Sounds truly groundbreaking. As a layman to me it’s amazing this works in the same “next-token” way as LLMs that have become so popular this year. I do wonder, how much memory is “too much”? Could we throw more memory at this problem and even if scaling is quadratic, is it feasible on some grand scale? We currently spend a lot of money on supercomputers, what it the same amount of resources was available for ClimaX?
@@Septumsempra8818 So basically in theory you could train a model to take certain economic assumptions to predict market behavior...but unlike physically based models those assumptions are not necessarily as objectively robost. So doing so, in theory if different competing models in the market used AI to guide their investment strategies...but those big firms made different assumptions...like for example that their firms model should beat the market...then that could easily spiral out of control if machines were making all the market decisions with goal of maximizing returns.
Hi Siva, the residual form simply involves moving all terms of the equation to one side, so we have F(x,u) = 0. If the ODE/PDE is satisfied then F(x,u) will equal zero, so minimising this term during training constrains the predicted variables to satisfy the residual, and thus the PDE/ODE. Hope that helps!
