In this episode of Beyond Words, host Garrett Oyama, MS, CCC-SLP, welcomes Dr. Benjamin Bergen, cognitive scientist at UC San Diego, to explore what research on human language and large language models (LLMs) can teach us about meaning, understanding, and communication.
Dr. Bergen shares insights from his research on embodied simulation, which examines how our brains may draw on sensory and motor experiences to understand language. Together, they discuss how humans process words, actions, and concepts, and what happens when we compare these processes to the remarkable language abilities of AI. The conversation explores what LLMs can reveal about human communication, including how humans and machines differ in their understanding of meaning, social interactions, and the world around them.
The discussion also addresses the growing ability of AI to imitate human conversation, persuade, and even deceive. Dr. Bergen shares findings from research on AI performance in Turing tests and the conditions under which language models may produce deceptive responses without being instructed to do so. For speech-language pathologists, these findings raise important questions about the use of AI in clinical practice, particularly when evaluating AI-generated information or considering how children and families interact with conversational AI. Throughout the episode, Garrett and Dr. Bergen examine why language that sounds convincingly human doesn't necessarily reflect human-like understanding, trustworthy information, or shared intentions.
Dr. Bergen shares insights from his research on embodied simulation, which examines how our brains may draw on sensory and motor experiences to understand language. Together, they discuss how humans process words, actions, and concepts, and what happens when we compare these processes to the remarkable language abilities of AI. The conversation explores what LLMs can reveal about human communication, including how humans and machines differ in their understanding of meaning, social interactions, and the world around them.
The discussion also addresses the growing ability of AI to imitate human conversation, persuade, and even deceive. Dr. Bergen shares findings from research on AI performance in Turing tests and the conditions under which language models may produce deceptive responses without being instructed to do so. For speech-language pathologists, these findings raise important questions about the use of AI in clinical practice, particularly when evaluating AI-generated information or considering how children and families interact with conversational AI. Throughout the episode, Garrett and Dr. Bergen examine why language that sounds convincingly human doesn't necessarily reflect human-like understanding, trustworthy information, or shared intentions.
Show more
Less
