We discuss how scientists at Google DeepMind are using a type of AI to find new answers to old questions about how fluids like water or air move around.
They are tackling complex math equations from hundreds of years ago, such as Euler and Navier-Stokes equations, which explain fluid flow but are really hard to solve, especially in three dimensions.
The team looked at what's called singularities, which are spots where things like speed or pressure in fluids get too big to measure, and used a tool called physics-informed neural network (PINNs) to get super accurate results.
They found new kinds of unstable singularities in fluid equations and noticed a pattern in a number called lambda, which measures how fast a singularity grows. This method might help AI tackle other hard problems in math, physics, and engineering that need very precise solutions.
