[HN Gopher] Fitting an elephant with four non-zero parameters
       ___________________________________________________________________
        
       Fitting an elephant with four non-zero parameters
        
       Author : belter
       Score  : 294 points
       Date   : 2024-07-14 14:27 UTC (1 days ago)
        
 (HTM) web link (arxiv.org)
 (TXT) w3m dump (arxiv.org)
        
       | EdwardCoffin wrote:
       | Freeman Dyson recounts the episode [1] that inspired this paper
       | in his web of life interviews (prepositioned to the fitting an
       | elephant bit) [2]
       | 
       | [1] https://youtu.be/hV41QEKiMlM
       | 
       | [2] https://youtu.be/hV41QEKiMlM?t=118
        
         | parker-3461 wrote:
         | Thanks for linking these, I was not very familiar of these
         | works/discussions taking place in the past, but these really
         | helped establish the context. Very grateful that these videos
         | are readily available.
        
           | EdwardCoffin wrote:
           | I listened to the whole series with Dyson some time in the
           | past year. It was well worth it. I also listened to the
           | series with Murray Gell-Mann [1] and Hans Bethe [2]. All time
           | well worth spending, and I've been thinking of downloading
           | all the bits, concatenating them into audio files, and
           | putting them on my phone for listening to when out on walks
           | (I'm pretty sure the videos do not add anything essential:
           | it's just a video of the interviewee talking - no visual
           | aids).
           | 
           | [1] https://www.youtube.com/playlist?list=PLVV0r6CmEsFxKFx-0l
           | sQD...
           | 
           | [2] https://www.youtube.com/watch?v=LvgLyzTEmJk&list=PLVV0r6C
           | mEs...
        
       | lazamar wrote:
       | Lol. Loved it.
       | 
       | This was a lovely passage from Dyson's Web of Stories interview,
       | and it struck a chord with me, like it clearly did with the
       | authors too.
       | 
       | It happened when Dyson took the preliminary results of his work
       | on the Pseudoscalar theory of Pions to Fermi and Fermi very
       | quickly dismissed the whole thing. It was a shock to Dyson but
       | freed him from wasting more time on it.
       | 
       | Fermi: When one does a theoretical calculation, either you have a
       | clear physical module in mind or a rigorous mathematical basis.
       | You have neither. How many free parameters did you use for your
       | fitting?
       | 
       | Dyson: 4
       | 
       | Fermi: You know, Johnny Von Neumann always used to say 'with four
       | parameters I can fit an elephant; and with five I can make him
       | wiggle his trunk'.
        
       | dheera wrote:
       | I wish there was more humor on arXiv.
       | 
       | If I could make a discovery in my own time without using company
       | resources I would absolutely publish it in the most humorous way
       | possible.
        
         | btown wrote:
         | There's plenty of humor on arXiv, and that's part of why it's
         | so incredible!
         | 
         | Some lists:
         | 
         | https://academia.stackexchange.com/questions/86346/is-it-ok-...
         | 
         | https://www.ellipsix.net/arxiv-joke-papers.html
        
           | azeemba wrote:
           | Consider posting this as a new post! It seems like a fun list
           | to read through
        
           | mananaysiempre wrote:
           | Joke titles and/or author lists are also quite popular, e.g.
           | the Greenberg, Greenberger, Greenbergest paper[1], a paper
           | with a cat coauthor whose title I can't seem to recall (but
           | I'm sure there's more than one I've encountered), or even the
           | venerable, unfortunate in its joke but foundational in its
           | substance Alpher, Bethe, Gamow paper[2]. Somewhat closer to
           | home, I think computer scientist Conor McBride[3] is the
           | champion of paper titles (entries include "Elimination with a
           | motive", "The gentle art of levitation", "I am not a number:
           | I am a free variable", "Clowns to the left of me, jokers to
           | the right", and "Doo bee doo bee doo") and sometimes code in
           | papers:                 letmeB this (F you) | you == me = B
           | this                           | otherwise = F you
           | letmeB this (B that)            = B that       letmeB this
           | (App fun arg)       = letmeB this fun `App` letmeB this arg
           | 
           | (Yes, this is working code; yes, it's crystal clear in the
           | context of the paper.)
           | 
           | [1] https://arxiv.org/abs/hep-ph/9306225
           | 
           | [2] https://en.wikipedia.org/wiki/Alpher%E2%80%93Bethe%E2%80%
           | 93G...
           | 
           | [3] http://strictlypositive.org/
        
             | gjm11 wrote:
             | > paper with a cat coauthor whose title I can't seem to
             | recall
             | 
             | You probably have in mind
             | https://en.wikipedia.org/wiki/F._D._C._Willard (coauthor of
             | multiple papers, sole author of at least one).
        
