Dr. Peter S. Park (Harvard Ph.D. '23, Princeton A.B. '17) conducts research at MIT as a Vitalik Buterin Postdoctoral Fellow in AI Existential Safety. He is grateful to be advised as a Postdoctoral Associate by Max Tegmark of MIT.
Dr. Park applies mathematics, the social sciences, and the cognitive sciences to research human-AI dynamics. To illustrate, OpenAI's mission is to create "highly autonomous systems that outperform humans at most economically valuable work." If and when we humans cede our decision-making to these highly autonomous AIs, will we come to regret this potentially irreversible choice? Dr. Park investigates this question by using a variety of tools, such as empirical studies of current AI systems, relevant data from the evolutionary or historical past, and mathematical models of what a hypothetical AI-led future may look like.
Dr. Park was advised by Joe Henrich of Harvard's Human Evolutionary Biology department (Culture, Cognition and Coevolution Lab). He was advised by Martin Nowak from AY2019-2020 spring to AY2020-2021 fall, and by Eric Maskin from AY2018-2019 fall to AY2019-2020 fall. Before that, he was an undergraduate studying math at Princeton University, where he was advised by Peter Sarnak (senior thesis) and Manjul Bhargava (junior independent work).
Please find Dr. Park's CV here.
Selected publications:
-
Divide-and-conquer dynamics in AI-driven disempowerment (with M. Tegmark). arXiv CS preprint (2023).
- Coverage: Thread by Peter S. Park; Interview on Jacob Haimes' Into AI Safety Podcast — discussed in Round 1 (Platforms: Spotify, Apple Podcasts, Amazon Music) and Round 2 (Platforms: Spotify, Apple Podcasts, Amazon Music)
-
AI deception: A survey of existing examples, risks, and potential solutions (with S. Goldstein, A.O’Gara, M. Chen, and D. Hendrycks). arXiv CS preprint (2023). Pre-Accepted at Patterns.
- Coverage: Op-ed (coauthored with Simon Goldstein) in The Conversation; Articles in BGR, Tweaktown, TS2's website, and I-HLS; Thread by Dan Hendrycks; Research Summary (coauthored with Aidan O'Gara) in the AI Ethics Brief
-
Diminished diversity-of-thought in a standard large language model (with P. Schoenegger and C. Zhu). Behavior Research Methods (2024).
- Coverage: Article in Marginal Revolution, Thread by Ethan Mollick, Thread by Philipp Schoenegger
- Cooperation in alternating interactions with memory constraints (with M. A. Nowak and C. Hilbe). Nature Communications. 13 (2022), Article no. 737.
- The evolution of cognitive biases in human learning. Journal of Theoretical Biology. 541 (2022), 111031.
List of publications:
- On pairwise intersections of the Fibonacci, Sierpinski, and Riesel sequences (with D. Ismailescu). J. Integer Seq. 16 (2013), 13.9.8.
- Linnik's theorem for Sato–Tate laws on elliptic curves with complex multiplication (with E. Chen and A. A. Swaminathan). Res. Number Theory 1(28) (2015), 1-11
- On logarithmically Benford sequences (with E. Chen and A. A. Swaminathan). Proc. A.M.S. 144 (2016), 4599-4608.
- The "Riemann hypothesis" is true for period polynomials of almost all newforms (with Y. Liu and Z. Q. Song). Res. Math. Sci, 3(31) (2016), 1-11.
- The van der Waerden complex (with R. Ehrenborg, L. Govindaiah, and M. Readdy). J. Number Theory 172 (2017), 287-300
- Bounded gaps between products of distinct primes (with Y. Liu and Z. Q. Song). Res. Number Theory 3(26) (2017), 1-28.
- Elliptic curve variants of the least quadratic nonresidue problem and Linnik's theorem (with E. Chen and A. A. Swaminathan). Int. J. Number Theory 14(1) (2018), 255-288.
- Conjugacy growth of commutators. J. Algebra. 526 (2019), 423-458.
- Cooperation in alternating interactions with memory constraints (with M. A. Nowak and C. Hilbe). Nature Communications. 13 (2022), Article no. 737.
- The evolution of cognitive biases in human learning. Journal of Theoretical Biology. 541 (2022), 111031.
- AI deception: A survey of existing examples, risks, and potential solutions (with S. Goldstein, A.O’Gara, M. Chen, and D. Hendrycks). arXiv CS preprint (2023). Pre-Accepted at Patterns.
- Which Humans? (with M. Atari, M. Xue, D. Blasi, and J. Henrich). PsyArXiv preprint (2023).
- Divide-and-conquer dynamics in AI-driven disempowerment (with M. Tegmark). arXiv CS preprint (2023).
-
Large language model prediction capabilities: Evidence from a real-world forecasting tournament (with P. Schoenegger). arXiv CS preprint (2023).
- Coverage: LSE Blog; Thread by Ethan Mollick; Thread by Philipp Schoenegger
- Perils and opportunities in using large language models in psychological research (with S. Abdurahman et al.). OSF preprint (2023).
- Diminished diversity-of-thought in a standard large language model (with P. Schoenegger and C. Zhu). Behavior Research Methods (2024).
-
AI-augmented predictions: LLM assistants improve human forecasting accuracy (with P. Schoenegger, E. Klein, and P. E. Tetlock). arXiv CS preprint (2024).
- Coverage: Thread by Philip E. Tetlock; Thread by Ethan Mollick; Thread by Philipp Schoenegger
-
Wisdom of the silicon crowd: LLM ensemble prediction capabilities rival human crowd accuracy (with P. Schoenegger, I. Tuminauskaite, and P. E. Tetlock). arXiv CS preprint (2024).
- Coverage: Astral Star Codex; Marginal Revolution; Thread by Ethan Mollick; Thread by John Nay; Thread by the AI Database; Thread by Philipp Schoenegger; Correction by Philipp Schoenegger
- Devising and detecting phishing: Large language models vs. smaller human models (with F. Heiding, B. Schneier, A. Vishwanath, and J. Bernstein). To appear in IEEE Access (2024).
- A theory of specialization, exchange, and innovation in human groups (with Viacheslav Savitskiy and Joseph Henrich). Presented at the Cultural Evolution Society Conference (2022). Journal version in progress.
- The iterated prisoner’s dilemma with implementation errors (with M. A. Nowak and C. Hilbe). In progress.
- The alternating prisoner's dilemma with cognitive costs (with C. Hilbe). In progress.
- A theory of social competition (with E. Schnell and M. Muthukrishna). In progress.
- Environments as networks, cultural systems as organic structures and the rich-lean hypothesis (with I. Kroupin and T. C. Zeng). In progress.
- Prestige in animal societies? (with A. Sarkar, J. Henrich, R. Reddy, and L. Samuni). In progress.
Expository writings:
- Siegel's theorem.
- The uniformization theorem for elliptic curves.
- The Bombieri–Vinogradov theorem.
- de Rham's theorem.
- Hodge theory.
- Probability laws for the distribution of geometric lengths when sampling by a random walk in a Fuchsian fundamental group
- Will powerful AI systems of the future pose societal risks?