Skip to main content
Aller à la page d’accueil de la Commission européenne (s’ouvre dans une nouvelle fenêtre)
français fr
CORDIS - Résultats de la recherche de l’UE
CORDIS

Development of Visual Cognition in Infants and Machines

Objectif

Like infants, modern artificial intelligence (AI) often begins by learning patterns in large streams of unstructured data. While impressive in some tasks, in others it falls short; it is worse than infants at generalising to new tasks and requires vastly more data and energy, at considerable financial and ecological cost. However, infant-inspired algorithms might close this gap. In turn, it would be invaluable for developmental science to have computational models of how understanding develops from the interaction of the brain’s intrinsic constraints with an infant’s experience. Developmental science and AI would benefit enormously from being brought together but remain deeply disconnected due to the difficulty of measuring infants’ cognition from their behaviour. This has led developmental science to focus on proof-of-concept “toy problems” while in contrast AI has focused on rich naturalistic datasets relevant to applications. To bridge the disciplines, we will use new neuroimaging tools that allow rich measurement of infant’s internal states. With awake functional magnetic resonance imaging, we will measure the brain representations of rich naturalistic visual stimuli in infants from 2-months-old, and with optically pumped magnetometer magnetoencephalography, we will obtain complementary measures of dynamic processing, which we will compare to AI models. We will focus on a dynamic research area in developmental science and AI, visual cognition, the process through which humans and machines understand what they see. Our objectives will be to investigate the understanding of people and objects; their physical and social relations; and predicting what they will do next. Using DNN models of the development of visual cognition, we aim to elucidate the mystery of how infants learn to see and understand. In turn, we aim to inspire a paradigm for the development of the next generation of AI that performs better, is more adaptive and efficient, and ultimately, more human.

Champ scientifique (EuroSciVoc)

Mots‑clés

Programme(s)

Thème(s)

Régime de financement

HORIZON-ERC - HORIZON ERC Grants

Voir tous les projets financés dans le cadre de ce programme de financement

Appel à propositions

(s’ouvre dans une nouvelle fenêtre) ERC-2025-ADG

Voir tous les projets financés au titre de cet appel

Institution d’accueil

THE PROVOST, FELLOWS, FOUNDATION SCHOLARS & THE OTHER MEMBERS OF BOARD, OF THE COLLEGE OF THE HOLY & UNDIVIDED TRINITY OF QUEEN ELIZABETH NEAR DUBLIN
Contribution nette de l'UE
€ 3 499 845,00
Adresse
COLLEGE GREEN TRINITY COLLEGE
D02 CX56 Dublin
Irlande
Région
Ireland Eastern and Midland Dublin
Type d’activité
Higher or Secondary Education Establishments
Liens
Coût total
Aucune donnée

Bénéficiaires (1)