On the Opacity of Deep Neural Networks

Canadian Journal of Philosophy:1-16 (forthcoming)
  Copy   BIBTEX

Abstract

Deep neural networks are said to be opaque, impeding the development of safe and trustworthy artificial intelligence, but where this opacity stems from is less clear. What are the sufficient properties for neural network opacity? Here, I discuss five common properties of deep neural networks and two different kinds of opacity. Which of these properties are sufficient for what type of opacity? I show how each kind of opacity stems from only one of these five properties, and then discuss to what extent the two kinds of opacity can be mitigated by explainability methods.

Categories

Add categories

Keywords

Reprint years

DOI

10.1017/can.2024.1

Links

PhilArchive



    Upload a copy of this work     Papers currently archived: 91,928

External links

Setup an account with your affiliations in order to access resources via your University's proxy server

Through your library

My notes

Similar books and articles

Toward a sociology of machine learning explainability: Human–machine interaction in deep neural network-based automated trading.Bo Hee Min & Christian Borch - 2022 - Big Data and Society 9 (2).
A Puzzle concerning Compositionality in Machines.Ryan M. Nefdt - 2020 - Minds and Machines 30 (1):47-75.
Big Data and Deep Learning Models.Daniel Sander Hoffmann - 2022 - Principia: An International Journal of Epistemology 26 (3):597-614.
Computer Simulations, Machine Learning and the Laplacean Demon: Opacity in the Case of High Energy Physics.Florian J. Boge & Paul Grünke - forthcoming - In Andreas Kaminski, Michael Resch & Petra Gehring (eds.), The Science and Art of Simulation II.
Three Strategies for Salvaging Epistemic Value in Deep Neural Network Modeling.Philippe Verreault-Julien - manuscript
Understanding Deep Learning with Statistical Relevance.Tim Räz - 2022 - Philosophy of Science 89 (1):20-41.
Should we replace radiologists with deep learning? Pigeons, error and trust in medical AI.Ramón Alvarado - 2021 - Bioethics 36 (2):121-133.
Symbolic Deep Networks: A Psychologically Inspired Lightweight and Efficient Approach to Deep Learning.Vladislav D. Veksler, Blaine E. Hoffman & Norbou Buchler - 2022 - Topics in Cognitive Science 14 (4):702-717.
Computational Functionalism for the Deep Learning Era.Ezequiel López-Rubio - 2018 - Minds and Machines 28 (4):667-688.
Moving beyond content‐specific computation in artificial neural networks.Nicholas Shea - 2021 - Mind and Language 38 (1):156-177.
Deep Learning Opacity in Scientific Discovery.Eamon Duede - 2023 - Philosophy of Science 90 (5):1089 - 1099.
Some Neural Networks Compute, Others Don't.Gualtiero Piccinini - 2008 - Neural Networks 21 (2-3):311-321.
Estimation and application of matrix eigenvalues based on deep neural network.Zhiying Hu - 2022 - Journal of Intelligent Systems 31 (1):1246-1261.
Chess, Artificial Intelligence, and Epistemic Opacity.Paul Grünke - 2019 - Információs Társadalom 19 (4):7--17.
Deep learning: A philosophical introduction.Cameron Buckner - 2019 - Philosophy Compass 14 (10):e12625.

Analytics

Added to PP
2024-03-26

Downloads
27 (#589,794)

6 months
27 (#110,477)

Historical graph of downloads
How can I increase my downloads?

Author's Profile

Anders Søgaard
University of Copenhagen

Citations of this work

No citations found.

Add more citations

References found in this work

Intentionality: An Essay in the Philosophy of Mind.John R. Searle - 1983 - New York: Cambridge University Press.
Understanding from Machine Learning Models.Emily Sullivan - 2022 - British Journal for the Philosophy of Science 73 (1):109-133.
Transparency in Complex Computational Systems.Kathleen A. Creel - 2020 - Philosophy of Science 87 (4):568-589.
The unreliability of naive introspection.Eric Schwitzgebel - 2006 - Philosophical Review 117 (2):245-273.

View all 13 references / Add more references