Overlap in meaning is a stronger predictor of semantic activation in GPT-3 than in humans.


Journal

Scientific reports
ISSN: 2045-2322
Titre abrégé: Sci Rep
Pays: England
ID NLM: 101563288

Informations de publication

Date de publication:
28 03 2023
Historique:
received: 23 12 2022
accepted: 24 03 2023
medline: 30 3 2023
entrez: 28 3 2023
pubmed: 29 3 2023
Statut: epublish

Résumé

Modern large language models generate texts that are virtually indistinguishable from those written by humans and achieve near-human performance in comprehension and reasoning tests. Yet, their complexity makes it difficult to explain and predict their functioning. We examined a state-of-the-art language model (GPT-3) using lexical decision tasks widely used to study the structure of semantic memory in humans. The results of four analyses showed that GPT-3's patterns of semantic activation are broadly similar to those observed in humans, showing significantly higher semantic activation in related (e.g., "lime-lemon") word pairs than in other-related (e.g., "sour-lemon") or unrelated (e.g., "tourist-lemon") word pairs. However, there are also significant differences between GPT-3 and humans. GPT-3's semantic activation is better predicted by similarity in words' meaning (i.e., semantic similarity) rather than their co-occurrence in the language (i.e., associative similarity). This suggests that GPT-3's semantic network is organized around word meaning rather than their co-occurrence in text.

Identifiants

pubmed: 36977744
doi: 10.1038/s41598-023-32248-6
pii: 10.1038/s41598-023-32248-6
pmc: PMC10050205
doi:

Types de publication

Comparative Study Journal Article

Langues

eng

Sous-ensembles de citation

IM

Pagination

5035

Informations de copyright

© 2023. The Author(s).

Références

Q J Exp Psychol (Hove). 2008 Jul;61(7):1036-66
pubmed: 17853227
Psychon Bull Rev. 2021 Feb;28(1):40-80
pubmed: 32885404
Nature. 2022 May;605(7908):21
pubmed: 35484343
J Exp Psychol. 1971 Oct;90(2):227-34
pubmed: 5134329
Psychol Rev. 2021 Jul 22;:
pubmed: 34292021
Psychon Bull Rev. 2000 Dec;7(4):618-30
pubmed: 11206202
Psychol Res. 2002 Aug;66(3):180-94
pubmed: 12192447
Mem Cognit. 1980 Nov;8(6):493-512
pubmed: 7219170
Front Neurosci. 2021 Oct 06;15:750639
pubmed: 34690686
Evolution. 2011 Jul;65(7):1827-40
pubmed: 21729041
Psychon Bull Rev. 2003 Dec;10(4):785-813
pubmed: 15000531
J Exp Psychol Learn Mem Cogn. 1989 Nov;15(6):1003-19
pubmed: 2530303
Proc Natl Acad Sci U S A. 2023 Feb 7;120(6):e2218523120
pubmed: 36730192
Behav Res Methods. 2013 Dec;45(4):1099-114
pubmed: 23344737
Front Psychol. 2018 Mar 15;9:345
pubmed: 29599739

Auteurs

Jan Digutsch (J)

Leibniz Research Centre for Working Environment and Human Factors at the Technical University of Dortmund, Dortmund, Germany. jan.digutsch@unisg.ch.
Institute of Behavioral Science and Technology, University of St. Gallen, St. Gallen, Switzerland. jan.digutsch@unisg.ch.

Michal Kosinski (M)

Stanford University, Stanford, CA, 94305, USA.

Articles similaires

[Redispensing of expensive oral anticancer medicines: a practical application].

Lisanne N van Merendonk, Kübra Akgöl, Bastiaan Nuijen
1.00
Humans Antineoplastic Agents Administration, Oral Drug Costs Counterfeit Drugs

Smoking Cessation and Incident Cardiovascular Disease.

Jun Hwan Cho, Seung Yong Shin, Hoseob Kim et al.
1.00
Humans Male Smoking Cessation Cardiovascular Diseases Female
Humans United States Aged Cross-Sectional Studies Medicare Part C
1.00
Humans Yoga Low Back Pain Female Male

Classifications MeSH