The Science and implications of Generative AI
Harvard Kennedy School
This is a collection of videos and course content from a Harvard Kennedy School course on generative AI. The course is designed to equip students with a solid introduction and understanding of how generative AI works, how to use it, and the larger opportunities and challenges it poses for society.
The course explores the inner workings of modern AI models, particularly focusing on Large Language Models (LLMs) such as ChatGPT, while engaging in practical applications of these systems. We will assess the advantages and risks of deploying generative AI in a variety of contexts, learning to harness AI technology responsibly for the benefit of society. Throught the course, we also engaged professionals from the field of AI in a series of discussions we called Beyond the Classroom.
Click on the SESSION links below to experience each class, including pre-class work, videos, and in-class activities.
🎥 If instead you want to watch all course videos as a YouTube playlist, you can do so here.
Unit 1: How generative AI works (Science)
SESSION 1: INTRODUCTION TO GENERATIVE AI [90 MIN]
In this section, we will start with a general introduction to Generative AI and LLMs, and then explore an application an University Admissions: can you tell which essay has been written by AI?
SESSION 2: DEEP NEURAL NETWORKS [60 MIN]
What is a deep neural network, and how does it really work? Learn the fundamental concepts and explore the key functionalities in this section.
SESSION 3: THE ALIGNMENT PROBLEM [70 MIN]
How can we make sure that AI systems pursue goals that are aligned with human values? Learn how to detect and analyze misalignment, and how to design aligned systems.
Unit 2: How to use generative AI (Individuals, Organizations)
SESSION 4: PROMPT ENGINEERING [90 MIN]
How can we guide Generative AI solutions to give us what we are really looking for? In this class, we learn to master the main tools and techniques in Prompt Engineering.
SESSION 5: BEYOND CHATBOTS [70 MIN]
How can we tailor our chatbots to our specific goals, and how can we use different modalities to achieve them? Learn about system prompts, retrieval augmented generation (RAG) and fine-tuning.
SESSION 6: WHEN AND HOW TO USE GENERATIVE AI [60 MIN]
How can we decide whether AI can help us achieve our specific goals, and whether we should use it? In this class, we discuss some criteria to help us make these decisions, and test them through the lenses of examples.
SESSION 7: USING AI TOOLS IN PRACTICE: A CASE STUDY [120 MIN]
What are some of the most effective ways of using AI in political campaigns? In this session we put everything we've learned so far together and explore different use cases for AI, from social media engagement to debate preparation for political candidates.
Unit 3: The implications of generative AI (Society)
SESSION 8: RISKS OF GENERATIVE AI [90 MIN]
What are the biggest risks of generative AI? In this class, we start looking into four main categories of risk (known limitations, misuse, society-wide disruptions and existential risks) and designing mitigation strategies.
SESSION 9: GENERATIVE AI AND COPYRIGHT [80 MIN]
What is copyright law, and what does it have to do with generative AI? In this session, we discuss the rationale behind copyright, and we discuss whether model training or AI outputs infringe copyrights, or if they can be copyrighted themselves.
SESSION 10: GENERATIVE AI AND THE FUTURE OF WORK [110 MIN]
What is the relationship between AI and the economy? In this session, we start thinking about what effect AI will it have on jobs and inequalities, and we investigate policy solutions that might be more effective.
SESSION 11: MIS/DISINFORMATION [90 MIN]
In what ways is AI facilitating the process of creating and spreading misinformation and disinformation content? In this session, we use practical examples to investigate the most critical risks and the most promising solutions.