Terence Tao: The shift to data-driven science, why verification is as crucial as creativity, and the challenges of AI in peer review | Dwarkesh

AI-generated theories challenge traditional peer review, prompting a shift in scientific evaluation methods.

Listen on Dwarkesh Podcast

Share

Add us on Google by Editorial Team Mar. 23, 2026

Key Takeaways

• Scientific discovery often begins with empirical data analysis before theoretical frameworks are developed.
• The balance between idea generation and data verification is crucial in scientific research.
• Modern scientific methods have shifted towards data-driven analysis rather than traditional hypothesis-driven research.
• The quantity of data is critical in forming reliable scientific conclusions.
• The traditional peer review system is struggling to keep up with the volume of AI-generated scientific theories.
• Scientific structures need to evolve to effectively evaluate the massive scale of theories generated by AI.
• Cultural and societal factors heavily influence the adoption and success of scientific ideas.
• Initial scientific theories may seem less accurate but can lead to deeper understanding over time.
• Copernicus’ simpler theory laid the groundwork for future advancements despite being initially less accurate.
• We are experiencing a cognitive revolution, similar to the Copernican revolution, regarding human intelligence.
• Historical scientific shifts can provide valuable insights into current evolutions in intelligence.
• The interplay between scientific innovation and societal acceptance is crucial for understanding how ideas evolve.

Guest intro

Terence Tao is a professor of mathematics and the James and Carol Collins Chair at the University of California, Los Angeles (UCLA). He earned his PhD from Princeton University at age 20 and received the Fields Medal in 2006 for contributions to partial differential equations, combinatorics, harmonic analysis, and additive number theory. With Ben Green, he proved the Green-Tao theorem demonstrating arbitrarily long arithmetic progressions of prime numbers.

The role of empirical data in scientific discovery

• Kepler’s laws of planetary motion were derived from empirical data rather than theoretical frameworks.

• He worked out the two laws of planetary motion of the ellipses also equal areas super equal times

  — Terence Tao

• Data analysis can precede theoretical understanding in scientific discoveries.
• Newton provided a theoretical explanation for Kepler’s laws a century later.
• Historical context highlights the difference between empirical observation and theoretical explanation.

• It took Newton a century later to give a theory that explained all three laws at once

  — Terence Tao

• Empirical findings can be critical in the initial stages of scientific discovery.
• Understanding the historical context of scientific discovery is essential for modern research methods.

Balancing creativity and verification in research

• Scientific discovery involves both idea generation and rigorous data verification.

• The ones we celebrate are these of eureka genius moments of idea generation

  — Terence Tao

• Verification is as important as creativity in scientific research.

• It has to be matched by an equal amount of verification otherwise it’s slop

  — Terence Tao

• The interplay between generating hypotheses and verifying them is crucial for progress.
• Creativity in science must be grounded in empirical validation.
• Celebrating only the idea generation overlooks the importance of verification.
• Scientific progress relies on both innovative ideas and their rigorous testing.

The shift towards data-driven scientific methods

• Scientific progress has shifted from hypothesis-driven research to data-driven analysis.

• Now it’s almost reversed; you collect big data first and then you try to get hypotheses from it

  — Terence Tao

• Modern scientific methods emphasize the importance of data in forming hypotheses.
• Traditional hypothesis testing has evolved with advancements in data analysis techniques.
• The evolution of scientific methods reflects the growing importance of data.
• Data-driven analysis allows for new hypotheses to emerge from large datasets.
• Understanding this shift is essential for adapting to modern scientific challenges.
• The role of data in scientific discovery has become increasingly prominent.

The importance of data quantity in scientific conclusions

• Kepler’s third law illustrates the importance of data quantity in scientific conclusions.

• Kepler had like six data points to like every planet he knew the length of the orbit and the distance of the sun

  — Terence Tao

• Limited data can hinder the reliability of scientific theories.
• The historical context of Kepler’s work highlights the need for sufficient data.
• Data sufficiency is critical in forming reliable scientific laws.

• That’s not enough data to be really reliable

  — Terence Tao

• The role of data quantity in scientific theories is crucial for accurate conclusions.
• Understanding the significance of data quantity is essential for scientific research.

Challenges in the peer review system due to AI

• The traditional peer review system is overwhelmed by AI-generated scientific theories.

• Human reviewers we just it’s just they’re already being overwhelmed

  — Terence Tao

• AI-driven submissions are flooding scientific journals.
• The impact of AI on scientific publishing presents significant challenges.
• Scientific structures need to adapt to the volume of AI-generated theories.

• We have to the rest of the aspects of science have to catch up

  — Terence Tao

• Verification and validation processes must evolve with technological advancements.
• The necessity for evolving scientific methodologies in response to AI is critical.

Cultural and societal influences on scientific ideas

• The adoption and success of scientific ideas depend heavily on cultural and societal factors.

• It depends on the future… it depends on also on the culture and society

  — Terence Tao

• Evaluating scientific progress requires understanding its societal context.
• Societal acceptance plays a crucial role in the evolution of scientific ideas.
• The interplay between innovation and societal factors is essential for scientific advancement.
• Understanding cultural influences is vital for assessing scientific ideas.
• The success of scientific ideas is intertwined with societal dynamics.
• Scientific innovation must consider cultural and societal contexts for broader acceptance.

The evolving nature of scientific theories

• Scientific progress often involves recognizing that initial theories may lead to deeper understanding.

• Science is always a work in progress

  — Terence Tao

• Initial theories may seem less accurate but can contribute to greater understanding.
• The historical context of scientific theories highlights their evolution over time.

• When you only get part of the solution it looks worse than a theory which is incorrect

  — Terence Tao

• The nature of scientific progress involves continuous refinement and understanding.
• Recognizing the work in progress nature of science is crucial for its advancement.
• Scientific theories evolve as new insights and data become available.

Simplicity versus accuracy in scientific theories

• The Copernican theory was simpler but initially less accurate than the geocentric model.

• Copernicus’ theory was a lot simpler but much less accurate

  — Terence Tao

• Simplicity can lay the groundwork for future scientific advancements.
• Kepler’s work improved the accuracy of Copernicus’ theory.
• The paradox of simplicity versus accuracy is evident in scientific theories.
• Understanding historical developments in astronomy provides insights into scientific evolution.

• There was only Kepler that made it more accurate than Thomley’s theory

  — Terence Tao

• The evolution of astronomical theories highlights the balance between simplicity and accuracy.

The cognitive revolution in understanding intelligence

• We are experiencing a cognitive version of the Copernican revolution regarding human intelligence.

• Now we’re actually seeing that there’s very different types of intelligence

  — Terence Tao

• Our perception of intelligence is undergoing significant reevaluation.
• Historical scientific shifts provide valuable insights into current evolutions in intelligence.
• The need for reevaluating our understanding of intelligence is highlighted by this revolution.
• The analogy with the Copernican revolution frames the current evolution of intelligence.
• Understanding the implications of this cognitive revolution is essential for future advancements.
• The evolution of intelligence requires a reevaluation of tasks and their associated intelligence.

Historical context and implications for scientific progress

• Historical scientific shifts can provide valuable insights into current scientific challenges.
• Understanding the historical context of scientific discovery is essential for modern research.
• The evolution of scientific methods reflects the growing importance of data.
• Historical developments in astronomy provide insights into scientific evolution.
• The interplay between scientific innovation and societal acceptance is crucial for progress.
• The historical context of scientific theories highlights their evolution over time.
• Recognizing the work in progress nature of science is crucial for its advancement.
• Historical shifts in intelligence perception highlight the need for reevaluation in modern contexts.

Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our Editorial Policy.