This course will explore whether and how generative AI can be developed to support human values and promote human autonomy, and how the context of the deployment of AI may impact answers to this question. Entire industries are being transformed by AI technology, much of which is driven by the recent meteoric advances in generative AI: the variety of AI that produces full content, such as documents, images, speech, and video. These advances have enabled many people to do things they previously were incapable of - such as essay writing or adding special effects to home movies - but have also brought about a series of ethical questions around their development and use - such as the role of AI in Hollywood brought into the public eye through the 2023 writer’s strike. These developments raise fundamental questions around whether it is even possible to develop generative AI technology that empowers rather than replaces people, and which serves human values such as rights, justice, and dignity. It also raises the question: Is generative AI different from other technologies that can be used toward both positive and negative ends? Different disciplines have different ways of answering questions around human values, whether it’s the social sciences, the humanities, or computer science. In this course, you will not only learn about the challenges of developing values-centered generative AI technology, but also actively participate in crafting tomorrow’s solutions.
Week Starting | ||
---|---|---|
01/27 | Welcome - What is intelligence? | Machine learning and computer intelligence
Reading: Artificial Intelligence: A Guide for Thinking Humans (Mitchell), Chapters 1-3 Writers’ Strike (Guardian) |
02/03 | Exploring design of generative AI Systems Watch: "Ethics and AI" from code.org/ai Read: The Turing Test (Stanford Encyclopedia of Philosophy), sections 1, 2, and 5 Read: AI and Music (NYT) |
Neural networks and deep learning I Watch: "Machine Learning" and "Neural Networks" from code.org/ai Read: Artificial Intelligence: A Guide for Thinking Humans (Mitchell), Chapter 6 Read: Incident 339 (and Slate) |
02/10 | Deep Learning II |
How do Generative text models work Part 1 Read: LearnPrompting Basics ("Introduction to AI" through "Formalizing Prompts" (you may skip "Embeds") Read: Incident 314( and TechCrunch) |
02/17 | How do Generative text models work Part II | How do Generative Image and Video models work
Reading: An Introduction to Diffusion Models for ML (Acharya) Optional reading (for more math): Introduction to Diffusion Models for Machine Learning Read: Incident 351 ( and NY Post) |
Instructor: Mohammad Nayeem Teli (nayeem at cs.umd.edu)
Office: IRB 2224
Office Hours: F 11 AM - 12 PM
Name | Email (at umd.edu) |
---|---|
Ziyue Li | litzy619 |
Hsien-Yu Meng | mengxy19 |
Alex Stein | astein0 |
Monday | Ziyue: 11:00 AM - 1:00 PM |
Tuesday | Hsien-Yu: 2:00 PM - 4:00 PM Alex: 10:00 AM - 12:00 PM |
Wednesday | Ziyue: 11:00 AM - 1:00 PM |
Thursday |
Hsien-Yu: 2:00 PM - 4:00 PM Alex: 10:00 AM - 12:00 PM |
Please note that a TA may need to leave 5 minutes before the end of the hour in order to go to his/her class. Please be understanding of their schedules.
Homework | Due Date* |
---|