Position: The Platonic Representation Hypothesis

Minyoung Huh, Brian Cheung, Tongzhou Wang, Phillip Isola
Proceedings of the 41st International Conference on Machine Learning, PMLR 235:20617-20642, 2024.

Abstract

We argue that representations in AI models, particularly deep networks, are converging. First, we survey many examples of convergence in the literature: over time and across multiple domains, the ways by which different neural networks represent data are becoming more aligned. Next, we demonstrate convergence across data modalities: as vision models and language models get larger, they measure distance between datapoints in a more and more alike way. We hypothesize that this convergence is driving toward a shared statistical model of reality, akin to Plato’s concept of an ideal reality. We term such a representation the platonic representation and discuss several possible selective pressures toward it. Finally, we discuss the implications of these trends, their limitations, and counterexamples to our analysis.

Cite this Paper

BibTeX
@InProceedings{pmlr-v235-huh24a, title = {Position: The Platonic Representation Hypothesis}, author = {Huh, Minyoung and Cheung, Brian and Wang, Tongzhou and Isola, Phillip}, booktitle = {Proceedings of the 41st International Conference on Machine Learning}, pages = {20617--20642}, year = {2024}, editor = {Salakhutdinov, Ruslan and Kolter, Zico and Heller, Katherine and Weller, Adrian and Oliver, Nuria and Scarlett, Jonathan and Berkenkamp, Felix}, volume = {235}, series = {Proceedings of Machine Learning Research}, month = {21--27 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v235/main/assets/huh24a/huh24a.pdf}, url = {https://proceedings.mlr.press/v235/huh24a.html}, abstract = {We argue that representations in AI models, particularly deep networks, are converging. First, we survey many examples of convergence in the literature: over time and across multiple domains, the ways by which different neural networks represent data are becoming more aligned. Next, we demonstrate convergence across data modalities: as vision models and language models get larger, they measure distance between datapoints in a more and more alike way. We hypothesize that this convergence is driving toward a shared statistical model of reality, akin to Plato’s concept of an ideal reality. We term such a representation the platonic representation and discuss several possible selective pressures toward it. Finally, we discuss the implications of these trends, their limitations, and counterexamples to our analysis.} }
Endnote
%0 Conference Paper %T Position: The Platonic Representation Hypothesis %A Minyoung Huh %A Brian Cheung %A Tongzhou Wang %A Phillip Isola %B Proceedings of the 41st International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2024 %E Ruslan Salakhutdinov %E Zico Kolter %E Katherine Heller %E Adrian Weller %E Nuria Oliver %E Jonathan Scarlett %E Felix Berkenkamp %F pmlr-v235-huh24a %I PMLR %P 20617--20642 %U https://proceedings.mlr.press/v235/huh24a.html %V 235 %X We argue that representations in AI models, particularly deep networks, are converging. First, we survey many examples of convergence in the literature: over time and across multiple domains, the ways by which different neural networks represent data are becoming more aligned. Next, we demonstrate convergence across data modalities: as vision models and language models get larger, they measure distance between datapoints in a more and more alike way. We hypothesize that this convergence is driving toward a shared statistical model of reality, akin to Plato’s concept of an ideal reality. We term such a representation the platonic representation and discuss several possible selective pressures toward it. Finally, we discuss the implications of these trends, their limitations, and counterexamples to our analysis.
APA
Huh, M., Cheung, B., Wang, T. & Isola, P.. (2024). Position: The Platonic Representation Hypothesis. Proceedings of the 41st International Conference on Machine Learning, in Proceedings of Machine Learning Research 235:20617-20642 Available from https://proceedings.mlr.press/v235/huh24a.html.

Related Material