Comment by jbay808 | Hacker News Reader

Comment by jbay808 | Hacker News Reader

Parent Full thread

I've long been a big fan of "more is different" myself, but perhaps you and I have different takes. I think that essay elegantly defines the quintessential nature of emergent behaviour, and the limits of reductionism. But not by refuting Feynman's claim; rather by exploring why it is that when a lot of stuff follows simple rules, the results aren't at all in keeping with the character of those rules, and appear to be better described by rules of their own.

In other words, once there are enough pieces on the board, you get blindsided as your queen is captured by a rook in a move you never saw coming, even though in retrospect you can go back and see that, yes, the rook really did move in a straight line.

That's why, even if biology really is "just" applied chemistry and so on, you'd never predict the wonderful behaviour of biology if you had only studied systems of less than ten molecules.

And I don't think Feynman would disagree at all, nor would he claim that knowing just the basic laws of physics would let you anticipate the character of cell biology or turbulent fluids, or all the other wonderful systems that have nontrivial degrees of freedom.

Ok good point, I think I got side tracked with my comment.

Perhaps a better analogy would be: Someone who claims to be a chess player, but keeps getting disqualified from tournaments for making illegal moves.

For example, when it comes to QCD and the structure of the nucleus, we have no idea if our theories are correct or even make sense. One of the millon dollar Millenium prizes is currently being offered for anyone that can prove the equations of QCD are even well defined or if it is possible to even have a bound state (i.e for particles to exist at all) within a QCD framework.

So my point is the underlying theory is really only described in the perturbation theory setting. I.e., in the high energy limit (smashing things together). Now the equations and mathematics behind QCD are very attractive and probably contain a great many aspects of what we will eventually become the full theory, but until we can predict, for example, the mass of a nucleus from those equations (or anything at all whatsoever), then they are certainly not proven.

So far all we know is that these equations can predict what comes out (and in what direction) when we smash particles together.

It's true that a mathematically rigorous proof that... well... any QFT exists at all escapes us for all but extremely special cases.

But, given you've given a whole spiel about about EFTs, you can think of the lattice formulation of QCD as an EFT whose cutoff is the lattice spacing. We can compute the mass of hadrons and take the continuum limit / cutoff to infinity. And when we take that limit holding some physical hadronic observables fixed, the spectrum matches perfectly. So I think it's relatively unfair and misleading to say that because we can't do it with pen and paper we can't make any low-energy QCD predictions at all whatsoever.

What always drives me crazy about stories like the OP is that strongly coupled theories are automatically technically natural. But most phenomenologists don't like them because they, not being proficient supercomputer programmers, can't calculate anything.

Last I checked the there were enough free parameters in these calculations to make the calculations of less than convincing. For example those 'hadronic observables' make up such a big part of the actual results that what is left over is within the error bounds anyway.

However, that was over 7 years ago, and someone else in the comments linked an article talking about the breakthroughs since then, and they seem significant, so things may well have improved since then.

Regardless, lattice QCD has always been real science and always been quite impressive. Any direct attempt to match theory with known experiments always is real science regardless of whether you agree with the methods.

> Last I checked the there were enough free parameters in these calculations to make the calculations of less than convincing. For example those 'hadronic observables' make up such a big part of the actual results that what is left over is within the error bounds anyway.

Indeed, this is more like a fifteen-year-old understanding. For example in 2008 , the authors of https://inspirehep.net/literature/804184 tuned to reproduce the pion and kaon masses (to fix the quark masses) and then predict ~ a dozen further masses, and in 2015 they computed the proton/neutron mass splitting https://inspirehep.net/literature/1300659 . In https://inspirehep.net/literature/1675748 we reproduced the nucleon axial coupling to ~1%.

Reply on news.ycombinator.com