[jiggawatts]

An analogy is that one can walk in the right direction for an awfully long way before getting stuck at the end of a valley, staring up a proverbial cliff. Someone walking parallel to you on a ridge might be nearby the whole time, but avoid being stuck.[1] Anyone studying AI/ML would know about getting stuck in local minima, and having to essentially restart training to shake things loose. Same thing.

My theory is that physics is stuck in a local minima where it's not sufficient to change just one, or even two or three fundamental things to get unstuck. That's too big a leap via traditional incremental publishing of new theories. Any one change to the status quo won't work, and is rejected. Multiple changes are too complex, and might need to have occurred too early in the timeline. It simply might not occur to people that the whole industry took a run turn... 100 years ago.

This is why I like crazy papers. They make you reconsider fundamental notions, the type that were in textbooks decades ago and are seen as foundational and unquestionable.

[1] This literally happened in New South Wales. For decades(!) nobody could cross the Blue Mountains, until Blaxland, Lawson, and Wentworth tried walking on the ridge tops instead of the valleys: https://en.wikipedia.org/wiki/1813_crossing_of_the_Blue_Moun...

[JackFr]

> My theory is that physics is stuck in a local minima where it's not sufficient to change just one, or even two or three fundamental things to get unstuck. That's too big a leap via traditional incremental publishing of new theories. Any one change to the status quo won't work, and is rejected.

Makes me think of Stephen Wolframs current work. Guy's a genius, who's current stuff kind of reads like he's a crank. But at the same time I'm kind of rooting for a revolutionary paradigm that's gonna totally upend things.

[jiggawatts]

He's starting from a totally blank slate and hoping he'll end up at Physics. In some sense, that's guaranteed by definition -- any sufficiently complex foundational system or algebra can represent any other, including the current models of reality. But this has no predictive power. It's like saying digital circuitry is a theory of physics because a computer can run a physics simulation!

My approach is more akin to assigning a lower probability of validity to papers that have long been generally accepted as 100% true. Then I try to hold all of them in my brain simultaneously while reassigning joint probabilities, almost like those computer game map generators that use "quantum decoherence".

The idea is to find a parallel path that goes through most of established physics but avoids the trap of local minima. The challenge is that it's really unclear which existing theories are the traps, and which are true and need to be kept.

Something like this is clearly needed, because existing theories are either contradictory or inconsistent. They can't all be right. Something somewhere must be discarded.