Class Schedule¹

I recommend reading the papers in the order in which they are listed.

Week 1 (Wednesday, September 3, 2025)

What is learnable?

• Readings (in class): Sutton (2019); Agüera y Arcas (2025);
• Related resources: Halevy et al. (2009); talk by Alyosha Efros

Week 2 (Wednesday, September 10, 2025)

How do LLMs work?

• Readings: Rumelhart (1989); Lee (2025); Either Vaswani et al. (2017) or Phuong & Hutter (2022) .
• Other resources for understanding transformers (very useful; strongly recommended!): The illustrated transformer; Transformers, the tech behind LLMs video and the next chapter focusing on attention;
• Extra: McCoy et al. (2024)

Week 3 (Wednesday, September 17, 2025)

Learning language from language

Special guest - Steven Piantadosi

• Readings: Elman (1990); Boleda (2020); Piantadosi (2024)

Week 4 (Wednesday, September 24, 2025)

Learning about the world from language

Special guest - Marina Bedny

• Readings: Plato: The allegory of the cave; Lupyan & Lewis (2017); Yildirim & Paul (2024); Wang et al. (in press)

• Recommended: response to Yildirim and& Paul: Goddu et al. (2024); Counter-response: Yildirim & Paul (2024)

Week 5 (Wednesday, October 1, 2025)

Do large language models understand us?

• Readings: Mitchell & Krakauer (2023); Agüera y Arcas (2022); Piantadosi & Hill (2022)
• Short reflection essay: So what do you think? Do large language models understand us? (~800 words).
• Extra: Feel free to also browse the talks in this series organized by Stevan Harnad, e.g., the different kinds of understanding by Chalmers.

Week 6 (Wednesday, October 8, 2025)

Stress testing embodiment

• Lots of reading this week!
• Readings: Barsalou (1999); Mollo & Millière (2023); Pavlick (2023)
• Extra: A philosophical take by David Chalmers on the possibility of “pure thinkers”

Week 7 (Wednesday, October 15, 2025)

Stress testing the language of thought hypothesis: learning to learn and represent

• Readings: Quilty-Dunn et al. (2022) (also read at least 2 positive and 2 negative commentaries); Lake & Baroni (2023); Watch this talk on learning compositionality by Pavlick
• Extra: Griffiths et al. (2025); Binz et al. (2024);

Week 8 (Wednesday, October 22, 2025)

Stress testing numerical cognition

• Readings: Leslie et al. (2008); Banerjee et al. (2025); O’Shaughnessy et al. (2023); Nanda et al. (2023) (don’t worry about all the technical details. Focus on the big picture. The authors do have a useful video walkthrough).
• Extra: Gordon (2004); O’Shaughnessy et al. (2021);

Week 9 (Wednesday, October 29, 2025)

Stress testing concepts: representational format; flexibility; data efficiency

• Readings: Tentative: Casasanto & Lupyan (2014); Barsalou (2016);
• Related resources: This ~2010 talk by Efros on the role of categories in recognition,

Week 10 (Wednesday, November 5, 2025)

Stress testing reasoning: the role of context

• Readings: (Tentative) Evans (2008); Lampinen et al. (2024)

Week 11 (Wednesday, November 12, 2025)

Stress testing methodology of cognitive psychology/cognitive (neuro)science

• Readings: Bower & Clapper (1989) (read or skim depending on your level of familiarity); Jonas & Kording (2017); Read 2 of the 12 case studies from Lindsey et al. (2025)
• Extras: Nickels et al. (2022); Yarkoni (2020)

Week 12 (Wednesday, November 19, 2025)

Stress testing theories of intelligence

Special guest - Blaise Aguera y Arcas

• Readings: (Tentative) Legg & Hutter (2007); Chollet (2019); Agüera y Arcas (2025) (Chapter 12).
• Extras: van der Maas et al. (2006); Legg & Hutter (2007); Szathmáry & Smith (1995)

Week 13 (Wednesday, November 26, 2025)

Thanksgiving - no class

• Project proposals due by 8pm.

Week 14 (Wednesday, December 3, 2025)

Revisiting old and new questions

• Readings: Review previous readings in light of what you’ve learned
• Short reflection essay: Is all the AI madness of the last few years good for or bad for understanding the human mind? (~800 words).

Week 15 (Wednesday, December 10, 2025)

Final presentations

Group project presentations [20 min presentation + 8 min Q&A].