Does the No Miracles Argument Apply to AI?

Darrell P. ROWBOTTOM; William PEDEN; André CURTIS-TRUDEL

doi:10.1007/s11229-024-04524-z

Does the No Miracles Argument Apply to AI?

Darrell P. ROWBOTTOM^*, William PEDEN, André CURTIS-TRUDEL

^*Corresponding author for this work

Department of Philosophy

Research output: Journal Publications › Journal Article (refereed) › peer-review

Abstract

According to the standard no miracles argument, science’s predictive success is best explained by the approximate truth of its theories. In contemporary science, however, machine learning systems, such as AlphaFold2, are also remarkably predictively successful. Thus, we might ask what best explains such successes. Might these AIs accurately represent critical aspects of their targets in the world? And if so, does a variant of the no miracles argument apply to these AIs? We argue for an affirmative answer to these questions. We conclude that if the standard no miracles argument is sound, an AI-specific no miracles argument is also sound.

Original language	English
Article number	173
Number of pages	20
Journal	Synthese
Volume	203
Issue number	5
Early online date	13 May 2024
DOIs	https://doi.org/10.1007/s11229-024-04524-z
Publication status	Published - 13 May 2024

Bibliographical note

We are grateful for feedback on earlier versions of the paper from audiences at Cambridge University, Kyoto University, Technical University Berlin, and Zhejiang University.

Publisher Copyright: © The Author(s) 2024.

Funding

Open Access Publishing Support Fund provided by Lingnan University. This work was funded by Research Grants Council, University Grants Committee, SRFS2122-3H01.

Keywords

Artificial intelligence
Scientific realism
Machine learning
Scientific progress
Scientific representation
AlphaFold

Access to Document

10.1007/s11229-024-04524-zLicence: CC BY

Cite this

@article{42a68176ec5b45b2a7d633502f047a15,

title = "Does the No Miracles Argument Apply to AI?",

abstract = "According to the standard no miracles argument, science{\textquoteright}s predictive success is best explained by the approximate truth of its theories. In contemporary science, however, machine learning systems, such as AlphaFold2, are also remarkably predictively successful. Thus, we might ask what best explains such successes. Might these AIs accurately represent critical aspects of their targets in the world? And if so, does a variant of the no miracles argument apply to these AIs? We argue for an affirmative answer to these questions. We conclude that if the standard no miracles argument is sound, an AI-specific no miracles argument is also sound.",

keywords = "Artificial intelligence, Scientific realism, Machine learning, Scientific progress, Scientific representation, AlphaFold",

author = "ROWBOTTOM, {Darrell P.} and William PEDEN and Andr{\'e} CURTIS-TRUDEL",

note = "We are grateful for feedback on earlier versions of the paper from audiences at Cambridge University, Kyoto University, Technical University Berlin, and Zhejiang University. Publisher Copyright: {\textcopyright} The Author(s) 2024.",

year = "2024",

month = may,

day = "13",

doi = "10.1007/s11229-024-04524-z",

language = "English",

volume = "203",

journal = "Synthese",

issn = "0039-7857",