Comment by jiggawatts | Hacker News Reader

Comment by jiggawatts | Hacker News Reader

Parent Full thread

To me, this always felt like programmers excusing abhorrent code with "but it works".

Quantum theory isn't. It's not a single, cohesive, consistent theory.[1] There is no recipe you can apply, it's just a bunch of guessworks and heuristics that happen to produce the numerically correct equations if you keep trying long enough. This isn't secret or some sort of external criticism, you'll find this front-and-centre in the foreword of many a QM text!

Schrödinger famously arrived at his equation basically through numerical methods. He just tried things until it "fit" the desired output.

Now, there's nothing wrong with this, per-se. It's a perfectly viable approach for getting going, for getting something and using it as a starting point. But it isn't the endpoint, because approaches like this often have virtually no explanatory power.

A similar example is collecting insects, categorising them, giving them Latin names, and putting them up for display in a museum. You can learn a lot, amass enormous amounts of information, but without a theory of genetics and natural selection you will always be blind to the underlying truth of it all.

QM is just like bug collecting. We're collecting numerical equations that work, but we have essentially no clear understanding why. We've built a tree of life, and nobody has had the lightbulb moment that explains why it's a tree.

[1] There was a paper published a few years back where a bunch of working quantum physicists were asked some simple multiple-choice questions about the fundamentals of the theory. There was no consensus opinion on anything! PS: You'll get similar results if you ask priests of a random Christian sects about the basics of religion. Conversely, you will get nearly zero disagreement if asking Chemists about the basics of their science.

> because approaches like this often have virtually no explanatory power.

the famous feynman have produced a response to this train of thought - usually from the laymen - where they demand an "explanatory theory". The answer is usually "don't know"; why gravity work the way it does? - don't know, but we can calculate orbits to great precision; why does photons or electrons travel the way it does? - don't know, but we can make very precise predictions via Quantum Electrodynamics theory.

The laymen demands an explanatory theory, because there's an underlying assumption they don't even verbalize - that such an explanatory theory can be understood _without the maths_, that the principles of the laws of physics operate on basic, understandable lego pieces. What is somewhat unconceivable is that the explanation _is_ the maths. To whit, we don't even have an explanation for the idea of inertia! We just observe it, and measure it (as mass). There's no explanatory theory.

You could say that the "explanation" of inertia is that which preserves the symmetry of space translations under Noether's theorem. But that doesn't tell a physicist any more than they already knew, and it tells a layman even less.

"don't know" -- that doesn't mean we can't know, it just means we don't currently have an explanation. If you had asked similar fundamental questions about the "why" of beetle species a few centuries ago, even the most prolific beetle collector in the world would have answered "don't know".

"The laymen demands an explanatory theory" -- and so do scientists! Why do you assume scientists are somehow "above" explanations, and are content in ignorance?

"explanation _is_ the maths." -- People had good rules of thumb for evolution (such as dog breeding practices), and Gregor Mendel could have come up with his mathematical rules for selection before Darwin published his explanatory theory of natural selection. That doesn't mean that it's worthless to try to aim for an explanatory theory of QM instead of a purely numerical / observational one.

Confounding variables are a strong argument against this. Give me two sets of data that have no causal relationship, and it's very possible (if not likely) I will be able to demonstrate not only a causal relationship but also observational predictability through the existence of confounding variables. Let me use ML models and this becomes trivial.

Do school names determine how good a school will be? Of course not. However, schools have names which tend to be reused: cultural icons, scientists, etc.. And schools in low performing areas are going to have a different distribution of chosen names than schools in high performing areas. So you can create a model that will not only demonstrate this but also offer legitimate observational predictability that a school's name will, on average, determine how well it will do.

The only way you could really refute this hypothesis is by understanding the true confounding variable, or by running some huge scale multiple decades long experiment where you changed the names of new schools and measured the performance difference. That experiment is clearly not really practicable, so all you're left with is understanding.

And this is really where the risk is. If we ever begin to build upon something that is unsound then it risks everything from that point onward being also invalid. At best everything collapses and you realize you must have made a mistake. At worse, you simply end up adding endless epicycles until at some point everything starts to completely stall out muddled in its own invalid fundamentals and irreconcilable complexity.

I'm trying to remember which textbook or professor was the one who said that "shut up and calculate" was the Feynman interpretation of QM.

Feynman must have driven his wife mad.

  "Honey did you take the garbage out?"
  "When you say garbage, have you considered if garbage can even really exist?"

In analogy to this discussion, it'd be the opposite, actually -- situational Feynman would be the person who knew he could practically gather specific unwanted matter into a bin or sack and take it out to the collection bin for most practical volumes, while situational wife would be the person asking for an explanation "but how do you know that what you have taken out is the garbage and that it has in fact been removed"?

Still, it's fun to play with this idea via figures who are a little more philosophically rarified as it were:

https://www.tiktok.com/@davecolumbo/video/706085573209017886...

Come on. We have plenty of good heuristics for why things are the way they are. Quantum theory is no more a bunch of "guessworks and heuristic" than Newtonian mechanics.

Also, physicists disagreeing on interpretational issues related to quantum foundations is not the same thing as disagreeing about the fundamentals of the theory.

The things that Schrodinger tried are still valid physics, just for different systems. He didn't know that, but now we do, this has little bearing on modern physics.

Also chemists have the luxury of agreeing because the fundamentals of their work (not the day to day act of doing chemistry) gets to assume the presence of physics i.e. it's stampkeeping (depending on who you ask).

