Ref link: Miles Cranmer – The Next Great Scientific Theory is Hiding Inside a Neural Network (April 3, 2024)
Academia and science are finally cottoning onto the idea of how to put AI and neural networks TO work to avail themselves of rapid foundation and equation discovery and thus, incorporation to validation of theory and equation.
They intend to make these models open to the world, which is amazing and may well be the offset to lobby and corporation-driven control.
This is likely one of the better overviews I’ve seen, albeit extraordinarily math heavy and assumptive of all manner of prior knowledge. I will continue to post insightful overviews as I find them, this is but the first conceptual overview of what will assuredly be a rapidly moving and expanding practice.
One thing I will note is the rising realization that simplicity is mostly familiarity and not actual simplicity. As I once said, if you asked a human to open a door and asked a robot to open a door, the human would do it while the robot couldn’t possibly. Not because the robot cannot be instructed to do so, but because before you can teach it how, you essentially have to program all the knowledge and methods and movements required, and do it well enough that it can (at least) approximately the human’s cumulative education on how to interpret and respond to such a request.
It begins!