Astrophysics in the Era of Artificial Intelligence powered by Large Language Models
Abstract
In recent times there is considerable unease about the way Artificial Intelligence (AI) powered by large language models is used to do astrophysics, and, assuming that these AIs will get far better in future, several alarming situations are also discussed. In a ``white paper,'' Hogg argues that extreme cases like allowing AI to write all astrophysics papers or strictly controlling AI are not good choices and posits that it is difficult to adopt moderate policies. I examine the impact of AI on astrophysics by making two rather extreme assumptions: first, that the modern astrophysicist has become increasingly preoccupied with the administrative task of publication over the act of physical discovery (astrophysicists are busy writing papers and have stopped doing astrophysics); and second, that while AI possesses the capacity to synthesize literature and generate manuscripts, they lack the intrinsic ability to perform original astrophysical inquiry (AI can write papers but cannot do astrophysics). I provide some plausible reasons for adopting these assumptions, but, more importantly, extrapolate the implications of these assumptions to argue that AI can have a very positive impact on astrophysics. By situating AI as agents of publication rather than discovery, the focus of the human researcher can be directed towards fundamental astrophysics, leading to very positive growth in the field.
Related
- The Robots are Coming: Artificial Intelligence Searching for Extraterrestrial Intelligence
- Future of Artificial Intelligence for Science in Japan 2024 Community Report
- Mephisto: Self-improving Large Language Model–based Agents for Automated Interpretation of Multiband Galaxy Observations
- Generative artificial intelligence for reconstructing neutron-star matter
- Detection of HI Holes in NGC 6946 Using Artificial Intelligence: New Candidates from moment-0 Maps
BibTeX
@misc{rao2026astrophysics,
title = {{Astrophysics in the Era of Artificial Intelligence powered by Large Language Models}},
author = {A. R. Rao},
howpublished = {arXiv},
year = {2026},
month = sep,
doi = {10.48550/arxiv.2609.13301},
eprint = {2609.13301},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2609.13301},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - Astrophysics in the Era of Artificial Intelligence powered by Large Language Models
AU - A. R. Rao
PY - 2026
DA - 2026/09/10
PB - arXiv
DO - 10.48550/arxiv.2609.13301
UR - https://arxiv.org/abs/2609.13301
AB - In recent times there is considerable unease about the way Artificial Intelligence (AI) powered by large language models is used to do astrophysics, and, assuming that these AIs will get far better in future, several alarming situations are also discussed. In a ``white paper,'' Hogg argues that extreme cases like allowing AI to write all astrophysics papers or strictly controlling AI are not good choices and posits that it is difficult to adopt moderate policies. I examine the impact of AI on astrophysics by making two rather extreme assumptions: first, that the modern astrophysicist has become increasingly preoccupied with the administrative task of publication over the act of physical discovery (astrophysicists are busy writing papers and have stopped doing astrophysics); and second, that while AI possesses the capacity to synthesize literature and generate manuscripts, they lack the intrinsic ability to perform original astrophysical inquiry (AI can write papers but cannot do astrophysics). I provide some plausible reasons for adopting these assumptions, but, more importantly, extrapolate the implications of these assumptions to argue that AI can have a very positive impact on astrophysics. By situating AI as agents of publication rather than discovery, the focus of the human researcher can be directed towards fundamental astrophysics, leading to very positive growth in the field.
ER -