ERC: The Self-Teaching Brain (BrainPlay)
The BrainPlay grant will approach the foundation of the human intellect, the mystery of biological learning, by using new theoretical concepts to navigate a new cognitive and behavioural paradigm. Specifically, our analysis of brain function will shift from the study of learning to the analysis of self-teaching. If we succeed, we will deliver the learning algorithms that subserve biological intelligence in terms of the neural cost functions that they optimise the signals and mechanisms that implement them, the cognitive functions that they target and the behaviours that they require to drive them. In doing so, we will do justice to play – a key self-teaching behaviour. Play has been neglected by mainstream neuroscience, but we showed it results in astounding cognitive benefits. The BrainPlay project will not just open a new window into the inner workings of our minds but might also have profound implications for medical and technological applications. We will pioneer a new functional characterization of synaptic plasticity, transfer powerful concepts of machine learning, such as maximal margin learning and backpropagation, to biologically plausible models of spiking neural networks. We will uncover the neural underpinnings of the brain state of play and of the cognitive improvements of intense playing. The distinguishing feature of the BrainPlay grant is its interdisciplinary ambition: We will put playfulness centre stage in the study of synaptic plasticity, a link that is entirely unexplored. We study brains, while they play and engage in seemingly useless playful behaviours, behaviours driven by the subject’s imagination rather than the experimenters reward schedule. We bridge from human psychophysics to ego-shooter games to self-teaching in novel “playful” neural network architectures. We got together to find a new field of research, BrainPlay, the study of the self-teaching brain.
The BrainPlay grant brings together very diverse researchers, who started to uncover the remarkable cognitive benefits from playful learning (Bavelier), who have started to record neural activity in playing animals (Brecht), who are able dissect the molecular and cellular mechanisms of synaptic change in self-teaching brains (Schmitz) and accomplished breakthroughs in the theory of neuronal self-teaching (Gütig).
Mittelgeber
H20: ERC Synergy Grant
Laufzeit
Projektstart: 08/2019
Projektende: 01/2026
Forschungsbereiche
Neurowissenschaften
Forschungsfelder
Neurowissenschaften
