IFCO 2735 - How AI Works: Intelligence, Learning, Society

This course provides an overview of the field of artificial intelligence (AI), with an emphasis on the intersection of its social and technical dimensions. The course has two main objectives: (1) to familiarize students with the core concepts and tools of statistical learning; and (2) to enable students to become informed and critical participants in public discourse about the history, organization, and impact of AI systems. Some key questions we will focus on include: What is “intelligence”? What does it mean to “learn”? Where does “data” come from? What are the goals of statistical learning methods, and what are they used for? How do we evaluate whether a learning system “works” for its intended purposes? We will begin with an introduction situating AI in the field of statistical learning. Some facts about statistical reasoning and some examples of statistical fallacies will be discussed. For several machine learning methods, we will then explain how they work: we will focus on clustering, linear models, discriminant analysis, logistic regression, neural networks, nearest neighbors and tree-based methods. Students will be invited to test in practice on their own laptops the methods discussed in class, using the open-source statistical software R. Tutorial sessions on statistical learning during the semester will provide additional opportunities to practice recognizing the design and purpose of statistical learning techniques. Other subjects that will be discussed include the history of AI as a topic of scientific and technological development; the relationship between AI and measures of human intelligence; the scientific evaluation of modern AI systems; regimes of data collection in modern societies; discrimination and social inequality in automated decision-making; and popular depictions of AI in news and entertainment media. Assignments will help students understand the capabilities and uses of AI systems in detail, and to evaluate applications of AI in a thorough, detail-oriented, and critical way.

Alexander KINDEL,Aurélie FISCHER,Marina GOMTSYAN

Cours magistral seul

English

Required readings, quiz, and final project. Students are invited to begin work on the final project early and will have opportunities to receive feedback on their ideas throughout the semester.

It is an introductory course. While there are no prerequisites in mathematics or statistics, this course has a specific focus on introducing students to the technical and mathematical foundations of AI systems. Consequently, it is important for students to be interested in this aspect of AI and to be eager to understand how things work in technical detail, in addition to being interested in the historical and sociological dimensions of AI.

Spring 2024-2025

Quiz (40%) & Project (60%) Toward the end of the semester, there will be a quiz on the most statistical part of the course in class (duration 45min). The quiz will address some concepts of statistical reasoning, with questions about the purpose and behavior of statistical learning algorithms, about recognizing and distinguishing different kinds of learning problems on examples. The final project asks students to practice producing and evaluating AI-generated text. The project has two parts, each worth half of the final project grade: First, working in small teams (2-4 students), you will develop and execute a strategic plan for using AI to generate a short essay that answers a question we will select collectively at the beginning of the term. Strategic plans will describe your approach to generating, editing, and recombining text as well as your approach to evaluating the quality of the final product. Second, working individually, you will compare the essays in a structured evaluation exercise that asks you to rate the essays along several dimensions, as well as to provide a few short responses (not using AI) that explain the evaluative criteria your ratings reflect. The final project will be completed outside of class. The strategic plan is due week 7; final AI-generated essays are due by week 10, and the final evaluation is due at the end of the semester.