Ask HN: Is physics moving forward?
Looking back over last 20 years, technology (esp. related to computers) had made extraordinary leaps forward and the pace is accelerating.
As a total layman in the physics world, it appears (looking from the benches) that things are crawling at the same speed they did 20 years ago.
Is that true ? Is the speed of physics advances accelerating as well ? What has been happening that we might of been missing ?
EDIT: Especially in the applied & practical worlds
74 comments
[ 4.9 ms ] story [ 169 ms ] threadMaybe you mean more theoretical physics? I don't really have enough knowledge to be useful there.
The 30-40 years have seen tech advancements roughly in line with Moore's law, which is fantastic. But that's the result of engineering advancements. There's no precedent I'm aware of to expect the same results out of the hard sciences.
The trouble in Physics is that the new tools look like the LHC.
The delay between a discovery in physics and a Nobel prize is getting bigger and bigger [1]. It's true for all fields, but the effect is particularly strong for physics.
'''It is safe to say that late 1920s and early 1930s were the “Golden Age” of 20th century physics, when the progress was lightning-fast and new discoveries lay like low-hanging fruits. In the 1940s Dirac commented bitterly, in view of problems quantum field theory was having at the time: “Then, a second-rate physicist could do first-rate work – now, it takes a first-rate physicist to do second-rate work”. Every physicist would love to live in such “interesting times”, when a new unexplored scientific territory opens up.''' [2]
So, its started decreasing quite some time. Also, as there are more an more people involved, the brainmass gets diluted. As of now there is no chance to get a photo of a concentration of luminaries as from the famous Solvay conferences [3].
So, if you (like me) read Feynman lectures of physics and know physics from 1920s-1950s, you are likely to get disappointed by the current pace.
...and just compare to recent progress in machine learning, when we can play on our computers with things, which a few years ago were thought to be out of our reach.
(But it shouldn't be surprising, technologies and science do have their growth time, and usually it's finite time.)
[1] http://priceonomics.com/why-nobel-winning-scientists-are-get...
[2] https://woodtickquarterly.wordpress.com/2011/11/17/graham-fa... (BTW: I recommend this book a lot)
[3] https://en.wikipedia.org/wiki/Solvay_Conference
The issue is that what appears simple, 100 years from now, takes a significant shift in how we're currently thinking. The problems are complex and the prior knowledge needed to get to a breakthrough is always growing, but I really do believe its self defeating to look at 100 years ago and think they had it easy as I believe people will undoubtedly say the same thing about us 100 years from now.
Economics is a good example as well I believe... Up until the second half of the 20th century many economists were not basing their theories on data but rather under assumptions that we are all "rational" actors. I will concede that the 'data' was not really available until the last 50yrs anyway but the 'behavioral' movement has hopefully set Economics as a whole on a 'straight' path and no doubt the next 100yrs will show the low hanging fruit which now dangle above us.
How poetic.
EDIT - The economics example I gave was very sweeping. Just an example.
The problem is that success has to be so compelling that people who are working within different paradigms (economics have many) generally figure out that they are thinking about it wrong and need to restructure everything in terms of behavioral economics, rather than looking at behavioral economics as a low order correction term to their own way of understanding them.
Richard Thaler puts it much better than me when he roughly says that the best economics work in the last decade is by classic economists doing good empirical work. Which, at least in my understanding, is the real 'gift' of the 'behavioral' approach.
But in psychology or economics as soon as you learn something new, that new knowledge is going to be a new variable to account for, for example as soon as you are made aware of the Bystander effect[1] it is trivial to avoid it, but you can't really predict what will come out of it. You can as well find ways to game any economic prediction by being aware of it, I guess you can find stable models that sort of work until something like the 2008 financial crisis happens and they don't anymore.
So either the investigators stay out of the economy/society or they will mess with their own results by making them public.
