The final exam will be a 3-hour open-book on-line exam (Canvas).

Questions will focus on your understanding of the material that has been presented in the course and can be found in the assigned readings and lecture notes, with an emphasis on the main concepts and topics and your ability to holistically comprehend these.

If you are unsure of which topics are the main ones, and which are sub-topics or less important, look at the lecture notes: If it's mentioned there it is important. If it's mentioned more than once it's even more important. But above all, your comprehension of the relationships between topics, and your ability to put this in context with the present state and future of AI, is what this test focuses on.

Below are some example questions. Please note that this is not an exhaustive list of the types of questions that may appear on the final exam, it is representative of some questions that may appear on the final exam, and are provided to help you prepare for the final exam. These are examples only, the specific question below may (or may not) appear - in this or modified form - on your final exam.

However you answer and whatever you write, make sure you present strong and clear arguments for your answer, referring back to the most relevant material covered in the course.

1. For a task of space junk cleanup, the European Space Agency (ESA) is planning to build an autonomous robotic satellite craft that can obliterate orbiting space junk with a powerful laser. Each encounter of the craft with a piece of junk must be documented through a camera and uploaded to a computer as soon as it is finished. The laser must be tuned based on the junk's weight, density, and speed of motion relative to the trash satellite for each encounter. For most of the junk that the craft will encounter there exists no records of size, weight, speed, or other information. Everything onboard the craft, including its small electric-power thrusters, is powered by its batteries, which are charged via its solar panels. The craft will stay in orbit for at least twenty years.

1a. ESA must decide what kind of control system to use for the craft. They ask you to write a report with an analysis of the requirements of the control system and recommendations for how it should be designed. They are especially interested in novel ideas and approaches. (i) What will your report say about the control system requirements? (ii) What will your recommendation look like? (Make sure you present strong and clear arguments for your answer.) [7%]

1b. A few months later ESA adds the following issue: They have determined that the is so much space junk that the craft will likely have to fire at multiple objects simultaneously. This puts extra strain on the scheduling of events, including charging batteries, keeping track of junk items, making sure the craft is not hit by debris, among other things. The company wants to develop a system based on the AGI-aspiring NARS system to control the photography unit. They want you to propose a way to do this. Please give an outline of your recommendations for how to use NARS for this purpose. Make sure you list the strengths and weaknesses of NARS, to help ESA come to the right conclusion of whether to choose NARS for this purpose.[7%]

2. How does AGI differ from “narrow AI”? [5%]

3. What is cognitive growth? Give some examples of cognitive growth. Why is cognitive growth an issue in constructivist AI but not in “good old fashioned AI” (GOFAI) ? [10%]

4. Why is architecture important in constructivist AI? [5%]

5. With specific reference to “task-environments”, discuss how you would approach designing a methodology to evaluate the generality of an AI system.

6. How does the concept of 'control' relate to the concepts of 'goal', 'plan', and 'model'? [10%]

7. What are some of the main methodologies used for building AI systems? Describe two limitations that they inherently have and explain how they might be overcome. [10%]

2021(c )Kristinn R. Thórisson

EOF