Jove
Visualize
Contact Us
JoVE
x logofacebook logolinkedin logoyoutube logo
ABOUT JoVE
OverviewLeadershipBlogJoVE Help Center
AUTHORS
Publishing ProcessEditorial BoardScope & PoliciesPeer ReviewFAQSubmit
LIBRARIANS
TestimonialsSubscriptionsAccessResourcesLibrary Advisory BoardFAQ
RESEARCH
JoVE JournalMethods CollectionsJoVE Encyclopedia of ExperimentsArchive
EDUCATION
JoVE CoreJoVE BusinessJoVE Science EducationJoVE Lab ManualFaculty Resource CenterFaculty Site
Terms & Conditions of Use
Privacy Policy
Policies

Related Experiment Videos

A model for discovering 'containment' relations.

Shimon Ullman1, Nimrod Dorfman1, Daniel Harari1

  • 1Weizmann Institute of Science, Department of Computer Science and Applied Mathematics, 234 Herzl Street, Rehovot 7610001, Israel.

Cognition
|November 13, 2018
PubMed
Summary
This summary is machine-generated.

Related Concept Videos

You might also read

Related Articles

Articles linked to this work by shared authors, journal, and citation graph.

Sort by
Same author

IRF1 is a context-dependent homeostatic gatekeeper of basal immunity and antiviral readiness.

The Journal of biological chemistry·2026
Same author

Neural correlates of minimal recognizable configurations in the human brain.

Cell reports·2025
Same author

Human-like scene interpretation by a guided counterstream processing.

Proceedings of the National Academy of Sciences of the United States of America·2023
Same author

Paradoxical activation of chronic lymphocytic leukemia cells by ruxolitinib <i>in vitro</i> and <i>in vivo</i>.

Frontiers in oncology·2023
Same author

Gelatin Stabilizes Nebulized Proteins in Pulmonary Drug Delivery against COVID-19.

ACS biomaterials science & engineering·2022
Same author

Gaze following requires early visual experience.

Proceedings of the National Academy of Sciences of the United States of America·2022
Same journal

Corrigendum to "Productivity matters for the neural processing of novel words, but not existing ones" Cognition Volume 274 (2026) 106593.

Cognition·2026
Same journal

Investigating the origins of partisanship: What motivates children to preferentially endorse their ingroups' claims?

Cognition·2026
Same journal

People make graded judgments about the inconceivable.

Cognition·2026
Same journal

The self as an image: Appearance and belief in visual representations of one's own face.

Cognition·2026
Same journal

Corrigendum to 'Consonant, vowel, and tone cues in early wordform recognition: Evidence from Cantonese-learning infants' [Cognition 275 (2026) 106624].

Cognition·2026
Same journal

Identifying distinct sources of whole number interference in children's decimal comparison: the role of numerical magnitude and inhibitory control.

Cognition·2026
See all related articles

Infants learn complex concepts like

Area of Science:

  • Cognitive Science
  • Developmental Psychology
  • Artificial Intelligence

Background:

  • Current AI models struggle to explain infant learning, particularly concept acquisition without external supervision.
  • Infants learn complex concepts, such as 'containment', remarkably early (around 3 months) and in a specific developmental order.
  • Existing methods rely on large labeled datasets, which do not reflect natural infant learning processes.

Purpose of the Study:

  • To present a computational model explaining how infants learn 'containment' and related concepts through visual experience alone.
  • To elucidate the developmental trajectory of early spatial relation learning in infants.
  • To demonstrate a mechanism for unsupervised concept learning applicable to both cognitive science and artificial intelligence.

Main Methods:

Keywords:
Computational modelContainment relationDevelopmental trajectoryInfants’ perceptual learningSpatial relations learningUnsupervised learning

Related Experiment Videos

  • Developed a model that learns 'containment' using only perceptual processes available in early infancy.
  • Employed 'paradoxical occlusion' detection as a source of internal supervision, eliminating the need for external labels.
  • Simulated the learning process based on infants' natural visual experience ('just looking').

Main Results:

  • The model rapidly acquires 'containment' and related concepts without external guidance.
  • The model's learning trajectory mirrors the observed developmental order of these concepts in infants.
  • Detection of paradoxical occlusion provides effective internal supervision for concept formation.

Conclusions:

  • Infant-like learning of 'containment' is achievable through unsupervised, internally supervised mechanisms.
  • Implicit internal supervision, such as paradoxical occlusion, can guide the acquisition of meaningful concepts.
  • This approach offers a pathway for developing more capable artificial intelligence systems with reduced reliance on labeled data.