Not quite stamp collecting, although botany or anatomy head in that direction. Call chemistry “just applied physics” as engineering is “just applied science”.

It's a quote ascribed to Rutherford ("all science is physics or just stamp collecting"). I also heard it when studying protein structures in the 90s (the idea being, if you're just a crystallographer cranking out more structures, you're doing stamp collecting; the real reason to determine structures is to elucidate the general principles of protein folding and function).

Note that protein structure prediction finally fell to machine learning and it was "mostly stamp collecting" (IE, the accurate predictions come from subtle analysis of rich protein sequence alignments, not from understanding the fundamental principles of protein folding).

We absolutely understand protein folding. The fact there isn't a closed form simple solution to tell us the shape of a protein shouldn't be surprising at all. It's just hard to calculate. It's like the 3 body problem, but there are 90. I don't know why you think there would be some simple solution to that that doesn't involve simulation and heavy computation. You'd need 8 dimensional 100x100 matrices or something to describe something like a protein in pure math, which is way more like programming than arithmetic anyway.

We do not understand protein folding. I'm an expert in this field with publications and what you just said makes no sense at all.

The simplified protein folding problem states that proteins fold to their energy minimum, and a succesful heuristic can find the "correct" fold by finding the structure argmin(energy). Although this is a real simplification that doesn't represent actual proteins, even that is not something you'd solve with matrices; combinatorial math.

More importantly, "protein folding" is the biophysical process that proteins experience; what DeepMind did was solve "structure prediction", which is another simpler problem, and they didn't do it by energy minimization, they did it by exploiting sequence similarity to provide structural constraints that massively reduced the search space.

I used matrix math as a hand wavey example of how one might purely numerically describe the angles in a structure. What don't we understand? We know the forces of all the atoms that act on each other, what is the mystery at this point other than calculating it is hard?

"We know the forces of all the atoms that act on each other, what is the mystery at this point other than calculating it is hard?"

might be true, but it's computationally intractable and even if you did find a way around it you'd just learn that your force field was inaccurate, or that proteins don't actually fold to their energy minimum.

Not just hard, but NP-hard. In fact, the "protein folding problem is NP hard" is misleading; in fact, the "static protein structure problem" is NP hard, and protein folding is "harder" than that (it's a superproblem of static structure prediction).

The way to describe angles for a protein is this: you use degrees or radians to describe the torsion angles of the backbone (2 torsions per amino acid, 360 possible positions per torsion, times the number of backbone atoms in the protein). Treat side chains with rotamers, that's already a solved problem. However, you can't compute self-collisions in angle space you need to embed the protein structure in cartesian coordinates, which is how they are normally represented (as a graph representing the bond topology, and an N*3 array of positions).

Right, my point is that it's as solved as it can really be as far as human understanding is concerned. There are too many variables at play to have any sort of elegant solution like Ohm's law or calculating an orbit. The best we can do is find different/more efficient/novel methods to calculate it. Am I wrong?

The problem is ill-posed, not that we lack a solution.

Can you elaborate on why it's ill-posed?

When you say:

> it's as solved as it can really be as far as human understanding is concerned

You handwave away several interesting parts of the problem without providing real justification for why that would be necessary or acceptable.

It's like saying "politicians are made up of atoms, and since we know how atoms interact, our understanding of politics is reasonably complete".

proteins don't fold to their energy minimum, they typically "collapse" to an intermediate state and then sample many different states (kinetically), rather than adopting the absolute thermodynamic minimum. In many proteins, the structure snapshot doesn't even correspond to a functionally active protein. The "protein folding problem is NP-hard is an entirely inaccurate view of the biophysics of folding", and if solved, would not address any useful question about folding.

What DeepMind solved was a far simpler problem: reproduce the structures that get produced by some experimental method. That's a well posed problem but ultimately a less interesting one, even if it's immediately "useful".

I prefer to call engineering "physics ruined by philistines" (sometimes at least, engineers seem to have a unique ability to take beautiful physics and notate it in the most cumbersome "safe" way possible)

If you asked chemists questions as "fundamental" as the ones posed to quantum physicists, they'd be exactly as lost because they'd be exactly the same questions. Shockingly, when you have a level of abstraction beneath you you get to have pure empirics, trusting that the foundations add up to get the right answer. If every field of knowledge that relies on empirical evidence rather than a grand unifying theory was stamp collecting, QM would be one of very few that wasn't.

To use the bug collecting analogy, I like the idea that by capturing and describing these particles perhaps we've killed them in the process.

The analogy I like to use is that high-energy particle physics is like trying to learn a game like chess or baseball by analysing the statistics of past games.

That's literally what goes on at CERN. They run effectively trillions of experiments, collect millions of them (after some filter), and then they draw histograms.

Unless you know nothing about mathematics, you know that histograms are inherently statistical beasts. It's meaningless to talk about the histograms of one experiment.

Could you learn to play, say, tennis by drawing histograms of player movements? Ball bounce locations? No? Why not!? It's tons and tons of data! Accurate, scientific data!

Quantum physicists would argue that only statistics exists. I like to point to the fact that one hydrogen atom can exist, and it can have on electron. Or zero electrons. In nice countable, integral quantities. You can place a single gold atom on a crystal lattice with an AFM or even draw cute pictures with them: https://cen.acs.org/analytical-chemistry/imaging/30-years-mo...

Just because some people think only collecting bugs is what biology is all about, doesn't mean that there isn't more to it...

sounds like how ML/neuron network works today, computer comes up a super complex math equation that no human could understand to fit the data fed

Reply on news.ycombinator.com