 - Example

Lecture 1

The course in the academic year 2006-2007 will be given by IIE lecturers Christian Gibson and Tiberiu Lupascu

Both Christian and Tiberiu are currently involved in research projects for the digital life knowledge center at the IIE, so we hope that their own experience will prove to be a valuable addition to the course material.

Lecture 1 was presented by Christian. The material contained in the slide show deals with the background to Artificial Intelligence. Most textbooks on the subject begin with an examination of the meaning of Artificial Intelligence. There is a reason for this.

'Intelligence' is a concept which is not easily defined. Perhaps because  lack of intelligence can used in a discriminatory fashion to evaluate human performance the word sometimes has a pejorative connotation. Whatever the reason, in ordinary discourse the meaning of intelligence seems to be flexible. Currently 'emotional intelligence' is an addition to the concept which is gaining general acceptance. The original Intelligence Quotient tests - although still commonly known as IQ tests - are referred to as General Factor tests by some psychologists, who claim that the tests do not actually measure 'intelligence' as such.

Lecture 1 addresses this question. We ask students to think clearly about what they understand by the word intelligence. We then look at how the word is used for animal behaviour. What is a 'highly intelligent' dog? Can the word be used to describe the behaviour of laboratory mice? And would it have any meaning to speak of an intelligent worm? Clearly the discussion of the meaning of intelligence is an important question.

The first serious reference to intelligent machines was made by Alan Turing in 1937 in a paper which he published on the concept of a universal machine. From the late 1940s Artificial Intelligence became a scientific discipline. The question as to whether a machine could actually think was circumvented by Turing. He claimed that in his imitation game if a machine could simulate responses to questions in such a way that it was indistinguishable from a real human being then to all intents and purposes the machine was intelligent. This lead to the famous Turing test.

What Turing was referring to was a concept which came to be know as 'weak' intelligence in contrast to 'strong intelligence'. Weak intelligence is simply simulation of intelligence whilst strong intelligence refers to intelligence which arises from the same thinking and learning processes which take place in the human brain. The problem is that nobody really knows exactly what processes are involved in thinking and learning. The models which we build with intelligent systems are claimed by many to actually demonstrate strong intelligence, but there are also other scientists who would dispute this.

Proponents of the strong intelligence thesis claim that there are at least three definitive factors involved in classifying a model as intelligent. These are:

- discovery

- learning

- adaptation

After dealing with the question of what we really understand by intelligent systems, lecture 1 continues with a short historical record of the developments in artificial intelligence. Finally, we refer to a number of paradigms which will be individually dealt with during the rest of the course lectures:

- expert rule-based systems

- frames and fuzzy logic

- neural networks

- Bayesian networks, knowledge discovery and data mining

- hybrid systems

After completing the lecture, the students form discussion groups for the workshop assignments.

Return

Assignment

Lecture 1

During the workshop, the students are shown examples of various 'more or less' intelligent systems. A video is shown of the 'intelligent' humanoid robot Asimo. The Microsoft paper clip, spam filters and similar programs are also considered. Students then discuss these examples within the group and attempt to answer three questions relating to these examples. The questions are:

1)     What is an intelligent system?

2)     What sort of problems kan be solved by intelligent systems?

3)     What are the limitations of intelligent systems?

 Return

How was the assignment executed?

In general the assignments were well thought out. One or two students took the easy way and simply stated a bald answer to the questions, without any insight into the thoughts and arguments which lead them to these answers.
Other students took a more serious approach and presented their arguments in the form of basic statements and observations with their reasoning and their conclusions set out in the document. The assignments will be assessed by the lecturers using the following criteria:

·         argumentation (what is the origin of the basic standpoints proposed?)

·         a clear relation between basic propositions and conclusions

·         structure (clearly recognisable argument form)

·         has the question been answered? (Did the student read the question correctly?)

·         creativity/originality

Not all the students succeeded in submitting an assignment. The consequence of failing to submit an assignment  is 2 penalty points in the final grade. In some circumstances, the lecturers will be prepared to accept assignments submitted after the deadline. As a rule, for late submissions 1 penalty point will be subtracted from the final grade.

Return

An example of a completed assignment is shown here:

Question 1.
Wat is een intelligent systeem?

Een systeem wil ik definiëren als iets - een object - dat activiteiten op een voorspelbare wijze uitvoert. Betekent dit dat een systeem geen willekeurig gedrag kan uitvoeren? Ik denk het niet. Dat zou te beperkend zijn. Een systeem bijvoorbeeld dat een willekeurig lampje laat branden is toch systematisch bezig. Als het systeem wordt ingeschakeld, dan laat hij een lampje branden. Dat is systematisch gedrag. Het feit dat het lampje willekeurig gekozen wordt is niet bepalend voor het classificeren van het object.

Nu kijk ik naar de kwalificatie ‘intelligent’. In de gewone taal wordt de term ‘intelligent’ gebruikt om de capaciteit van denken en begrijpen aan te duiden (zie o.a. Collins Essential English Dictionary). Wij gebruiken het woord niet slechts om menselijk gedrag, maar ook om dierlijk gedrag te kwalificeren. Als wij de betekenis willen verleggen om machinaal gedrag te kwalificeren, dan raken wij de kern van het begrip ‘intelligent systeem’.

Een intelligent systeem is volgens deze benadering:

“een object dat activiteiten uitvoert op een voorspelbare wijze en dat de capaciteiten bezit om te kunnen denken en begrijpen”.

Question 2.

Wat voor problemen kunnen door intelligente systemen oplost worden?

Vanuit deze vraag rijst de vraag hoe ik problemen kan classificeren. Wij kunnen problemen kwalificeren door:

X      Complexiteit

X      Urgentie

X      Oplosbaarheid

X      Abstractieniveau

De vraag welke problemen in aanmerking komen voor oplossing door intelligente systemen is een praktische vraag. Het vereist ook inzichten in de verschillen tussen intelligente systemen en menselijke intelligentie. Kort samengevat berusten deze verschillen m.i. meer op complexiteit dan op wezenlijke fundamentele verschillen. Ik maak hier ook geen onderscheid tussen ‘sterke’ en ‘zwakke’ intelligentie. Intelligente systemen zijn beperkt door de beschikbare technologie (volgende vraag). Wel kunnen intelligente systemen onbeperkt voor 24 uur per dag worden ingezet. Met het oog op de bovenstaande kwalificatie van problemen concludeer ik dat:

X      minder complexe

X      minder urgente,

X      concrete problemen

X      die voor mensen moeilijk oplosbaar zijn

het meest in aanmerking komen voor oplossing door intelligente systemen.

Question 3. 

Wat zijn de beperkingen van intelligente systemen?

Deze vraag is nauw aan de voorgaande vraag verbonden. Ik zie technologische beperkingen als de belangrijkste beperking van intelligente systemen. Ik zie geen principiële reden waarom intelligente systemen in de toekomst niet in staat zouden zijn om alle menselijke taken over te nemen. Ik verwerp hiermee de stelling dat de mens  oftewel de menselijke geest, wat dat ook moge zijn  in een fundamenteel, niet na te bootsen wijze, van de machine verschilt.

Return