         | msp26 wrote:
         | Pretraining on the Test Set Is All You Need
         | 
         | https://arxiv.org/abs/2309.08632
        
         | paulpauper wrote:
         | There is . it is called the General Math section. what is more
         | funny than a 2 page proof of the Reiman Hypothesis?
        
       | boywitharupee wrote:
       | what's the purpose of this? is it one of those 'fun' problems to
       | solve?
        
         | jfoutz wrote:
         | This quote might help -
         | https://en.wikipedia.org/wiki/Von_Neumann%27s_elephant#Histo...
         | 
         | yes, a fun problem, but also a criticism of using to many
         | parameters.
        
       | Steuard wrote:
       | Sadly, the constant term (the average r_0) is never specified in
       | the paper (it seems to be something in the neighborhood of 180?):
       | getting that right is necessary to produce the image, and I can't
       | see any way _not_ to consider it a fifth necessary parameter. So
       | I don 't think they've genuinely accomplished their goal.
       | 
       | (Seriously, though, this was a lot of fun!)
        
         | rsfern wrote:
         | They say in the text that it's the average value of the data
         | points they fit to. I think whether to count it as a parameter
         | depends on whether you consider standardization to be part of
         | the model or not
        
           | Steuard wrote:
           | I see your point, that it's really just an overall
           | normalization for the size rather than anything to do with
           | the _shape_. I can accept that, and I 'll grant them the
           | "four non-zero parameters" claim.
           | 
           | Though in that case, I would have liked for them to make it
           | explicit. Maybe normalize it to "1", and scale the other
           | parameters appropriately. (Because as it stands, I don't
           | think you can reproduce their figure from their paper.)
        
       | lupire wrote:
       | IIUC:
       | 
       | A real-parameter (r(theta) = sum(r_k cos(k theta))) Fourier
       | series can only draw a "wiggly circle" figure with one point on
       | each radial ray from the origin.
       | 
       | A compex parameter (z(theta) = sum(e^(z_ theta))) can draw more
       | squiggly figures (epicycles) -- the pen can backtrack as the
       | drawing arm rotates, as each parameter can move a point somewhere
       | on a small circle around the point computed from the previous
       | parameter (and recursively).
       | 
       | Obligatory 3B1B https://m.youtube.com/watch?v=r6sGWTCMz2k
       | 
       | Since a complex parameter is 2 real parameters, we should compare
       | the best 4-cosine curve to the best 2-complex-exponential curve.
        
       | elijahbenizzy wrote:
       | This is humorous (and well-written), but I think its more than
       | that.
       | 
       | I'm always making the joke (observation) that ML (AI) is just
       | curve-fitting. Whether "just curve-fitting" is enough to produce
       | something "intelligent" is, IMO, currently unanswered, largely
       | due to differing viewpoints on the meaning of "intelligent".
       | 
       | In this case they're demonstrating some very clean, easy-to-
       | understand curve-fitting, but it's really the same process --
       | come up with a target, optimize over a loss function, and hope
       | that it generalizes, (this one, obviously, does not. But the
       | elephant is cute.)
       | 
       | This raises the question Neumann was asking -- why have so many
       | parameters? Ironically (or maybe just interestingly), we've done
       | a _lot_ with a ton of parameters recently, answering it with
       | "well, with a lot of parameters you can do cool things".
        
         | visarga wrote:
         | > Whether "just curve fitting" is enough to produce something
         | "intelligent" is, IMO, currently unanswered
         | 
         | Continual "curve fitting" to the real world can create
         | intelligence. What is missing is not something inside the
         | model. It's missing a mechanism to explore, search and expand
         | its experience.
         | 
         | Our current crop of LLMs ride on human experience, they have
         | not largely participated in creating their own experiences.
         | That's why people call it imitation learning or parroting. But
         | once models become more agentic they can start creating useful
         | experiences on their own. AlphaZero did it.
        
           | soist wrote:
           | AlphaZero did not create any experiences. AlphaZero was
           | software written by people to play board games and that's all
           | it ever did.
        
             | visarga wrote:
             | AZ trained in self-play mode for millions of games, over
             | multiple generations of a player pool.
        