[1] https://en.wikipedia.org/wiki/Bystander_effect
My reasoning on this is that since everyone has a finite lifespan, understanding/work will be lost as the researcher retire, move to a different field, or die, unless this information is transmitted to someone else during their life time. However, as we understand more, we have to teach people more. This brings in two problems that I see:
- It takes longer for people to learn up to the frontier of science, which means they might be in their 30s, 40s, and that number may keep getting larger. - We might be able to teach these things in the same amount of time, but the understand of these topics by the younger generation might be less overall, which makes it difficult for them to make certain connections, as they might not have the deep enough understanding to see a connection.
My worry is the eventually we will reach a stage where the amount of knowledge we have is so much that no one can feasibly master a field. We can subdivide the fields, but no one will know enough general knowledge to make insightful connections that simplifies our models, or something to that effect.
As an example: my fluid dynamics professor said that the fluids mechanics I course we take in second year undergrad is what people worked on as their masters/phd thesis 100 years ago, whereas we don't even derive the formulas anymore now simply because we do not have the time to do so.
Similarly, it's considered basic undergrad work to understand Fourier decomposition; but the notations and clarity we get the insight from is the result of two centuries from the 1807 paper by Fourier. It was certainly PhD level at the time.
Certainly science is compacted and that's how it gets transmitted to younger generations. And remember that our brains get bigger, so maybe we'll never hit any "limit" to understanding? :)
Since we are teaching people less in depth stuff as they need to go very far into science, it seems like building knowledge with unstable foundation. The lack of branching out may also inhibit discovering certain connections between fields that may proved to be enlightening.
I've heard it said that string theory could never be scientifically tested in a traditional sense. People sometimes compare it to a religious philosophy because of this quality of untestability.
Maybe the logic of the universe at the lowest levels are hidden from us. If the universe is a computer simulation then we almost certainly are not exposed to the underlying mechanisms of how it works at the lowest level.
The fluid in a fluid simulation doesn't have access to the knowledge of how the computer on which it is running functions. The fluid is just information. That information is manipulated by the running program, but the code of the simulation is not built into the information representing the fluid.
Maybe the matter in our universe exists in the same way, as information. Maybe the mechanics that drive the motion of the smallest particles are not exposed.
I'm sure I'm not the only person to postulate on this. People have probably written very scientific papers on this very topic. I honestly don't know, but it's what my programmer brain leads me to believe.
I'm also reminded of Dr. Nick Bostrum's writings on the topic of simulated existence.
(1) It takes much less time to teach something, than was spent to figure things out the first time.
(2) When teaching improves, the essential points of a topic can be taught in less time, then what was spent when the topic was new. Then a topic is new, usually it is first taught following the lines and chronology on how it was discovered. Later improved teaching can figure out ways to get free of the chronology of the original story, and just teach the plain fact.
While Darwin's finches are still sometimes mentioned when teaching evolution, the whole story of measuring all kinds of properties of those birds in different islands, and presenting this data, and what can we deduct from it, is rarely presented. We just jump to the main point: species adapt and evolve.
Arguable, teaching physics still almost everyplace follows the historical chronology: mechanics, electromagnetism, quantum mechanics, which usually pushes quantum mechanics to the 3rd year in a university. The mathematical machinery of quantum mechanics however shouldn't take two full years to master, so maybe there could be a route to proceed quantum mechanics faster, not trying to cover all of classical physics first.
(3) I assume the success of future researchers is party build upon that they had access to good teachers, or good textbooks, which made it possible for them to quickly absorb the present state of knowledge. But there are voices complaining that producing good teaching material is not incentivized in the present academic system.
One of the biggest problem is fragmentation. People used to be "philosophers of the natural science", now they are experts in "implementations of quantum algorithms with cold ions". So its unlikely to get so many good people in one place (as the number of fields is astonishingly high) and to get progress between/outside of fields (many fields are a bit historical/arbitrary) - as many scientist (even eminent) lack of even basis knowledge in other sciences (or even - subfields). So things requiring bigger picture may be already out of scope.
When it comes to switching fields or dying - did you read: http://blog.computationalcomplexity.org/2015/07/will-our-und... ?
Largely hardware advances that are heavily interrelated with physics. Even still, it has been incremental (though still impressive) production advances and not radical architectural redesign. Intel only got bounds checking in hardware (MPX extensions) a couple of years ago, even though this was first done over half a century ago.
