AI tackles a millennium problem in fluid dynamics

Fluids lacking rigid structures, such as gases, water, and oceanic currents, take the shape of their containers, and their movement integrates with Isaac Newton's laws of motion.

These equations integrate with Isaac Newton's laws of motion to explain various physical phenomena, defining fluids broadly to include gases, water, and entire oceans.

Astrophysical processes like star and galaxy formation are essential to model as fluids because matter commonly behaves this way in space, according to researcher Um.

Most matter observed in astrophysics behaves as a turbulent fluid, meaning these equations are consistently applicable to both terrestrial events like weather patterns and cosmic-scale occurrences.

By redirecting airflow to minimize drag, researchers like Mordecai-Mark Mac Low demonstrate that micro-turbulence in golf ball dimples improves aerodynamic performance rather than acting as a purely detrimental force.

For instance, golf ball dimples create controlled micro-turbulence that helps air flow smoothly around the ball, reducing drag and increasing travel distance.

Formulated in the 19th century, the Navier-Stokes equations remain the subject of a long-standing debate regarding whether they can mathematically 'blow up' by reaching an infinite value, or singularity.

A long-standing debate has persisted over whether these equations are always stable and finite, or if they can yield regions where fluid speed reaches infinity, indicating a failure of the model to perfectly describe reality.

Physicists actively seek smooth solutions that avoid such singularities, ensuring the equations remain valid and predictive over time and across all conditions.

OpenAI utilized AI to identify a counterexample to the Navier-Stokes smoothness hypothesis, a process that cost $6 million in GPU time over 88 hours.

The objective was to prove if the equations are always smooth or if they can lead to infinite values, with OpenAI's AI identifying a scenario where they are not smooth.

The result was verified using 'Lean,' a mathematical verification language, and the proof process reportedly required approximately $6 million in GPU costs over 88 hours.

The proof involves approximately 26,000 sub-theorems, which necessitates significant reliance on the reliability of the Lean verification software and raises questions about whether the underlying model was trained on private inputs.

The Navier-Stokes equations assume a continuous, indivisible fluid, which simplifies the physical reality that fluids like gas consist of discrete atoms.

Boltzmann equations, which describe systems of particles, provide a more granular account of physical realities that Navier-Stokes' continuous fluid model simplifies, making the Navier-Stokes an approximation.

Deriving Navier-Stokes from Boltzmann's framework requires the assumption that changes between regions are small; however, when gradients become large, such as in cases leading to infinity, this approximation fails, resulting in 'non-smooth' mathematical scenarios.

Physicists view this as identifying a gap or mismatch between purely mathematical descriptions and the more complex realities of physical phenomena.

Partial differential equations, such as Navier-Stokes, are fundamental to understanding non-solid fluid dynamics, including blood flow and gas movements.

Solid mechanics also heavily relies on specific sets of partial differential equations, underscoring their broad applicability across physics.

Essentially, physics functions as a complex system of partial differential equations integrated with set theory, forming the basis of our understanding of the world.

Human mathematicians provide context and readable logic, which AI black boxes currently lack, as evidenced by the unreadable nature of the vast 'Lean' statements generated by OpenAI's solution for human researchers.

Human-led research, such as the ongoing work by Buckmaster and Alp, prioritizes understanding the process and the 'why' behind the solution, contrasting with AI's focus on simply providing an answer.

The speaker primarily utilizes AI as a tool for navigating and exploring scientific literature to support his astrophysics research.

This application involves using tools that guarantee the reality of cited sources, even though the AI may still generate interpretive hallucinations or errors in authorship.

Mordecai-Mark Mac Low uses AI to explore research topics such as interstellar dust and radiation.

The tool ensures that all identified academic sources are real documents, significantly aiding in the discovery of relevant papers for research.

The system frequently identifies valid academic papers for user queries, though the guest warned that it often fails by inventing incorrect authors for those specific documents.

The guest noted that while the existence of a paper is verifiable, manual checking of authorship data remains essential due to these AI-generated inaccuracies.

Representatives from OpenAI and other major AI firms are currently engaging with world leaders during UN week, highlighting AI's emergence as a significant geopolitical and societal force.

Neil deGrasse Tyson emphasized that AI is a real and present technology, urging continuous public vigilance over its long-term societal effects, regardless of the immediate consequences of specific technical challenges like those in physics.