               | soist wrote:
               | I am familiar with the literature on reinforcement
               | learning.
        
               | pharrington wrote:
               | They're saying the board games AlphaZero played with
               | itself _are_ experiences.
        
               | soist wrote:
               | And I am saying they are confused because they are
               | attributing personal characteristics to computers and
               | software. By spelling out what computers are doing it
               | becomes very obvious that there is nothing that can be
               | aware of any experiences in computers as it is all simply
               | a sequence of arithmetic operations. If you can explain
               | which sequence of arithmetic operations corresponds to
               | "experiences" in computers then you might be less
               | confused than all the people who keep claiming computers
               | can think and feel.
        
               | nyssos wrote:
               | > By spelling out what computers are doing it becomes
               | very obvious that there is nothing that can be aware of
               | any experiences in computers as it is all simply a
               | sequence of arithmetic operations.
               | 
               | By spelling out what brains are doing it becomes very
               | obvious that it's all simply a sequence of chemical
               | reactions - and yet here we are, having experiences.
               | Software will never have a _human_ experience - but
               | neither will a chimp, or an octopus, or a Zeta-Reticulan.
               | 
               | Mammalian neurons are not the only possible substrate for
               | intelligence; if they're the only possible substrate for
               | consciousness, then the fact that we're conscious is an
               | inexplicable miracle.
        
               | soist wrote:
               | This is a common retort. You can read my other comments
               | if you want to understand why you're not really
               | addressing my points because I have already addressed how
               | reductionism does not apply to living organisms but it
               | does apply to computers.
        
               | Dylan16807 wrote:
               | The comments where you demand an instruction set for the
               | brain, or else you'll dismiss any argument saying its
               | actions can be computed? Even after people explained that
               | lots of computers don't even have instruction sets?
               | 
               | And where you decide to assume that non-computable
               | physics happens in the brain based on no evidence?
               | 
               | What a waste of time. You "addressed" it in a completely
               | meaningless way.
        
               | godelski wrote:
               | If an algorithmic process is an experience and a
               | collection of experiences is intelligence then we get
               | some pretty wild conclusions that I don't think most
               | people would be attempting to claim as it'd make them
               | sound like a lunatic (or a hippy).
               | 
               | Consider the (algorithmic) mechanical process of screwing
               | in a screw into a board. This screw has an "experience"
               | and therefore intelligence. So... The screw is
               | intelligent? Very low intelligence, but intelligent
               | according to this definition.
               | 
               | But we have an even bigger problem. There's the metaset
               | of experiences, that's the collection of several screws
               | (or the screw, board, and screwdriver together). So we
               | now have a meta intelligence! And we have several because
               | there's the different operations on these sets to
               | perform.
               | 
               | You might be okay with this or maybe you're saying it
               | needs memory. If the later you hopefully quickly realize
               | this means a classic computer is intelligent but due to
               | the many ways information can be stored it does not solve
               | our above conundrum.
               | 
               | So we must then come to the conclusion that all things
               | AND any set of things have intelligence. Which kinda
               | makes the whole discussion meaningless. Or, we must need
               | a more refined definition of intelligence which more
               | closely reflects what people actually are trying to
               | convey when they use this word.
        
               | nyssos wrote:
               | > If an algorithmic process is an experience and a
               | collection of experiences is intelligence
               | 
               | Neither, what I'm saying is that the observable
               | correlates of experience are the observable correlates of
               | intelligence - saying that "humans are X therefore humans
               | are Y, software is X but software is not Y" is special
               | pleading. The most defensible positions here are
               | illusionism about consciousness altogether (humans _aren
               | 't_ Y) or a sort of soft panpsychism (X really does imply
               | Y). Personally I favor the latter. Some sort of threshold
               | model where the lights turn on at a certain point seems
               | pretty sketchy to me, but I guess isn't ruled out. But
               | GP, as I understand them, is claiming that biology
               | doesn't even supervene on physics, which is a _wild_
               | claim.
               | 
               | > Or, we must need a more refined definition of
               | intelligence which more closely reflects what people
               | actually are trying to convey when they use this word.
               | 
               | Well that's the thing, I don't think people are trying to
               | convey any particular thing. I think they're trying to
               | find some line - any line - which allows them to write
               | off non-animal complex systems as philsophically
               | uninteresting. Same deal as people a hundred years ago
               | trying to find a way to strictly separate humans from
               | nonhuman animals.
        