Software hasn't.
"New physics", physics that we don't already have an explanation for, is becoming increasingly rare. Our definition of "new physics" is also expanding to including things that aren't fundamentally new, just small holes in our understanding of the equations.
The flip side of this is that while we already know the broad strokes, there is still a lot of work to be done in filling in the details. This work is just less glamorous and doesn't make the headlines.
We don't understand why galaxies rotate the way they do, as the visible mass does not correspond to the rotation according to known laws of gravity. The solution is to postulate dark matter and dark energy, until things match again.
If you believe Lee Smolin then advocates of String Theory have also had an oppressive effect on opposing ideas, making it hard to get tenure if you're not working on it, effectively choking off competing theories (e.g. Doubly Special Relativity, Loop Quantum Gravity, etc)
I have no data or first-hand evidence, as I was only a mere Physics undergrad, but found his arguments pursuasive. String Theory (and his oscillating pals) have certainly been dominant in the scientific press, and don't seem to have come up with much. Could be a false negative though. We'll only know when someone opens the next big door.
Other fields of physics have had much more interesting discoveries though (e.g. graphene).
Government spending in the sciences has been extremely schizophrenic over the last 5 years. The NSF is a little more stable, but the NSF doesn't fund much of anything over 100M, which would be a relatively small experiment when spread out over 4-5 years.
If you look on log plots of parametrized experimental progress, progress remains linear, so it's a Moore's law like improvement on many fronts.
The emergence of precision cosmology has really transformed astrophysics in the last twenty years. The solar neutrino problem is now solved entirely (and even \theta_{13} has been measured!). The lynchpin of the Standard Model (the Higgs) has been found. LIGO is likely to make first detection in the next couple of years. Graphene and topological insulators have the solid-state community buzzing. Fluorescence microscopy and nanopore techniques are making waves in the biophysics community. And more, of course. Heck, this week, the most compelling evidence yet for the long-sought pentaquark appeared.
For the probes of the dark sector and of gravity, though we haven't found anything, huge swaths of parameter space (i.e. possible theories) have been ruled out. EDM searches are relentless in their searches for new physics. Someday, someone will find a reliable anomalous signal, but we can't predict when.
And, if you're looking for fun hints of new physics, check out "muon g-2". The next 5-10 years will be exciting there, too, to see if the existing discrepancy between measurement and the Standard Model will survive closer experimental scrutiny.
We go slow because we will go far.
http://www.int.washington.edu/talks/WorkShops/int_08_3/Peopl...
Maldecena and Susskind are both well regarded and accomplished.
The Standard Model (including the yet-to-be discovered Higgs and top quark) was completed in the mid '70s around when the bottom and charm quark were found. The pace of discovery of new fundamental particles has clearly slowed dramatically; the top quark was in 1994 and Higgs was just a couple of years ago, with serious pessimism that anything else will come out of the LHC (even with the luminosity upgrades). The existence of an LHC successor is uncertain, and even if it is built it is unlikely to find anything new.
The solar neutrino, compared to just about any other fundamental physics question in the 20th century, is boring. It was solved by adding a few additional free parameters to the SM with little conceptual insight and few implications for further post-SM physics. As you and I commented elsewhere, pentaquarks are a neat window into low-energy QCD, but they aren't new fundamental physics.
Precision cosmology's biggest result was to confirm the ad hoc and simplest inflationary models from the 1980's, with little hope in the near-term future to do anything besides rule out poorly motivated variations designed explicitly to be testable. What is the inflaton like, and what does it do besides adjust a couple of parameters in the CMB? No one knows, and know one has a serious expectation of finding out anytime soon.
None of this, obviously, means that these meager accomplishments weren't hard or that the people who made them aren't extremely clever. But they are undeniably less exciting then what was see before I was born in 1985, and pale in comparison to the absolute revolutions of the first half of the 20th century (quantum mechanics, relativity, GR, field theory).
I really want to call this out:
> huge swaths of parameter space (i.e. possible theories) have been ruled out.
