International Journal of Science, Engineering and TechnologyAuthors: Rudy Shoushany
Abstract: The quest for Artificial General Intelligence (AGI) has shifted from specialized, narrow AI systems toward generalized foundation models capable of cross-domain reasoning. This paper explores the pivotal role of multimodal neural networks (MNNs) in this transition. By integrating diverse data streams—including text, vision, audio, and sensory inputs— MNNs mimic the human cognitive process of cross-modal alignment. We analyze current breakthroughs in native multimodal architectures, the shift from strong to weak semantic correlation learning, and the emergence of embodied AI as a critical path toward AGI. Our findings suggest that while MNNs provide the necessary perceptual framework for AGI, the integration of autonomous reasoning and self-correcting feedback loops remains the final frontier.
DOI: http://doi.org/10.5281/zenodo.18383308