               | gowld wrote:
               | Continuing this reductio ad abusrdum, you might reach the
               | fallactious conclusion, as some famous cranks in the past
               | did, that intelligence is even found in plants, animals,
               | women, and even the uncivilized savages of the new
               | continent.
               | 
               | Intelligence appears in gradients, not a simple binary.
        
             | rocqua wrote:
             | Are you launching into a semantic argument about the word
             | 'experience'? If so, it might help to state what essential
             | properties alphago was missing that makes it 'not having an
             | experience'.
             | 
             | Otherwise this can quickly devolve into the common useless
             | semantic discussion.
        
               | soist wrote:
               | Just making sure no one is confused by common
               | computationalist sophistry and how they attribute
               | personal characteristics to computers and software.
               | People can have and can create experiences, computers can
               | only execute their programmed instructions.
        
               | HeatrayEnjoyer wrote:
               | On what priors are you making that statement?
        
               | soist wrote:
               | Rephrase your question. I don't know what you're asking.
        
               | nertirs wrote:
               | I think he meant to ask, what is the difference between
               | an experience and a predefined instruction?
        
               | gowld wrote:
               | Username `soist is an abbreviation for `solipsist, then?
        
           | elijahbenizzy wrote:
           | There are a whole bunch of assumptions here. But sure, if you
           | view the world as a closed system, then you have a decision
           | as a function of inputs:
           | 
           | 1. The world around you 2. The experiences within your
           | (really, the past view of the world around you) 3. Innateness
           | of you (sure, this could be 2 but I think it's also something
           | else) 4. The experience you find + the way you change
           | yourself to impact (1), (2), and (3)
           | 
           | If you think of intelligence as all of these, then you're
           | making the assumption that all that's required for (2), (3),
           | and (4) is "agentic systems", which I think skips a few steps
           | (as the author of an agent framework myself...). All this is
           | to say that "what makes intelligence" is largely unsolved,
           | and nobody really knows, because we actually don't understand
           | this ourselves.
        
           | godelski wrote:
           | > Continual "curve fitting" to the real world can create
           | intelligence.
           | 
           | I'm going to need a citation on this bold claim. And by that
           | I mean in the same vein as what Carl Sagan would say
           | Extraordinary claims require extraordinary evidence
        
             | Grimblewald wrote:
             | I'd simply argue what we do is precisley that, so either we
             | are intelligent, or we are not, however we might define
             | that intelligence.
        
           | kgeist wrote:
           | >It's missing a mechanism to explore, search and expand its
           | experience.
           | 
           | Can't we create an agent system which can search the internet
           | and choose what data to train itself with?
        
             | Xcode23 wrote:
             | you need to define what the utility function of the agent
             | is so it can know what to actually use to train itself. If
             | we knew that this whole debate about human intelligence in
             | computers would either be solved already or well on its way
             | to being solved.
        
         | luplex wrote:
         | I mean the devil is in the details. In Reinforcement Learning,
         | the target moves! In deep learning, you often do things like
         | early stopping to prevent too much optimization.
        
           | soist wrote:
           | There is no such thing as too much optimization. Early
           | stopping is to prevent overfitting to the training set. It's
           | a trick just like most advances in deep learning because the
           | underlying mathematics is fundamentally not suited for
           | creating intelligent agents.
        
             | rocqua wrote:
             | Is over fitting different from 'too much optimization'?
             | Optimization still needs a value that is optimized. Over
             | fitting is the result of too much optimization for not
             | quite the right value (i.e. training error when you want to
             | reduce prediction error)
        
               | soist wrote:
               | What value is being optimized and how do you know it is
               | too much or not enough?
        
               | godelski wrote:
               | I think the miscommunication is due to the proxy nature
               | of our modeling. From one perspective, yes you're right
               | because it's just on your optimization function and
               | objectives. But if we're in the context where we
               | recognize the practical usage of our model replies on it
               | being an inexact representation (proxy) then certainly
               | there is too much optimization. I mean most of what we
               | try to model in ML is intractable.
               | 
               | In fact, that entire notion of early stopping is due to
               | this. We use a validation set as a pseudo test set to
               | inject information into our optimization products without
               | leaking information from the test set (why you shouldn't
               | choose parameters based on test results. That is
               | spoilage. Doesn't matter if it's status quo, it's
               | spoilage)
               | 
               | But we also need to consider that a lack of divergence
               | between train/val does not mean there isn't overfittng.
               | Divergence implies overfittng but the inverse statement
               | is not true. I state this because it's both relevant here
               | and an extremely common mistake.
        
