Brainwaves: Is AI actually thinking?
Summary
Host Meghna Chakrabarti explores whether AI truly thinks by testing its ability to create humor with AI expert Melanie Mitchell. Linguist Kyle Mahowald and psychologist Anna Ivanova join to discuss dissociating language from thought, noting that while AI mimics human conversation, it lacks genuine self-awareness or embodied experience. The episode examines how studying these models helps scientists better understand the complex mechanisms of human cognition.
Topics discussed
Intro, sponsors, and On Point Club promotion
Episode intro: Defining human thinking vs AI
Experiment: Asking Claude to write a funny joke
Claude's inappropriate second joke attempt
History of AI: From logic-based to brute force
How LLMs work: Next-word prediction vs human learning
AI capabilities: Riddles, plagiarism, and language limits
Guest Kyle Mahowal: Defining thought and language
AI introspection: Can AI report its internal state?
Sponsors and ad breaks
Self-awareness: Human consistency vs AI memory
AI progress in coding, math, and physical tasks
Using AI to model and understand human brain activity
Outro and teaser for next episode on near-death
Listen ad-free on Castria