Terence Tao – Kepler, Newton, and the true nature of mathematical discovery - Dwarkesh Podcast Recap

Podcast: Dwarkesh Podcast

Published: 2026-03-20

Guests: Terence Tao

What Happened

Johannes Kepler's groundbreaking work on planetary motion started with a flawed theory involving platonic solids but eventually led to the discovery that planets orbit in ellipses, not circles. This revelation was supported by Tycho Brahe's extensive astronomical data, culminating in Kepler's three laws of planetary motion. Isaac Newton later provided the theoretical underpinnings for these laws, illustrating the iterative nature of scientific discovery where hypotheses are refined through data.

Today's scientific discovery often begins with data analysis, a shift from the traditional hypothesis-first approach. Terence Tao points out that Kepler's work was an early example of this data-driven method, though Kepler himself started with preconceived theories. AI has further transformed the landscape by reducing the cost of idea generation, moving the bottleneck to verification and validation of the vast number of theories generated.

AI tools have made significant strides in mathematics, solving 50 out of 1,100 Erdős problems. However, progress has plateaued, with AI excelling at applying standard techniques at scale but struggling with inventing new ones. This highlights the distinction between artificial cleverness and intelligence, as AI lacks cumulative progress and relies heavily on trial and error.

The social aspect of science is crucial, as the acceptance of theories often depends on effective communication and societal context. Historical examples include Charles Darwin's theory of evolution, which was accepted more readily due to effective communication, while Isaac Newton's work faced initial skepticism due to its complexity and Newton's secretive nature.

AI's role in mathematics is growing, with tools enhancing productivity by handling secondary tasks like generating plots and reformatting text. While AI is not yet a replacement for human mathematicians, it complements human work and is expected to take over many tasks within a decade.

Terence Tao identifies as a 'fox' who knows a little about many things, as opposed to a 'hedgehog' who knows one thing deeply. He emphasizes the importance of serendipity in scientific work, noting that unexpected interactions often lead to valuable insights. Remote meetings during COVID-19 reduced such serendipitous interactions, impacting academia.

Writing blog posts helps Tao retain knowledge, as he often forgets details over time. He predicts that hybrid human-AI collaboration will dominate mathematics for a long time, with AI allowing for non-traditional opportunities, enabling contributions from individuals at earlier stages of education.

Key Insights

• Kepler's work on planetary motion illustrates the iterative nature of scientific discovery, as his initial theory involving platonic solids was refined through data analysis to reveal elliptical orbits.
• AI has transformed scientific discovery by shifting the bottleneck from idea generation to verification and validation, as it generates vast numbers of theories at a reduced cost.
• AI tools in mathematics have addressed 50 out of 1,100 Erdős problems, highlighting their capability to apply standard techniques at scale while struggling with inventing new methods.
• The acceptance of scientific theories often depends on effective communication and societal context, as demonstrated by the different receptions of Darwin's and Newton's theories.