AlphaGo chronicles the now famous Go match between professional Lee Sedol (one of the greatest players of the game) and the artificial neural network AlphaGo designed by Google DeepMind. This film is interesting on a number of levels for our class, and it touches on the topics of machine learning and model search, which we have been discussing in recent weeks. In this assignment you will have a chance to express your own views on some of the ideas presented in the film. The verification of many of the claims made by figures in the film about AI is not straightforward, making this an excellent launching point for discussion.

Briefly respond to the following prompts in a PDF document (a few sentences for each one is fine, but feel free to write as much as you'd like). These responses will be summarized and discussed in class following the film. This film response is worth 50 points, which count towards your total class participation grade. It is due at 11:59pm on 9/23.

Prompt 1: Are games a good way to assess artificial general intelligence?

Demis Hassabis routinely describes Google DeepMind's work as the pursuit of artificial general intelligence (i.e., the replication of human intelligence). But DeepMind's pursuits are largely restricted to game playing. Are games a good way to assess artificial general intelligence? Why or why not? Explain your reasoning.

Prompt 2: Did Lee Sedol overestimate AlphaGo's abilities?

Nate Silver's book The Signal and the Noise recounts another famous match-up between an AI and a champion game player: IBM's Deep Blue Chess Computer and Gary Kasperov. Silver alleges that a bug-generated mistake made by Deep Blue was interpreted by Kasperov to be a sign of superior intelligence, which forced a change in his gameplay that resulted in his subsequent poor performance. Throughout AlphaGo, we see instances of surprising moves from the neural network, and an increasingly harried Sedol. Is it possible Sedol lost because he was simply psyched out? And if so, what does this tell us about this style of AI evaluation?

Prompt 3: What makes Go so different from other games?

Go maintains a reputation of being significantly more difficult to tackle with AI compared to other traditional board games. What, exactly, is so different about it?

Submitting Your Film Response Document

Send a copy of your response to the above prompts (in PDF format) via email to Sophia (sabraha2@nd.edu). Prof. Scheirer will grade these and lead a discussion based on the thoughts of the class during the week of 9/23.

If you have specific questions about this assignment before it is due, please reach out to Prof. Scheirer or Sophia during office hours, or via slack. This assignment is worth 50 points.