Cosma Rohilla Shalizi is a statistician at Carnegie Mellon and a co-author, with Henry Farrell, of AI as Social Technology (Knight Columbia, May 11 2026). His research applies statistics, machine learning, nonlinear dynamics, and statistical physics to complex systems, and he is a co-PI on the broader AI as Cultural Technology research program.
Background
Shalizi has been at Carnegie Mellon since 2005, where he is an associate professor (with tenure since 2014) in the Department of Statistics & Data Science, with affiliations with the Machine Learning Department. He is also external faculty at the Santa Fe Institute. His research applies tools from statistics, machine learning, nonlinear dynamics, and statistical physics to complex systems, focusing on prediction and inference for dependent, often high-dimensional data, and on the statistics of social networks. He is the author of the open statistics textbook Advanced Data Analysis from an Elementary Point of View, and has kept a long-running personal blog and notebook site, Three-Toed Sloth / bactra.org, known for its discursive essays on statistics, social science, and the limits of machine learning (Source: https://www.stat.cmu.edu/~cshalizi/) (Source: https://bactra.org/research/cv.pdf).
Work on AI as social and cultural technology
Shalizi co-authored AI as Social Technology (Knight Columbia, May 11 2026) with Farrell; see AI as Social Technology and AI as Social Technology (Farrell + Shalizi, Knight Columbia, May 11 2026). He is a co-PI on the broader AI as Cultural Technology research program with Farrell, Alison Gopnik, and James Evans.
His distinctive contribution to that framework is its statistical-modeling and coarse-graining core. The argument treats large language models not as nascent minds but as statistical models, with the characteristic trade-off of any statistical model between good average-case behavior and poor tail-case behavior, and treats deep neural networks as systems that build internal coarse-grainings of the world whose nature is often obscure even to their builders. This connects to a long-standing skeptical strand in his work, including his sympathy for the Rahimi-Recht (2017) "alchemy" characterization of practical deep-learning research as effective engineering running ahead of theoretical understanding.
Shalizi is also the source of one of the framework's load-bearing historical claims: his much-cited 2010 essay arguing that "the Singularity began two centuries ago, with the Industrial Revolution," that a self-reinforcing, faster-than-human optimization process (the market economy and the corporation) is already old news. That framing is the seed of the The Long Industrial Revolution concept and of the Farrell-Shalizi move to treat AI as the latest in a lineage of social and cultural technologies rather than as a sudden break.
Prior work cited
- Shalizi (2010) — "The Singularity began two centuries ago with the Industrial Revolution" framing.
- Rahimi & Recht (2017) — the "alchemy" critique of deep-learning practice; cited in Farrell-Shalizi 2026.
- Statistical-physics and complex-systems methodology applied to social science.
Relationships
- affiliated-with: Carnegie Mellon, Santa Fe Institute
- co-author-with: Henry Farrell, Alison Gopnik (AI as Cultural Technology program)
- author-of: AI as Social Technology (Farrell + Shalizi, Knight Columbia, May 11 2026)
- supports: AI as Social Technology, AI as Cultural Technology, The Long Industrial Revolution
- related: AI as Social Technology, Herbert A. Simon
Sources
- AI as Social Technology (Farrell + Shalizi, Knight Columbia, May 11 2026) — Farrell & Shalizi, AI as Social Technology, Knight Columbia, May 2026.
- (Source: https://www.stat.cmu.edu/~cshalizi/) — Shalizi's Carnegie Mellon faculty page (affiliation, research focus).
- (Source: https://bactra.org/research/cv.pdf) — Shalizi's curriculum vitae (positions, dates).