               | soist wrote:
               | Most practitioners seem to understand that what they are
               | doing is creating executable models and they don't
               | confuse the model based on numeric observations with the
               | actual reality. This is why I very much do not like all
               | the AI hype and how statistical models were rebranded as
               | artificial "intelligence" because the people who are not
               | aware of what the words mean get very confused and start
               | thinking they are nothing more than computers executing
               | algorithms to fit numerical data to some unspecified
               | cognitive model.
        
               | godelski wrote:
               | > Most practitioners seem to understand that what they
               | are doing is creating executable models and they don't
               | confuse the model based on numeric observations with the
               | actual reality.
               | 
               | I think you're being too optimistic, and I'm a pretty
               | optimistic person. Maybe it is because I work in ML, but
               | I've had to explain to a large number of people this
               | concept. This doesn't matter if it is academia or
               | industry. It is true for both management and coworkers.
               | As far as I can tell, people seem very happy to operate
               | under the assumption that benchmark results are strong
               | indicators of real world performance __without__ the need
               | to consider assumptions of your metrics or data. I've
               | even proven this to a team at a trillion dollar company
               | where I showed a model with lower test set performance
               | had more than double the performance on actual customer
               | data. Response was "cool, but we're training a much
               | larger model on more data, so we're going to use that
               | because it is a bit better than yours." My point was that
               | the problem still exists in that bigger model with more
               | data, but that increased params and data do a better job
               | at hiding the underlying (and solvable!) issues.
               | 
               | In other words, in my experience people are happy to be
               | Freeman Dyson in the conversation Calavar linked[0] and
               | very upset to hear Fermi's critique: being able to fit
               | data doesn't mean shit without either a clear model or a
               | rigorous mathematical basis. Much of data science is
               | happy to just curve fit. But why shouldn't they? You
               | advance your career in the same way, by bureaucrats who
               | understand the context of metrics even less.
               | 
               | I've just experienced too many people who cannot
               | distinguish empirical results from causal models. And a
               | lot of people who passionately insist there is no
               | difference.
               | 
               | [0] https://news.ycombinator.com/item?id=40964328
        
         | maitola wrote:
         | In the case of AI, the more parameters, the better! In Physics
         | is the opposite.
        
           | elijahbenizzy wrote:
           | One of the hardest parts of training models is avoiding
           | overfitting, so "more parameters are better" should be more
           | like "more parameters are better given you're using those
           | parameters in the right way, which can get hard and
           | complicated".
           | 
           | Also LLMs just straight up _do_ overfit, which makes them
           | function as a database, but a really bad one. So while more
           | parameters might just be better, that feels like a cop-out to
           | the real problem. TBD what scaling issues we hit in the
           | future.
        
           | will1am wrote:
           | A dichotomy between these fields
        
         | will1am wrote:
         | Your humorous observation captures a fundamental truth to some
         | extent
        
       | maitola wrote:
       | I love the ironic side of the article. Perhaps they should add
       | the reason for it, from Fermi's and Neumann's. When you are
       | building a model of reality in Physics, If something doesn't fit
       | the experiments, you can't just add a parameter (or more) variate
       | it and fit the data. The model should have zero parameters,
       | ideally, or the least possible, or, even at a more deeper level,
       | the parameters should emerge naturally from some simple
       | assumptions. With 4 parameters you don't know whether you are
       | really capturing a true aspect of reality of just fitting the
       | data of some experiment.
        
         | jampekka wrote:
         | This was mentioned in the first paragraph of the paper. The
         | paper is mostly humoristic.
         | 
         | That said, the wisdom of the quip has been widely lost in many
         | fields. In many fields data is "modeled" with huge regression
         | models with dozens of parameters or even neural networks with
         | billions of parameters.
         | 
         | > In 1953, Enrico Fermi criticized Dyson's model by quoting
         | Johnny von Neumann: "With four parameters I can fit an
         | elephant, and with five I can make him wiggle his trunk."[1].
         | This quote is intended to tell Dyson that while his model may
         | appear complex and precise, merely increasing the number of
         | parameters to fit the data does not necessarily imply that the
         | model has real physical significance.
        
           | karmakaze wrote:
           | That's how I feel about dark matter. Oh this galaxy is slower
           | than this other similar one. The first one must have less
           | dark matter then.
           | 
           | What can't be fit by declaring the amount of dark matter that
           | _must be present_ fits the data? It 's unfalsifiable, just
           | because we haven't found it, doesn't mean it doesn't exist.
           | Even worse than string/M-theory which at least has math.
        
