Syllabus

Course Meeting Times

Lectures: 2 sessions / week, 1.5 hours / session

Prerequisite

6.034 Artificial Intelligence

Course Description

This course analyzes seminal work directed at the development of a computational understanding of human intelligence, such as work on learning, language, vision, event representation, commonsense reasoning, self reflection, story understanding, and analogy. It reviews visionary ideas of Turing, Minsky, and other influential thinkers and examines the implications of work on brain scanning, developmental psychology, and cognitive psychology. There is an emphasis on discussion and analysis of original papers; students taking the graduate version complete additional exercises and a substantial term project.

Purpose

6.803/6.833 is designed to help you learn about progress toward the scientific goal of understanding human intelligence from a computational point of view.
It complements 6.034, because this course focuses on long-standing scientific questions, whereas 6.034 focuses on existing tools for building applications with reasoning and learning capability.

Why you should not take this course

  • You want a subject that will help you get a summer job. (Take a subject in machine learning instead. This course focuses on cognition and what makes humans special. It is not about popular statistical mechanisms.)
  • You cannot commit to on-time attendance and committed reading. (Because of the emphasis on reading and discussion, and the limitation on enrollment, regular attendance is obligatory, along with commitment to reading the papers. If you cannot picture yourself in class at 9:30 am or 11:00 am, every Monday and Wednesday, you should not register, so as to make room for others who would otherwise be excluded because of the enrollment limitation. A corollary is that you probably should not register for this course if you are taking five subjects or course equivalents, such as UROP. You definitely should not register if you are involved in a startup or you are taking six or more subjects or subject equivalents.)
  • You are not interested in understanding human intelligence from a computational point of view. (Believing that both mind-stretching and near-miss learning are educationally useful, some of the papers I have selected are boring, stupid, or nearly unintelligible. One goal of the subject is to develop the skill of gleaning useful ideas from such papers, but if you have little or no interest in understanding human intelligence, you should not subject yourself to the necessary reading. For more detail on what you will need to read, have a look at the previous year’s schedule.)
  • You already know everything you need to know about communication. (About one-third of the subject is devoted to discussing how to package ideas orally and in writing. You need to be enthusiastic about practicing the skills taught with a positive attitude.)
  • You are enrolled in another limited-enrollment artificial intelligence (AI) subject or an AI subject whose enrollment should be limited. (Alas, advanced AI subjects are scarce, and fairness dictates that they should be offered as broadly as possible. This fairness goal must be balanced, however, against the need to keep some of them small. If you are just generally interested in AI, you should take one of the graduate lecture-based subjects.)

Why you should take this course

  • You should take this course if you want to learn about the enterprise of explaining intelligence from a computational point of view. When you have finished the subject, you will understand the powerful ideas behind an optimistic view of what will be discovered in the next decade.
  • You should take this course if you want to develop a foundation for making personal contributions toward reaching the goal of understanding intelligence. When you have finished the subject, you will know about intriguing ideas begging for extension.
  • You should take this course if you want to learn how to dig the salient ideas out of a research paper without distraction by minutiae. When you have finished the subject, you will have learned to identify big ideas and ignore detritus.
  • You should take this course if you want to learn to present complex ideas effectively, as if you were presenting a thesis, delivering a job talk, chatting with a high-ranking official at breakfast, or making a presentation to a potential customer or venture capitalist. When you have finished the subject, you will have learned about heuristics that will improve your ability to do all these.

Why this course can be viewed as a humanities subject

  • This course is about computational theories of human thinking. Hence, it can be viewed as a special kind of psychology subject.
  • This course is about ferreting big ideas out of original sources through thoughtful reading, writing, and discussion. Hence, it can be viewed as a special kind of literature subject.
  • This course is about packaging ideas into a variety of formats, including abstracts, conclusions, slide shows, press releases, proposals, reports, letters, and conversation. Hence, it can be viewed as a subject in communication.
  • This course is about how to empower ideas through a clear statement of vision, an enumeration of concrete steps toward the vision, an articulation of new results with clarifying details, and a statement of contributions. Hence, it can be viewed as a subject in persuasion and leadership.

Level

6.803 is the undergraduate version of the course, and 6.833 is the graduate version. The two differ in that 6.833 may require you to attend some extra classes and will require you to complete a substantial term project. Both meet together ordinarily.

The graduate version forms a bridge between 6.034 and design/project/thesis work in artificial intelligence.

Content

The content of the course is largely based on papers identified in an informal survey of representative AI leaders, who were asked what has most influenced the way they think about human intelligence. The papers mentioned tend to fall into the following categories, ranked by frequency:

  1. Visionary thinking by the giants
  2. Computational models of perception and cognition
  3. Powerful computational ideas
  4. Neuroscience and human behavior

Style

  • You read parts or all of one or two papers for each class.
  • You discuss the content of those papers in class, occasionally with the authors.

The following mechanisms are used to ensure that you read the papers and absorb the material:

  • Homework, consisting of either short answers to questions about the papers or the preparation of abstracts, slide shows, and other forms of communication.
  • Verbal questions, often asked of random students during class.

Limit on Enrollment

Because of the emphasis on reading, discussion, and presentation, enrollment is limited.

Credit and Projects

Doing a substantial project is required for graduate credit.

Frequently Asked Questions

What should I do if I have to go to a job interview?

Explain to the interviewer that education comes first. Invite the interviewer to come to class with you.

What should I do if I haven’t done the assignment?

Show up anyway because the communications discussions are cumulative and there is no written backup. Just tell me you are not prepared to participate in the discussions either before class or when first called upon.

Course Info

Learning Resource Types

assignment Written Assignments