AI Related Discussion Questions for GEB
As an interesting exercise to myself and my Artificial Intelligence course, I had the undergrads read the first few chapters of Gödel Escher Bach alongside the regular coursework and respond to discussion questions, since the book is particularly interesting in this age of LLMs in my opinion. I figured I'd share the questions in case any book clubs or other courses wanted to work with them. Let me know if you particularly like these and I can work on some for the other chapters as well.
Intro
- Find a recording of one of the musical pieces introduced by Hofstadter; what do you think of it?
- There is a lot of talk about AI art and music now; relate the ideas from the book to these new advances in AI. Could an AI today make something like A Musical Offering?
- What makes this chapter a "Musico-Logical" Offering (a pun on A Musical Offering)?
- In your own words, what is a "strange loop"? Or what do you think he means?
- Create your own artwork that encompasses a strange loop. If you use an AI art tool, what prompt did you use to encode self reference? Was it successful? If you did it the old fashioned way, what was your thought process?
- How relevant are Hofstadter's points about AI to the current day?
Ch 1/2
- How did you approach the MU puzzle? How would you solve it using code?
- Could an AI "jump out of a system" without help from the person who coded it?
- What is the difference between manipulating symbols and understanding meaning?
- If a computer produces a correct output, does it “know” the answer? What role does the observer play in assigning meaning?
- Find a modern AI example (ChatGPT, Midjourney, music models, DeepMind, etc.). Which parts feel like symbol manipulation, and which parts feel like “real understanding”? Defend your view.
- Where do you think meaning comes from in humans?
- What would convince you that a machine truly understands something?
Ch 3/4
- Write some code to figure out the word that has ADAC in it? What about the word that starts and ends with HE. What does this have to do with the reading?
- Research online regarding "recursively enumerable vs recursive sets"; this is a topic that comes up in Theory of Computation classes. What can you prove about the latter vs the former?
- What’s an example of a field where relaxing a core assumption (like Euclid’s parallel postulate) opened up a completely new way of thinking?
- The record player metaphor is a bit dated unfortunately. Can you come up with another metaphor that is able to map to Gödel's incompleteness theorem?
- Figure vs. ground: Are there concepts in AI or computer science that are easier to define by what they are not rather than what they are? Or in general?
- Hofstadter distinguishes between blindly following rules (mechanical mode) and understanding meaning (intelligent mode). Can systems that only follow rules still appear intelligent? Where does meaning actually come from in AI systems?
- Chapter 3 suggests that some truths can’t be captured by any single rule-based system. What might this imply about the limits of AI, math, or automated reasoning?
- Consistency vs. completeness: Would you rather have a system that never produces false results but misses some truths (consistent but incomplete), or one that captures all truths but sometimes produces contradictions?
- Hofstadter argues that symbols don’t have meaning by themselves — meaning comes from interpretation. Does this apply to AI models and programming languages? Do computers “understand,” or just manipulate symbols?
- Systems that analyze themselves (AI evaluating AI, code that modifies itself, etc.) can become powerful but also risky. What are the benefits and dangers of self-referential systems?
Ch 5/6
- The Ch 5 dialogue is quite dense. What did you gather from it in comparison to the content of the chapter?
- Every dialogue title connects to a Bach piece. This one is the Harmonic Labyrinth, why did he choose that one? https://www.youtube.com/watch?v=0JVoKDxVgVI
- What is Djinn/Jinn in Arab Culture/Islam? What connections do you see?
- This book was not written for computer scientists. As such, a lot of the subjects (recursion, loops, etc) are described in a way that is more accessible to the average person (think your mom or dad, unless your mom or dad are in fact programmers). In your opinion, did the book do a good job of that? Did it make you think about these subjects deeper than having studied them in an intro programming course?
- Hofstadter talks about a recursive process for defining language. Do some research and see how parts of speech tagging is done now. Compare the statistical approaches to this recursive approach. Or maybe recursive approaches are still widely used; I'm not sure. You can also connect this to the CSPs we studied in class.
