Artificial Intelligence: A Guide for Thinking Humans

Mitchell frequently points out the hidden limitations of spectacular AI successes (like Deep Blue, Watson, and image recognition benchmarks). Analyze this contrast as a rhetorical strategy. How does she use narrative structure (moving from breakthroughs to “cracks”) shape readers’ perception of progress and hype in AI?

Discuss how Mitchell integrates insights from psychology, cognitive science, and linguistics into her account of AI. How do concepts like intuitive physics, metaphor, mental models, and analogy deepen her central claim that intelligence is more than data and computation? In what ways does this interdisciplinary approach challenge a purely technical framing of AI?

Consider Mitchell’s treatment of “commonsense knowledge” as both a scientific and philosophical problem. How does she use case studies (Cyc, SQuAD, Winograd schemas, science-question benchmarks, etc.) to demonstrate the difficulty of encoding or learning common sense? What view of human cognition emerges from these examples?

Analyze the ethical and social concerns Mitchell raises about contemporary AI systems (bias, opacity, adversarial attacks, surveillance, fake media, and automation). How does she balance caution with optimism, and what assumptions about responsibility (corporate, governmental, individual) underpin her discussion of AI governance?

Compare Mitchell’s account of AI with earlier narratives about human uniqueness and machines (e.g., Turing’s test, I. J. Good’s “intelligence explosion,” Kurzweil’s Singularity). How does she position her own perspective within or against these techno-utopian or techno-dystopian traditions, and what does that reveal about AI as both a cultural narrative and a scientific field?

Mitchell highlights analogy and abstraction as “core” to human intelligence. Choose one extended example (Bongard problems, Copycat, Metacat, or Situate) and analyze how she uses it to dramatize the distance between human and machine cognition. How effectively does this example support her broader thesis about what current deep learning is missing?

Explore the role of embodiment in Mitchell’s argument. How do her discussions of self-driving cars, robotics, and vision (including Karpathy’s locker-room photo) support the claim that intelligence may require an embodied, human presence? To what degree does she suggest that purely disembodied, data-driven AI is fundamentally limited?

Mitchell often foregrounds her own experiences as a researcher, reader of Gödel, Escher, Bach, and observer of public AI debates. How does this first-person voice influence the tone, credibility, and accessibility of the book? To what extent is Artificial Intelligence both a popular science text and a personal intellectual memoir?