For decades, string theory has been hailed as the leading candidate for the theory of everything in our universe. Yet despite its mathematical elegance, the theory still lacks empirical evidence.

One of its most intriguing, yet vexing, implications is that if all matter and forces are composed of vibrations of tiny strands of energy, then this allows for a vast landscape of possible universes with different physical properties, varieties of particles and complex spacetimes. How, then, can we possibly pinpoint our own universe within a field of almost infinite possibilities?

Since 2005, Cumrun Vafa(opens a new tab) at MIT has been working to weed out this crowded landscape by identifying which hypothetical universes lie in a ‘swampland’ with properties inconsistent with the world we observe. In this episode of The Joy of Why, Vafa talks to co-host Janna Levin about the current state of string theory, why there are no more than 11 dimensions, how his swampland concept got an unexpected lift from the BICEP array, and how close we may be to testable predictions.

Geometry is one of the oldest disciplines in human history, yet the worlds it can describe extend far beyond its original use. What began thousands of years ago as a way to measure land and build pyramids was given rigor by Euclid in ancient Greece, became applied to curves and surfaces in the 19th century, and eventually helped Einstein understand the universe.

Yang-Hui He sees geometry as a unifying language for modern physics, a mutual exchange in which each discipline can influence and shape the other. In the latest episode of The Joy of Why, He tells co-host Steven Strogatz how geometry evolved from its practical roots in ancient civilizations to its influence in the theory of general relativity and string theory — and speculates how AI could further revolutionize the field. They also discuss the tension between formal, rigorous mathematics and intuition-driven insight, and why there are two types of mathematicians — “birds” who have a broad overview of ideas from above, and “hedgehogs” who dig deep on one particular idea.

Large language models (LLMs) are becoming increasingly more impressive at creating human-like text and answering questions, but whether they can understand the meaning of the words they generate is a hotly debated issue. A big challenge is that LLMs are black boxes; they can make predictions and decisions on the order of words, but they cannot communicate the reasons for doing so.

Ellie Pavlick at Brown University is building models that could help understand how LLMs process language compared with humans. In this episode of The Joy of Why, Pavlick discusses what we know and don’t know about LLM language processing, how their processes differ from humans, and how understanding LLMs better could also help us better appreciate our own capacity for knowledge and creativity.