Bootstrapping Poverty of the Stimulus Through Random Projections

Language involves several levels of abstraction, from small sound units like phonemes to contextual sentence-level understanding. Large Language Models (LLMs) have shown an impressive ability to predict human brain recordings. For instance, while a subject is listening to a book chapter from Harry Potter, LLMs can predict parts of brain imaging activity (recorded by functional Magnetic Resonance Imaging or Electroencephalography) at the phoneme or word level. These striking results are likely due to their hierarchical architectures and massive training data. Despite these feats, they differ significantly from how our brains work and provide little insight into the brain’s language processing. We will see how simple Recurrent Neural Networks like Reservoir Computing can model language acquisition from limited and ambiguous contextual data better than LSTMs.My work includes modeling language comprehension, language acquisition from a robotic perspective, sensorimotor models, and extended models of Reservoir Computing to model working memory and hierarchical processing. I propose creating a new generation of neural-based computational models of language processing and production, utilizing biologically plausible learning mechanisms that rely on recurrent neural networks. This approach involves developing novel sensorimotor mechanisms to account for the shaping of action-perception, building hierarchical models from the sensorimotor to the sentence level, and embodying these models in robots.I aim to model general hierarchical sensorimotor processes; thus, our models are not only relevant to language or vocal learning, but are interesting for a larger set of sensorimotor tasks.