This is the sort of comment that only us physicists can say with a straight face. Parameter space is bounded only by one's cleverness and stamina. Ruling it out says very little about reality because most proposals have very little a priori probability. The entire industry of constructing complicated models and then ruling them out is the lamest kind of progress I can imagine. Perhaps the open questions have really just gotten so hard that this is the best that can be hoped for. But we should be honest that this is essentially like hitting a brick wall and digging into it with spoons.
From the particle physics end, the revolution stopped because accelerator technology stalled out. If the laser-wakefield (or similar) accelerators can pan out, there may be a resurgence of the 1950's era of progress.
Just because the Standard Model happens to have panned out doesn't mean that its experimental verification wasn't a triumph. The Standard Model didn't have to be true!
The solar neutrino thing is phenomenal, at least to me, we surely see CKM mixing in the quark sector, but to see leptons change flavor, and do so on such a grand scale is astounding. When the problem first came up, flavor violation wasn't even on the table.
Precision cosmology has really laid out that there's something really, really big about Physics that we're missing. Really big. And it says it with such statistical and systematic vehemence that even the most curmudgeonly skeptic (myself included) is forced to confront it. In combination with the Bullet Cluster, there's almost no escape from the conclusion that there's something really important to be found.
Totally agreed that parameter space, without qualifiers, is a vague term.
Some parameter spaces are more important than others. I'd argue that the absence of new signals at LHC, the absence of bright new lines in Fermi/GLAST data, the absence of new lines in ultra-high energy cosmic rays, the absence of any signal in any EDM experiment, the absence of any equivalence principle violation or anything else in the gravitational sector, the absence of any experimental signal for any dark matter candidate, the continued expected behavior of the Hulse-Taylor pulsar, etc. are all important, especially in light of the problems of CP-violation, the complete incompatibility of the Standard Model with General Relativity, and the extremely good agreement between observed cosmology/BBN with the \LambdaCDM model.
A lot of really good people have pushed really hard experimentally on entirely reasonable theories, and those ideas have been severely constrained. In my own field, precision tests of gravity, theory is forced to contort itself to avoid experimental constraint. LHC is putting the screws to supersymmetry; there's still plenty of room there, but there's a lot less than there was a few years ago.
As experimentalists, we've got little patience for contrived models of parameter space. As each experiment takes 5-10 years, we only get a few in our lifetime, so we try to make our work as meaningfully-impactful as we can.
To give an example of why "pouring money into it and getting slower and slower progress" could translate into linear progress on a log plot, imagine you have to pour in x² amount of (money, people, time) to get to level x of progress.
In that case "we go slow" means, "we'll gobble up larger and larger amounts of money, but everytime people get impatient we'll have some bone to throw at them".
Understanding things is certainly great, but you can't even claim you understand something new if you can not use that new understanding to achieve something you couldn't before it.
Or, more clearly, pure theory is good for mathematicians, but scientists can not even claim a theory is new if it has no application.
It's certainly not what I'm expecting. Even if it hadn't gone on to be central to our engineering, I would have liked to have known about general relativity or QCD.
Understanding things is certainly great, but you can't even claim you understand something new if you can not use that new understanding to achieve something you couldn't before it.
What do you mean by understand something new? This week we started to understand something new - understand many things new - about Pluto. Are you trying to claim that this isn't true unless we can use it somehow? What is it you're trying to say?
Or, more clearly, pure theory is good for mathematicians, but scientists can not even claim a theory is new if it has no application.
Is astronomy not a science? Or palaeontology? Or archaeology? Those are all observation-oriented sciences that don't conduct experiments but do gather information and develop knowledge.
I have no idea what it is you're trying to say about claim a theory is new - do you want to rephrase that? What would you say about the recent discover of pentaquarks? Pentaquarks are useless, but our knowledge of them is quite new.
The job market for physics has not changed substantially in the last twenty years.
By 'we go slow because we will go far', I meant only to emphasize that regular incremental improvements generally trump blitzkrieg efforts.
[1] https://en.wikipedia.org/wiki/Timeline_of_quantum_computing
In fundamental physics, the LHC is online, neutrino physics is hot, and lots is going on. Now there are a huge number of quantum theories of black holes, but no way to prove anything about them in site.