             | edflsafoiewq wrote:
             | It's easy to say "Epicycles! Epicycles!", but people are
             | going to continue using their epicycles until a Copernicus
             | comes along.
        
               | jampekka wrote:
               | There will be no Copernicus if everybody just studies
               | epicycles. E.g. there are massive resources put into the
               | desperate WIMP hunt that could be used for finding new
               | theories.
        
               | Dylan16807 wrote:
               | I don't see how those resources are fungible with each
               | other.
        
               | jampekka wrote:
               | Research funding is very competitive and scarce.
        
               | Dylan16807 wrote:
               | Building a physics machine, and thinking about how the
               | equations might work, are so very different.
               | 
               | Does the latter even _get_ funding?
        
               | jampekka wrote:
               | Of course. It's called theoretical physics.
        
               | fiddlerwoaroof wrote:
               | Well, the funny thing is Copernicus posits just about as
               | many epicycles in his theory as previous geocentric
               | theories. Only Kepler's discovery of the equal area law
               | and elliptical orbits successfully banishes epicycles.
        
               | aeneasmackenzie wrote:
               | The history of these discoveries is fascinating and shows
               | that Kuhn's scientific revolutions idea is wrong but it's
               | always rounded off to "Copernicus and Galileo" and
               | doesn't even get them right
        
             | rocqua wrote:
             | I used to think this, but dark matter does make useful
             | predictions, that are hard to explain otherwise.
             | 
             | This is partially because there are two ways to detect
             | dark-matter. The first is gravitational lensing. The second
             | is the rotatinal speed of galaxies. There are some galaxies
             | that need less Dark Matter to explain their rotational
             | speed. We can then cross check whether those galaxies cause
             | less gravitational lensing.
             | 
             | Besides that, the gravitational lensing of galaxies being
             | stronger than the bright matter in the galaxies can justify
             | is hard to explain without dark matter.
        
               | jampekka wrote:
               | The problem with dark matter is that there's no (working)
               | theory on how the dark matter is distributed. It's really
               | easy to "explain" gravitational effects if you can
               | postulate extra mass ad-hoc to fit the observations.
        
               | rocqua wrote:
               | If there are two different types of observations, and one
               | parameter can explain both, that is pretty strong
               | evidence. Put differently, dark matter is falsifyable,
               | and experiments have tried to falsify it without success.
               | 
               | Besides the idea 'not all mass can be seen optically' is
               | not that surprising. The many theories on what that mass
               | might be are all speculation, but they are treated as
               | such.
        
               | XorNot wrote:
               | It's worth noting that one dark matter explanation is
               | just: it's cold matter we just can't see through
               | telescopes. Or black holes without accretion disks.
               | 
               | Both of these are pretty much ruled out though: you can't
               | plausibly add enough brown dwarfs, and if it's black
               | holes then you should see more lensing events towards
               | nearby stars given how many you'd need.
               | 
               | But they're both concrete predictions which are
               | falsifiable (or boundable such that they can't be the
               | dominant contributors).
        
               | kaibee wrote:
               | I dunno if this is the correct way of thinking about it,
               | but I just imagine it as a particle that has mass but
               | does not interact with other particles (except at big-
               | bang like energy levels?). So essentially a galaxy would
               | be full of these particles zipping around never colliding
               | with anything. And over time, some/most of these
               | particles would have stable orbits (as the ones in
               | unstable orbits would have flown off by now) around the
               | galactic core. And to an observer, it would look like a
               | gravitational tractor ahead of the rest of the physical
               | mass of the galaxy (which is slower because it is
               | affected by things like friction and collisions?). And so
               | you'd see galaxies where the arms are spinning faster
               | than they should be?
        
               | nyssos wrote:
               | > I dunno if this is the correct way of thinking about
               | it, but I just imagine it as a particle that has mass but
               | does not interact with other particles (except at big-
               | bang like energy levels?).
               | 
               | Not even anything that extreme. What's ruled out is
               | interaction via electromagnetism (or if you want to get
               | really nit-picky, electromagnetic interaction with a
               | strength above some extremely low threshold).
        
