Hacker News Reader

Comment by RcouF1uZ4gsC | Hacker News Reader

Sometimes I think there is other information that gives the physics its meaning.

For example, if you took a Shingled Magnetic Recording hard drive filled with VP9 encoded movie files on a NTFS partition, even if you perfectly understood the physics, and figured out all the individual magnetic fields, you would still have a rough time making sense of the of what was on it without this higher level information information.

View on HN

Sure, but part of the issue is that we can make accurate predictions of pretty much everything we can readily observe today. If there is higher structure, we can't even see the noise yet.

There's much work to be done, but unless something really amazing happens it'll be a long time before we get anywhere near new fundamental(there's still loads of work to be done elsewhere) physics

> Sure, but part of the issue is that we can make accurate predictions of pretty much everything we can readily observe today.

Actually there are a lot of predictions we can’t make today. Stuff that are very important

1) Why do most young people do fine with COVID-19 but some apparently healthy people do not?

2) How exactly do proteins fold

3) Will it rain tomorrow at location x at time x

4) when will my pet tiger decide to maul me?

5) will a drug with this molecular structure for this disease actually treat the disease?

We know a load about the basic building blocks. About quarks and protons and electrons. But once we get to more complicated systems composed of these building blocks, we struggle to make predictions.

See for example the scrubbed SpaceX launch due to weather.

I am sure NASA and spaceX have scientists who are experts in physics, but they could not predict in advance that there would be bad weather on that exact day.

What we do now is make calculations on “spherical cows” and call everything else “chaotic” and “too many variables” and pat ourselves on the back for reducing everything to its basic blocks.

None of the things you listed depend on explanations outside of the realm of physics that we already understand. Theoretically with a simulator powerful enough and using the principles we already know, all of the items you listed could be solved.

I think there's a difference between understanding a system well enough to make predictions, and recreating a system (a simulation) and then seeing what it does. Many times in history before the physics was properly understood, the best scientists and engineers could do was create some kind of "dry run" and simulate/recreate whatever process they were testing, and use that to predict outcomes of the real one. But actually being able to look at the state of a live system and predict what it will do next accurately is quite different.

It seems to me to be more of a spectrum, where a 'well-understood' system is one we can simulate accurately with extremely pared-down models. (I.e., we can simulate it much more cheaply than running the whole atomic simulation) because we found symmetries/simplifications/shortcuts. Approximating ballistics arcs with parabolae, for example, is sort of a compression algorithm on the full simulation.

Some systems are simply random. No amount of simulation will tell you exactly when an unstable atom will decay.

We've spent the last century arguing about what this means for physics, and we still have no idea - even though quantum events can have macroscopic consequences. (To give a contrived example - a decaying atom may damage some human DNA and cause cancer in a historically important individual who dies decades earlier than they would have otherwise, changing the course of history.)

It's more accurate to say that we can model a small selection of mathematically 'pure" system types that are neither random nor emergent, on timescales that are comprehensible to us.

I think the problems you described are in two different groups though. The first group, is where we understand everything well, but we don't have the ability to track every variable accurately and record the amount of data required, in order to make good predictions. Weather falls into this category.

Then there are those like protein folding and animal behaviour where we understand the building blocks, but even of we could track every single building block there would be something missing for us to make predictions.

In case of protein folding it's mostly that the optimization problem is NP-hard.

That's what we do because it's all we can do. Meaning it's likely predicting all those things you listed are as hard to predict as the lotto numbers of next week, for which even if we knew the exact position of every ball in the lottery ball machine at atomic level we would need to predict the thoughts of the person who will spin the machine, for that level of predictions nothing short of a full simulation of reality (ala Matrix) would suffice, and even if we already had a machine powerful enough to do that inputting the state of every atom into that machine would still be outside our capabilities, and I'm afraid it will always be.

Fluid dynamics is one of those problems I was referring to at the end.

I'm not a CFD guy but it's partly an issue of not having enough computing power - fluid dynamics doesn't scale the same way as regular mechanics so you have to compute the whole picture.

That's a good one. I also wonder how many other maps there are, and what _they're_ good at. Consider physics really deep and full of leverage that's currently really helpful to humanity...but still only one general system of map design. The nice thing is that lots of alternatives are considered within physics, but we also like our "best" concept; take Wolfram finding the "best" universal rule system for example. This immediately begins discarding alternatives. And maybe they're just different, not just worse.

Even the map metaphor starts to break at this point because it's hard to think of an alternative map that's not a literal map, and maps all seem the same (i.e. resolving down to the geometry-based, literal "map system" we already know) if you squint your eyes enough. The map is the territory, and it's not, and we can say the same of the pencil, and your shoes, and the concept of intuition. Just as a way of putting "known physics is the map" into context.

Check out whitehead’s process philosophy. Deleuze has something similar.

There are certainly levels that we can't comprehend because we are physically incapable of it, just as an ant is incapable of understanding a book. Perhaps with neural augmentation we will discover new vistas previously unavailable to us.

Obligatory xkcd https://xkcd.com/638/

Your "ant" comment hits home :)

That's because the specific information on the disk was initially generated by, and is the consequence of, a great many previous events. Even if we assume the disk holds only the complete works of Shakespeare, we would need to understand vastly more than just the drive's physics, in order to understand it completely. That does not imply that the explanation for the drive, as we find it with its specific contents, is not reducible to physics, but performing such a reduction is not the path to understanding it.

Reply on news.ycombinator.com