The dark matter problem is a huge "anomaly" left to solve so there are still mountains to climb.
In terms of practical stuff there is lots of physics in how you build a 7nm microchip. Physicists collaborate a lot with "nanotechnology" people and biologists. For instance my thesis advisor worked with experimentalists who were stretching DNA with tweezers and figured out how the AIDS virus self-assembles.
Even the "dead" area of chaos theory is looking much better now that people at NASA have made a map of the earth-moon phase space which can give a km/sec or so free propulsion.
Regarding now vs. the 80s, we have much better tools for making assurance cases for critical software. So much progress has been made in both specification correctness and implementation correctness that I think the only way you could compare the two and say there hasn't been improvement is if you haven't tried to see what we can actually do now vs. then...
https://en.wikipedia.org/wiki/The_Trouble_with_Physics
Smolin makes a sosiological-historical argument, that from the 20s to 70s, the development in theoretical particle physics was done by trusting the "mathematical intuition", if the math was beautiful and predicted the existence of some particles, a bit later usually the experimental physicists found those particles. And the rewards and Nobel prizes went to those who did the math the fastest.
So the whole theoretical physics adopted this style of work, when someone proposes something new that looks interesting, everyone tries to do the math as quickly as possible, to be the first to get the results.
But then from the 70s to 00s, this flocking attitude was applied to string theory, and people just developed string math furiously, and it was left unnoticed that the theory was totally unrelated to any experiments.
So, Smolin suggests, for 30 years the best of theoretical physics went into a direction that may be totally separated from experiments. If this turns out to be true, theoretical physics pretty much lost 30 years.
You can't predict breakthroughs. A scientist might spend their entire career following their research to a dead end. That's not a failure, in my opinion. It just shows how much we already know, and I think we should honour those scientists just as much as the lucky ones.
Putting the measurement problems aside, progress in physics seems to be a lot less smooth and the big jump occurred in the first three decades of the last century with the discovery of special relativity and quantum mechanics plus the ongoing project of formalizing physics. That one was a complete paradigm shift towards mathematical models and towards an entirely different picture of reality. Since then the development of quantum field theories is basically just using the same trick as for quantum mechanics. ( Not trying to belittle the development of QFT, that is one of the monumental achievements of the human mind, it just pales in comparison to the development of QM. ) So, in this view the next big jump may be just around the corner or not possible for a human brain, but we will only know the answer after progress happened. ( I should cite one of the famous philosophers of science here, unfortunately I forgot which one.)
As an example, string theory is currently not even wrong, because we can not build the known experiments that would enable us to test string theory. However a lot of brain power and ink was expended on its development over the last thirty years, and we simply do not have a good idea if it was worthwhile. If someone suggests a experiment that can distinguish between string theory and other models of quantum gravity, and if a string theory passes this test, then it was probably worthwhile to spend all that effort.
In conclusion, I would argue that the question is ill defined and runs furthermore in epistemological problems, that is even if we would find a good definition we can not really know the answer. However, I am actually quite optimistic that a breakthrough is just around the corner. For example, I think that the connection of information theory and physics is not really understood, but concepts like entropy and information seem to crop up everywhere one looks.
Up till about the early 20th century, ground breaking physics was largely "two guys in a garage" territory -- single individuals such as Newton or Cavendish tinkering in their own private laboratories using fairly modest equipment. Teenagers replicate their experiments in school physics lessons with equipment costing no more than a few hundred pounds today.
Throughout the mid-twentieth century, ground-breaking discoveries were increasingly made by teams of researchers, which seem to have grown larger over time, with equipment that has become increasingly large and expensive, and sponsored by universities, companies and governments.
Nowadays it seems that most ground-breaking discoveries are made by large, national or multinational teams working with equipment costing billions of dollars and processing petabytes of data. I couldn't see two guys in a garage producing their own space telescope or particle accelerator any time soon.
And believe me, there are a lot of experimental physicists doing excellent work with two guys in a garage levels of equipment.