Welcome to the last section of the course!
The last section presented an approach to rational decision making in an uncertain static environment. This section presents an approach to rational decision making in an uncertain dynamic environment. In other words, we are going to combine the utility function with a dynamic Bayesian Network. The result is known as a Markov Decision Process, and it forms the basis of a lot of decision making models in AI.
We are then going to devote the last part of the course to AI and ethics. AI decision making algorithms already affect our daily lives quite profoundly: it is used in recruitment, mortgage lending, immigration, policing and more. So anyone working on models of decision making must be aware of the ethical issues and the challenges that we face in making AI trustworthy, transparent and accountable.
This folder consists of the following material:
- three videos of short lectures. They cover:
- Markov Decision Processes: Representation
- Markov Decision Processes: Computing Optimal Polices
- AI and Ethics
- Some required reading from Russell and Norvig.
Please watch the videos or attend the in-persion lecture. There is no quiz for this section, but as always, if there is anything you don't understand, then ask your question at the in-person lecture or post it on piazza.
Lecture 30 Slides: Whole!
30a: Markov Decision Processes: Representation
30a slides: 30a.pdf
30a video:
30b: Markov Decision Processes: Optimal Policies
30b slides: 30b.pdf
30b video:
30c: AI and Ethics
30c slides: 30c.pdf
30c video:
Required Reading
R&N Section 17.1–17.3 or NIE Chapter (17) "Making Complex Decisions", Sections 1–3.
NOTE: The abbreviation R&N refers to:
“Artificial Intelligence: A Modern Approach” Third Edition, Russell R & Norvig P, Prentice Hall, 2010 (R&N).
The abbreviation NIE stands for the following edition of the same book:
“Artificial Intelligence: A Modern Approach” Third Edition, Pearson New International Edition, Russell R & Norvig P, Pearson, 2014.