The Economics of Transformative AI

In this course, we explore the interplay between artificial intelligence and economic growth through the lens of foundational and emerging models.

Some claim that advanced AI will greatly:

1. Accelerate economic growth,
2. Displace labor, and/or
3. Risk existential catastrophe.

This two-week summer course covers a selection of tools in (mainly) economic theory relevant to evaluating the first two claims, and deciding what to do if some or all of them seem likely.

The course is designed primarily for graduate students in economics. Others may also benefit from the material, including advanced undergraduates and computer science students.

Opening Weekend (August 16-17): (Optional but recommended) We revisit key concepts in economics, including utility and production functions, substitution, and the basics of growth, setting the stage for more advanced topics.

Week 1 (August 18-22): We dive into AI in standard growth models, beginning with automation’s role in production (both reduced-form and task-based models) and moving to its impact on research alone, and research combined with production. The week concludes by examining scaling and its implications.

Week 2 (August 25-29): We shift to alternative growth models, exploring Schumpeterian ideas, finance and knowledge distribution, and the trade-offs between efficiency, equality, and long-term welfare. The week ends by addressing philosophical questions about long-term survival versus flourishing and strategies for risk mitigation.

Most lectures were given by Phil Trammell, a postdoc in economics at the Stanford Digital Economy Lab. Some lectures were given by Zach Mazlish, a doctoral student in economics at Oxford, while others were conducted by guest lecturers including Chad Jones, professor of economics at Stanford GSB, and Erik Brynjolfsson, Stanford professor and director of the Lab.

Each day consisted of a morning lecture and an afternoon lecture, except the opening Saturday, which only had an afternoon lecture. Each lecture featured two 45-minute halves separated by a 15-minute break.

The course was attended in August 2025 by 48 students from around the world: 32 pursuing or about to start PhDs in economics, 1 pursuing a PhD in computer science, 13 doing research or research assistance at a university or think-tank, and 2 undergraduates. We are proud to say that it was reviewed highly overall, with a significant majority who would strongly recommend it.

Among other comments, participants suggested that the program would have been improved with (1) more scheduled opportunities for networking and sharing research ideas; (2) more discussion concretely linking the microeconomic effects of AI observed today with the macroeconomic models of what may happen if AI greatly advances; and (3) a session reviewing the relevant material and terminology regarding AI itself, analogous to the opening weekend’s review of the relevant economics. We plan to incorporate this feedback going forward.

Course Materials (August 2025)

In some places, the lectures were re-recorded due to audio issues. Otherwise, they are taken from when the lectures were given live in August 2025. Note that the slides in the recordings sometimes contain errors that were corrected in the slides linked below.

Exercises accompanying some of the lectures may be found here.

The sources listed under each lecture are not exhaustive. In some cases, the majority of the lecture consists of an overview summarizing (or touching briefly on) too many sources to list individually, or of material not from any source.

We have linked to the Overleafs that produced the slides, as well as the slides themselves, in case you would like to use them. You are free to use them for any purpose, with or without modification.

Review of relevant economics

001 Optimization and substitution  Slides, Overleaf, Recording

002 Stylized facts of production and growth  Slides, Overleaf, Recording

003 Technological development  Slides, Overleaf, Recording

Part 1: Growth

01 Task-based models: theory  Slides, Overleaf, Recording

• Zeira (1998)
• Aghion et al. (2019)
• Acemoglu (2025)

02 The productivity J-curve (Erik Brynjolfsson)  Slides, Recording

• Brynjolfsson et al. (2021)
• Brynjolfsson, Chandar, Halperin, and Trammell (in progress)

03 Task-based models: evidence  Slides (a, b [Arjun Ramani]), Overleaf, Recording

• Acemoglu (2025), Aghion and Bunel (2024)
• Humlum and Vestergaard (2024)
• Levine (2025a,b)
• Brynjolfsson, Halperin, and Ramani (in progress)

04 Task-based models: selected research  Slides (a [Tomas Aguirre], b [Bharat Chandar]),

Recording

• Aguirre and Manning (in progress)
• Brynjolfsson, Chandar, and Chen (2025)

05 Automating production, homogeneous output  Slides, Overleaf, Recording

• Hanson (2001)
• Acemoglu (2003)
• Sachs and Kotlikoff (2012) 
• Nordhaus (2021)
• Mookherjee and Ray (2022)

06 Automating production, heterogeneous output  Slides, Overleaf, Recording

• Bessen (2018)
• Nordhaus (2021)
• Trammell (in progress)

07 Automating research: basics  Slides, Overleaf, Recording

• Sotala (2012)
• Aghion et al. (2019)
• Agrawal et al. (2019)
• Besiroglu et al. (2024)
• Eth and Davidson (2025), Davidson and Houlden (2025)

08 Automating research: bottlenecks  Slides, Overleaf, Recording

• Erdil and Barnett (2025)
• Whitfill and Wu (2025)
• Aghion et al. (2019), Appendix

09 Full automation and the Malthusian past  Slides (a, b), Overleaf, Recording

• Kremer (1993)
• Aghion et al. (2019)
• Roodman (2020), Davidson (2021)
• Davidson et al. (in progress)

10 Full automation: BOTECs and bottlenecks  Slides, Overleaf, Recording

• Davidson and Hadshar (2025)
• Trammell (2025a,b)

Part 2: Scaling, finance, risk

11 Scaling laws: basics  Slides, Overleaf, Recording

• Wijk et al. (2024)
• Merali (2024)
• Villalobos (2023)
• Ho et al. (2024)

12-13 Scaling laws: growth models  Slides (Anson Ho), Recording (12), Recording (13)

• Kokotajlo et al. (2025)
• Davidson (2023), Kannegeiter (2023)
• Erdil et al. (2025)

14 TAI and finance  Slides, Overleaf, Recording

• Chow et al. (2024)
• Andrews and Farboodi (2025)
• Eisfeldt et al. (2023)

15 AI safety  Slides (a, b [Max Reith], c [Eric Chen and Sami Petersen], d), Overleaf (a, d),

Recording

• Thornley (2024, 2025), Thornley et al. (2025)
• Chen et al. (2024)
• Trammell (2024)

16 TAI and social welfare  Slides (a, b), Overleaf (a, b), Recording

A Inequality 

• Nordhaus (2004)
• O’Keefe et al. (2020)
• Garfinkel (2021), Carvalho (2025)
• Korinek and Stiglitz (2021)

B Longtermism

• Bostrom (2003)
• Nordhaus (2009)
• Beckstead (2013), Greaves and MacAskill (2021)
• Thorstad (2022)
• Trammell and Aschenbrenner (2025), App. A

17 Existential risk vs. growth  Slides (a [Chad Jones], b [Tom Houlden]), Recording

• Jones (2024), Houlden (2024)
• Jones (2025), Shulman and Thornley (2024)

18 Existential risk and growth  Slides, Overleaf, Recording

• Trammell and Aschenbrenner (2025)

19 AI governance  Slides (a, b), Overleaf (a, b), Recording

• Armstrong et al. (2015), Trager, Dafoe, Jensen, and Emery-Xu (various)
• Acemoglu and Lensman (2024), Gans (2024), Koh and Sanguanmoo (2024)

20 Choosing our future  Slides, Overleaf, Recording

• MacAskill (2025)
• Trammell (in progress)
• Assadi (2023), Ely and Szentes (2024)
• Finnveden et al. (2022)