             | GuB-42 wrote:
             | The dark matter theory is falsifiable. Sure we can't see
             | dark matter (it doesn't interact electromagnetically), but
             | we can see its effects, and it has to follow the laws of
             | physics as we understand them today.
             | 
             | It is actually a satisfying theory with regard to the Occam
             | razor. We don't have to change our laws of physics to
             | explain the abnormal rotations of galaxy, we just need
             | "stuff" that we can't see and yet interact gravitationally.
             | When we have stuff like neutrinos, it is not that far
             | fetched. In fact, though unlikely given our current
             | understanding of physics, dark matter could be neutrinos.
             | 
             | If, as it turn out, the invisible stuff we call dark matter
             | doesn't follow the laws of physics as we know them, then
             | the dark matter theory is falsified and we need a new one
             | (or at least some tweaks). And it may actually be the case
             | as a recent paper claims that gravitational lensing doesn't
             | match the predictions of the dark matter theory.
             | 
             | The main competitor to dark matter is modified gravity,
             | which calls for no new stuff, but changes the equations for
             | gravity. For the Occam razor, adding some random term to an
             | equation is not really better than adding some invisible
             | but well characterized stuff, especially when we consider
             | that the equation in question is extremely well tested. It
             | is, of course, also falsifiable.
             | 
             | The problem right now is not that these theories are
             | unfalsifiable, it is that they are already pretty much
             | falsified in their current form (dark matter less than
             | modified gravity), and some rework is needed.
        
               | nyssos wrote:
               | > Sure we can't see dark matter (it doesn't interact
               | electromagnetically), but we can see its effects
               | 
               | Even this is granting too much: "seeing it" and "seeing
               | its effects" are _the same thing_. No one has ever
               | "directly seen", in the sense that internet DM skepticism
               | demands, anything other than a photon.
        
               | adrian_b wrote:
               | "Seeing" is indeed a poorly chosen word.
               | 
               | The problem with dark matter is that there does not exist
               | any second relationship from which to verify its
               | existence, like in the case of normal matter, which takes
               | part in a variety of interactions that lead to measurable
               | effects, which can be compared.
               | 
               | The amount and the location of dark matter is computed
               | from the gravitational forces that explain the observed
               | movements of the bodies, but there are no additional
               | relationships with any other data, which could
               | corroborate the computed distribution of dark matter.
               | That is what some people mean by "seeing".
        
               | nyssos wrote:
               | All major DM candidates also have multiple interactions:
               | that's the WI in WIMP, for instance. In fact I don't know
               | that anyone is seriously proposing that dark matter is
               | just bare mass with no other properties - aside from the
               | practical problems, that would be a pretty radical
               | departure from the last century of particle physics.
        
               | jampekka wrote:
               | No interactions have been found, despite a lot of
               | resources put into the search. So currently all dark
               | matter particle theories apart from "non-interacting"
               | have been falsified. And non-interacting theories are
               | probably unfalsifiable.
               | 
               | Radical departure may well be needed, for other reasons
               | too.
        
               | karmakaze wrote:
               | > The problem with dark matter is that there does not
               | exist any second relationship from which to verify its
               | existence.
               | 
               | This is exactly it! Dark matter is strictly defined by
               | its effects. The only 'theory' part is a belief that it's
               | caused by yet to be found particle that's distributed to
               | fit observations. Take all the gravitational anomalies
               | that we can't explain with ordinary matter, then
               | arbitrarily distribute an imaginary 'particle' that
               | solves them: that's DM.
               | 
               | The problem is that the language used to talk about DM is
               | wrong. It's not that DM doesn't interact with EM, or the
               | presence of DM is causing the galaxies to rotate faster
               | than by observed mass. These are all putting the cart
               | before the horse. What we have is unexplained
               | gravitational effects being attributed to a hypothetical
               | particle. If we discovered a new unexplained
               | gravitational property, we would merely add that to the
               | list of DM's attributes rather than say "oh then it can't
               | be DM".
        
               | nyssos wrote:
               | > Dark matter is strictly defined by its effects
               | 
               | All physical entities are defined by their effects!
               | Suppose we found axions and they had the right mass to be
               | dark matter. Would that mean we now "really knew" what
               | dark matter was, in your sense? No, it would just push
               | the defining effects further back - because all an axion
               | is is a quantum of the (strong CP-violation term promoted
               | to a field).
               | 
               | Just like the electromagnetic field is the one that acts
               | on charged particles in such and such a way, and a
               | particle is charged if the electromagnetic field acts on
               | it in that way. There's no deeper essence, no intuitive
               | "substance" with some sort of intrinsic nature. _All_
               | physical properties are relational.
        
