We’re using the Large Hadron Collider to make toast: Why AI is wasting its potential on email and what World Models could do instead

We're using some of the most sophisticated computational infrastructure ever built to... write better emails and summarise Zoom calls. Meanwhile, Yann LeCun keeps reminding everyone that we can't even replicate how a house cat understands physics. Our cats know unsupported objects fall. They plans complex sequences. They haz cheeseburgerz. Our "revolutionary" AI can write sonnets about quantum mechanics but has like no grasp of how a ball rolls down a hill.

This article argues we're wasting AI's potential on the wrong problems. Instead of automating knowledge work, we could be building world models that actually understand causality and physical reality, like robotics that handle chaotic disaster zones, or climate modeling sophisticated enough to help solve the crisis instead of just documenting it...

I think it's less about whether AI transforms everything (that ship has sailed). It's whether we build systems that replace human judgment or make us something more.