Some currently popular and successful deep learning architectures display certain pathological behaviors (e.g. confidently classifying random data as belonging to a familiar category of nonrandom images; and misclassifying miniscule perturbations of correctly classified images). It is hypothesized that these behaviors are tied with limitations in the internal representations learned by these architectures, and that these same limitations would inhibit integration of these architectures into heterogeneous multi-component AGI architectures. It is suggested that these issues can be worked around by developing deep learning architectures that internally form states homologous to image-grammar decompositions of observed entities and events.
展开▼