- Side Project Idea! With the advent of "Vibe Coding", it is quite easy to create algorithmic visualizations. I have done many for this class. Vibe-code an app (or do it by hand) to visualize Fibonacci, G, H, Q, INT, and maybe even allow users to enter their own functions. Deploy it to CloudFlare pages: https://pages.cloudflare.com/ - What did you learn from this?
Bonus - Share it on reddit - https://reddit.com/r/GEB
- If you have been struggling with the art and music analogies, maybe the physics one really clicked. Does this apply to you? Why?
- Where do you see recursion in nature?
- Next class we will look at Game Trees in great detail. What are some of your thoughts on how to solve them more efficiently. If you already know the common algorithms please don't answer this one.
- Hofstadter's Law is pretty famous on the internet, and it's seemingly recursive in its definition. How do you make sense of this law which is true, but does not make any sense if you break it down.
- Does the Ch6 dialogue give you any interesting ideas on how one might compose music algorithmically?
- Since 1979, what have we learned about the mapping between genotypes and phenotypes? Do we still not understand the "language"? Hint: Look up something called polygenic risk scores. The ideas you got from high school biology regarding punnet squares are largely wrong, seems like Mendel mostly got lucky with the traits he studied.
- What is your answer to any of Hofstadter's questions about sending certain things to space and aliens deciphering meaning from them?
- What do you think about this "Jukebox theory"? How does this relate to artificial intelligence? Especially when he refers to different types of intelligences.
- "However, it appears likely that within the next few decades there will be much progress made in
elucidating what human intelligence is. In particular, perhaps cognitive psychologists, workers in Artificial Intelligence, and neuroscientists will be able to synthesize their understandings, and come up with a definition of intelligence. It may still be human-chauvinistic; there is no way around that. But to counterbalance that, there may be some elegant and beautiful-and perhaps even simple-abstract ways of characterizing the essence of intelligence." - Have we done this as a society?
- Transmitting a DNA sequence was actually a plot point in the show Pluribus. Did any of you see that? How can you relate that to the book?
- From Ch6, what thoughts do you have about meaning and how it is transmitted?
Ch 7/8
- In the opening dialogue, the Tortoise resists accepting logical conclusions even when they seem obvious. What does this reveal about the difference between formal rules and human acceptance of logic?
- Why is Achilles unable to “force” the Tortoise to accept the logical meaning of “and”? What does this suggest about the foundations of reasoning?
- Hofstadter describes derivations as “typographical” and mechanical. Compare this to how modern machine learning models (like LLMs) generate outputs. Are they closer to mechanical derivations or something else?
- Discuss the debate between “Prudence” and “Imprudence.” Which side do you agree with?
- Listen to the crab cannon; what are your thoughts on it? If you can read sheet music; what did Bach do to make this piece of music work?
- Why is the ability for a system to describe itself such a powerful and potentially dangerous feature?
- What would it mean for an AI to “discover” a theorem vs. just derive it mechanically?
- Can you think of examples in computer science or AI where structure matters more than meaning?
- Create a visual or coded demonstration of recursion or self-reference (could be simple Python or pseudocode).
- Find a modern example of self-reference in technology, art, or internet culture (memes count) and explain how it connects to GEB.
- How do you feel about a story of Zen buddhism being proven by propositional calculus? Does this show any meaning in the system, or is the meaning made by us?
- In most discrete math textbooks, the symbol -> is used for "implies". Hofstadter uses ⊃, and actually looking at old proofs this symbol was more widely used. What do you think the reason is?
- What are some pieces of media which utilize the "push" rule and go deeper into certain levels, and have lower levels affecting higher levels? There's a famous american example, but an Iranian example is the movie "The Mirror".
- Translate some of Tortoise's statements into propositional calculus.