Can machines actually think?

In a world where chatbots draft emails, explain quantum mechanics, Quanta magazine's The Joy of Why podcast tackles the question on everyone's mind: do these systems actually understand anything? In the episode Will AI Ever Have Commonsense?, host Steven Strogatz talks to University of Washington computer scientist Yejin Choi about hidden strengths and limits of LLMs like ChatGPT.

Choi highlights the central mystery: these models are 'black boxes', as we don't know how they arrive at their answers. Yet, astonishingly, they've acquired a striking amount of what looks like common sense, an ability long thought impossible to encode in machines.

She explains how her lab is experimenting with ways to feed such everyday reasoning into neural networks, sometimes by mimicking how children learn - asking endless questions and building rules from them. The conversation balances technical insight with accessibility.


Strogatz's curiosity and Choi's clarity bring to life one of the defining puzzles of modern AI: whether intelligence can be reduced to statistical patterns, or something deeper is at play. Thought-provoking and timely, this episode is essential listening for anyone intrigued by the promises - and pitfalls - of AI.