Course syllabus

Course memo

TME285 / FIM762 Intelligent agents lp3 VT21 (7.5 hp)

The course is offered by the department of Mechanics and Maritime Sciences

General information

This year, with the ongoing pandemic, the course will be given in a different way than usual. There will be no classroom lectures. Instead, each lecture will consist of a video with a recorded presentation of the lecture (uploaded a few hours before the scheduled time of the lecture; see below), directly followed by an interactive session on Zoom, where I will discuss the contents of the lecture with the students, and also answer any questions that the students might have. In addition, students are welcome to ask questions or give feedback at any time (see also Contact details below).

Note: Due to the pandemic, we have had to make very large changes in the course format  (since we cannot, as previous years, have long work sessions in the classrooms). Therefore, the course material (which is being modified and adapted accordingly) will have to be provided gradually during the course.

Contact details

During the course, we strive to be available as much as possible. You are welcome to ask questions at any time, either in the Zoom session associated with each lecture or at other times, and you may also ask questions via e-mail or telephone. You are always welcome at our offices. However, due to the pandemic, we can only receive one (1) student at a time, and (of course) only if you are symptom free. You do not need to make an appointment, but since we are not always in our offices (in the current situation, we mostly work from home) it's a good idea to first check that we are there (e.g. via e-mail or telephone).

Lecturer and examiner: 
Professor Mattias Wahde, Tel: 772 3727, e-mail: mattias.wahde@chalmers.se
Course assistant: 
Björnborg Nguyen, e-mail: bjornborg.nguyen@chalmers.se

Finding our offices: Go to Hörsalsvägen 7, enter the building (nya M-huset), so that you have Café Bulten on your right as you enter. Then go up one flight of stairs, and enter the corridor (Vehicle Engineering and Autonomous Systems). If the door is locked, please dial the appropriate extension, as shown in the list beside the door (e.g. 3727 for Mattias).

Course purpose

The aim of the course is for the students to gain knowledge regarding intelligent agents in general, and conversational agents, in particular. Students will learn about both chatbots and (most importantly) task-oriented agents. Emphasizing interpretable systems (rather than black box models), the course covers both theoretical aspects and the practical aspects related to actual implementation of conversational agents. Applications and ethical aspects are considered as well.

Schedule

The course schedule is given below. The "Publish time" refers to the time when the pre-recorded video (for the lecture in question) will be uploaded, when applicable. (see the Modules page).  For each lecture there will also be a live Zoom session, starting at the nominal lecture time (see the corresponding column). The lectures can also be found in TimeEdit, but note that the duration of the lectures is as shown below.

The link to my Zoom page is:

https://chalmers.zoom.us/j/2039626788
passcode: 2020FFR105    (unchanged from the previous course)

Björnborg's Zoom page is at:

https://chalmers.zoom.us/j/2430787247

passcode: 2020FFR105 (i.e. same as mine).

Date	Publish time	Zoom session	Content
20210119	16.00 (20210118)	08.00-08.45	Course introduction and motivation; Brief description of the topics covered in the course; Introduction to conversational AI.
20210120	08.00	13.15-14.00	Conversational agents. Chatbots vs. task-oriented agents. Interpretable AI vs. blackbox models. Historical overview of conversational agents.
20210126	16.00 (20210125)	08.00-08.45	Chatbots (Pattern-based, information retrieval-based, generative) Handout of Assignment 1 (chatbots)
20210127	08.00	13.15-15.00	Background on C# .NET programming. Description of the C# .NET libraries and applications used in the course,
20210202	---	08.00-08.45	Assignment work session (teachers available on Zoom throughout the session)
20210203	---	---	No lecture
20210209	16.00 (20210208)	08.00-08.45	Task-oriented agents (finite-state, frame-based, model-based). The pipeline model. Intent detection, cognitive processing
20210210	08.00	13.15-15.00	Task-oriented agents: DAISY (part 1) Handout of Assignment 2 (task-oriented agents)
20210216	16.00 (20210215)	08.00-08.45	Task-oriented agents: DAISY (part 2).
20210217	08.00	13.15-15.00	Input and output modalities (speech synthesis, speech recognition, character animation, video processing, data acquisition). C# .NET: Advanced topics. Handin of Assignment 1
20210223	16.00 (20210222)	08.00-08.45	Applications (and evaluation) of conversational agents.  Handout of Assignment 3 (short essays on (a) interpretabilty, (b) ethical implications of conversational agents)
20210224	---	13.15-17.00	Assignment work session (teachers available on Zoom throughout the session)
20210302	16.00 (20210301)	08.00-08.45	Ethical aspects of conversational agents.
20210303	---	13.15-17.00	Assignment work session (teachers available on Zoom throughout the session)
20210309	---	08.00-09.45	Assignment work session (teachers available on Zoom throughout the session)
20210310	---	13.15-17.00	Assignment work session (teachers available on Zoom throughout the session). Handin of Assignment 2 and Assignment 3

Course literature

The course literature will consist of lecture notes (also with some links to various scientific papers), which will be provided gradually during the course.
Note that, due to the large changes in the course, the compendium from previous years is now outdated. All lecture notes will be provided on the Modules page.

Course design

The course consists of a sequence of lectures, usually two per week (but note that there are some exceptions; see above). Several sessions, especially towards the end of the course, are work sessions, during which the students will work with their assignments. 

Changes made since the last occasion (2020)

The course has been modified significantly, partly as an adaptation to the pandemic, and partly in order to emphasize more strongly the topic of human-machine dialogue.

Learning outcomes

After completion of the course the student should be able to...

• Implement chatbots (in C# .NET)
• Implement a task-oriented agent in (in C#. NET) 
• Implement various input and output modalities (e.g. speech) in conversational agents
• Describe and compare various different (applications of) intelligent agents.
• Describe and discuss interpretable AI, and be able to contrast interpretable models with black box models
• Describe and discuss the ethical implications of conversational agents.

Examination 

Examination: There will be three assignments, worth a total of 100 p, distributed as follows:

Assignment 1 (Chatbot): 20 p  (source code and report)

Assignment 2 (Task-oriented agent): 60 p (source code and report)

Assignment 3 (essays): 10p +10p (two reports)

The dates for handing in the assignments can be found in the schedule above. The detailed requirements will be given in connection with each assignment. In order to pass the course, students must hand in satisfactory solutions to all three assignments. Grades will then be set as follows:

Grade 3: Up to 60p

Grade 4: [61, 80] p

Grade 5: [81-100] p

Note: All deadlines are at 23.59.59 on the specified dates, and it is the time when the submission is received that counts. Submissions that are handed in late will receive a lower score, as described in each assignment document. The possibility of submitting assignments will be closed at midnight on 20210312.

Re-examination, grade improvement etc.: It is possible for students to improve their grade by resubmitting assignments. However, after the final submission deadline during the course (20210314) resubmission is only allowed in connection with the re-exam periods in August (2021) and January (2022). 

Note: All submissions must be handed in via the Canvas page. E-mailed submissions will not be considered.