For human-like models, train on human-like tasks.

Hermann, Katherine; Nayebi, Aran; van Steenkiste, Sjoerd; Jones, Matt

Hermann, Katherine; Nayebi, Aran; van Steenkiste, Sjoerd; Jones, Matt.

Afiliação

Hermann K; Google DeepMind, Mountain View, CA, USA hermannk@google.com.
Nayebi A; McGovern Institute, Massachusetts Institute of Technology, Cambridge, MA, USA aran.nayebi@gmail.com https://anayebi.github.io/.
van Steenkiste S; Google Research, Mountain View, CA, USA sjoerdvansteenkiste@gmail.com https://www.sjoerdvansteenkiste.com/.
Jones M; Google Research, Mountain View, CA, USA sjoerdvansteenkiste@gmail.com https://www.sjoerdvansteenkiste.com/.

Behav Brain Sci ; 46: e394, 2023 Dec 06.

Article em En | MEDLINE | ID: mdl-38054325

ABSTRACT

ABSTRACT

Bowers et al. express skepticism about deep neural networks (DNNs) as models of human vision due to DNNs' failures to account for results from psychological research. We argue that to fairly assess DNNs, we must first train them on more human-like tasks which we hypothesize will induce more human-like behaviors and representations.

Assuntos

Aprendizado Profundo; Redes Neurais de Computação; Humanos

Texto completo

Imprimir

XML

PubMed Links

Buscar no Google

Texto completo: 1 Coleções: 01-internacional Base de dados: MEDLINE Assunto principal: Redes Neurais de Computação / Aprendizado Profundo Limite: Humans Idioma: En Ano de publicação: 2023 Tipo de documento: Article

Texto completo

Imprimir

XML

PubMed Links

Buscar no Google