What makes you so excited about teaching Artificial Intelligence?
AI is exciting to teach because it’s everywhere and it’s changing the world so quickly! We all interact with AI constantly — everything from searching the web to playing video games. AIs plan our driving directions, filter our spam, and focus our cameras. AI lets you guide your phone with your voice and read foreign newspapers in English. But perhaps even more importantly, beyond today’s applications, AI will be at the core of many new technologies that will shape our future. From self-driving cars to household robots, advancements in AI help transform science fiction into real systems. In this course we are teaching many of the core AI concepts and principles that will drive this transformation.
What materials are covered in this course?
This course is about programming systems that make good decisions. For AI systems and humans alike, there are two main components to decision making: simulation and memory. The first half of the course is about simulation — modeling the future. Here, agents tell the good decisions from the bad ones by running them through a model of the world and planning ahead. By considering possible outcomes, agents search for sequences of actions that achieve desirable outcomes in the end. The agents’ worlds could be deterministic, they could have uncertainty, and they could have adversaries. Each of these settings comes with its own key ideas, abstractions, and algorithms. We will study modern-day techniques to optimize our agent’s behavior in each case.
The second half of the course is about memory — generalizing from past experiences. Here, we will consider how agents can act intelligently by accumulating experiences over time. The formal setting is reinforcement learning, where agents must experiment and learn, trading off between exploration (to acquire new knowledge) and exploitation (of current knowledge). One key idea throughout is that computation can produce complex behaviors from elegant, uniform algorithms.
Do MOOC students get to program an AI?
There will be three programming projects where students can program the AI for variations of the classic game of Pacman (™ Namco-Bandai). In each project, students will write the AI and watch as the the algorithms from class result in complex game-playing behaviors. Students’ AIs will find paths, dodge enemies, and learn from experience!
How does the MOOC relate to the on-campus offering, and what is this new feature that will allow students to assess themselves like Berkeley students?
CS188.1x corresponds to the first half of our on-campus course. As in our past two MOOCs, students will have access to video lectures, quizzes, online homework, online projects, and an online exam. As completely optional bonus material, which won’t contribute to the course grade, we will also release the PDF representing the actual Berkeley exam for the first half of the Berkeley course. If they choose, MOOC students will be allowed to complete the exam and submit answers into Gradescope, the system we use for grading Berkeley students’ exams. This way, MOOC students will be able to simulate taking a Berkeley exam with pen and paper, and then assess their performance through this system.
What have you enjoyed the most about your past MOOCs?
It’s great to be able to reach so many students from so many different backgrounds, and it’s always a pleasure to see the contributions from the students to the discussion forums. We also really enjoy how providing the course for an online audience forces us to keep making it better!
What should students do to prepare for this course in advance?
The course assumes familiarity with object-oriented programming, recursion, Python (or ability to learn Python quickly), data structures, and probability. If it has been a while since touching on these topics, it might be worthwhile to refresh them.
Sign up for Artifical Intelligence today and start your AI journey with BerkeleyX.
By Rachel Lapal, Communications Manager
16 Mar 2020