Nobel laureate Geoffrey Hinton, a pioneer in AI, has expressed his surprise at the ability of neural networks to understand natural language. He stated that these networks are "much better at processing language than anything ever produced by the Chomskyan school of linguistics."
However, Veena D. Dwivedi, a neuroscientist specializing in the study of human language, respectfully disagrees with the notion that AI can truly "understand." She emphasizes the distinction between written text and natural language, highlighting that the same language can be represented by different visual symbols, such as Hindi and Urdu, which are mutually intelligible but use different writing scripts.
Dwivedi points out that linguistic communication involves more than just words. It includes context, such as the speaker's tone, facial expressions, and the shared environment. She illustrates this with the example of the sentence "I'm pregnant," which carries different meanings depending on the context.
Her research shows that even an individual's emotional state influences brainwave patterns when processing sentences. Dwivedi clarifies that AI algorithms are not the same as the biological brain networks that characterize human understanding. She also addresses the claim that neural networks surpass Chomskyan linguistics, which focuses on universal grammar and the ease with which infants learn language.
Chomsky's work explores how humans acquire language, proposing an innate module for language learning. Dwivedi concludes that conflating AI with human understanding can have dangerous consequences, especially when scientific terms are misused or misapplied.