Using cognitive psychology to understand GPT-3
Proceedings of the National Academy of Sciences · 07.02.2023 · 120 (6)
DOI: 10.1073/pnas.2218523120 ↗Kurz gefasst
Fachartikel in der Zeitschrift Proceedings of the National Academy of Sciences, online erschienen am 02.02.2023. Verfasst von Marcel Binz und 1 weiteren Autor:innen. Beteiligte Einrichtung: Max Planck Research Group (MPRG) Computational Principles of Intelligence, Max Planck Institute for Biological Cybernetics, Tübingen 72076, Germany. Hauptthemen laut MEDLINE: Decision Making, Cognitive Psychology. Die Arbeit wurde bisher 489-mal zitiert (Crossref) und nennt 55 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY-NC-ND 4.0). Gefördert von: Volkswagen Foundation, Deutsche Forschungsgemeinschaft, Max-Planck-Gesellschaft.
Abstract (Originalsprache)
We study GPT-3, a recent large language model, using tools from cognitive psychology. More specifically, we assess GPT-3’s decision-making, information search, deliberation, and causal reasoning abilities on a battery of canonical experiments from the literature. We find that much of GPT-3’s behavior is impressive: It solves vignette-based tasks similarly or better than human subjects, is able to make decent decisions from descriptions, outperforms humans in a multiarmed bandit task, and shows signatures of model-based reinforcement learning. Yet, we also find that small perturbations to vignette-based tasks can lead GPT-3 vastly astray, that it shows no signatures of directed exploration, and that it fails miserably in a causal reasoning task. Taken together, these results enrich our understanding of current large language models and pave the way for future investigations using tools from cognitive psychology to study increasingly capable and opaque artificial agents.
Schlagwörter
MeSH-Schlagwörter (MEDLINE)
Stichwörter der Autor:innen
Autor:innen und Institutionen
- Max Planck Research Group (MPRG) Computational Principles of Intelligence, Max Planck Institute for Biological Cybernetics, Tübingen 72076, Germany.
ORCID-Profile: Marcel Binz
Förderung
- Volkswagen Foundation
- Deutsche Forschungsgemeinschaft
- Max-Planck-Gesellschaft
Zitieren
APA
Binz, M., & Schulz, E. (2023). Using cognitive psychology to understand GPT-3. Proceedings of the National Academy of Sciences, 120(6). https://doi.org/10.1073/pnas.2218523120
Vancouver
Binz M, Schulz E. Using cognitive psychology to understand GPT-3. Proceedings of the National Academy of Sciences. 2023;120(6). doi:10.1073/pnas.2218523120
BibTeX
@article{Binz2023,
author = {Binz, Marcel and Schulz, Eric},
title = {Using cognitive psychology to understand GPT-3},
journal = {Proceedings of the National Academy of Sciences},
year = {2023},
volume = {120},
number = {6},
doi = {10.1073/pnas.2218523120}
}
EUVIMED