               | andrewflnr wrote:
               | > For the Occam razor, adding some random term to an
               | equation is not really better than adding some invisible
               | but well characterized stuff...
               | 
               | You're being too kind. It's worse. Especially when (in my
               | understanding anyway) that added term doesn't even
               | explain all the things dark matter does.
        
               | adrian_b wrote:
               | Adding any finite number of parameters is strictly better
               | than adding an infinity of parameters (i.e. an arbitrary
               | distribution of dark matter chosen to match the
               | observations).
        
               | Dylan16807 wrote:
               | The distribution has to be consistent forward and
               | backwards in time. It's a lot less arbitrary than you're
               | implying, and adding a hundred parameters (or similar
               | finite number) to gravity is not better.
        
               | adrian_b wrote:
               | If we add an arbitrary amount of dark matter everywhere,
               | to match the observed motions of the celestial bodies,
               | that adds an infinity of parameters, and not even a
               | enumerable one.
               | 
               | This obviously can match almost anything and it has
               | extremely low predictive power (many future observations
               | may differ from predictions, which can be accounted by
               | some dark matter whose distribution was previously
               | unknown), so it is a much worse explanation than a
               | modified theory of gravity that would have only a finite
               | number of additional parameters.
        
               | smallnamespace wrote:
               | > to match the observed motions of the celestial bodies
               | 
               | The point is that even with current observational data
               | there's no reasonable distribution of dark matter that
               | correctly explains all evidence that we have.
               | 
               | Your intuition that "if I have an infinite number of
               | degrees of freedom anything at all can be fit" is leading
               | you astray here.
        
               | adgjlsfhk1 wrote:
               | the reason this isn't true is that by the hypothesis of
               | dark matter, it follows gravity but not electromagnetism.
               | as such it only fits distributions recoverable from
               | evolving gravity. e.g. if we require a certain
               | distribution today, it fixes the distribution at all
               | other points in time, and we can use light speed delay to
               | look into the past to verify whether the distributions
               | have evolved according to gravity.
        
               | Retric wrote:
               | All observations of individual galaxies occur at a
               | specific point in time. We can't use light speed delay to
               | see the evolution of individual galaxies only completely
               | different galaxies at some other point in time. As such
               | each galaxy gets its own value for the amount of dark
               | matter.
               | 
               | At minimum this is a ~200 billion parameter model, and
               | more if you're looking at smaller structures.
        
               | beeforpork wrote:
               | 'dark matter' is not a theory, it is the name of an
               | observational problem.
               | 
               | There are many theories to explain dark matter
               | observations. MOND is not a competitor with 'dark
               | matter', because MOND is a theory and it tries to explain
               | some aspects (spiral galaxy rotation) of what is observed
               | as the dark matter problem, which consists of many more
               | observations. There is no competition here. There are
               | other theories to explain dark matter, like dark matter
               | particle theories involving neutrinos or whatever, and
               | these may be called competitors, but dark matter itself
               | is not a theory, but a problem statement.
        
               | XorNot wrote:
               | Yes and no...MOND's core proposition is that dark matter
               | doesn't exist, and instead modified gravity does.
               | 
               | Whereas you can have many proposals for what dark matter
               | is, provided it is capable of being almost entirely only
               | gravitationally interacting, and there's enough of it.
               | 
               | MOND has had the problem that depending which MOND you're
               | talking about, it still doesn't explain _all_ the dark
               | matter (so now you 're pulling free parameters on top of
               | free parameters).
        
               | wavemode wrote:
               | I've seen this opinion before, but can't seem to square
               | it with any reliable source.
               | 
               | Wikipedia has: "dark matter is a hypothetical form of
               | matter that appears not to interact with light or the
               | electromagnetic field ... Although the astrophysics
               | community generally accepts dark matter's existence, a
               | minority of astrophysicists, intrigued by specific
               | observations that are not well-explained by ordinary dark
               | matter, argue for various modifications of the standard
               | laws of general relativity. These include modified
               | Newtonian dynamics, tensor-vector-scalar gravity, or
               | entropic gravity."
               | 
               | Even its name, "dark matter", sort of strongly implies
               | this. If someone were just trying to refer to the
               | observations, rather than a specific explanation for the
               | observations, wouldn't they just say "abnormal galaxy
               | rotation curves" rather than "dark matter"?
               | 
               | I'm not saying Wikipedia is an end-all-be-all source on
               | this, I'm just asking where you're getting this alternate
               | definition. If it is somewhere reliable then perhaps the
               | article needs to be rephrased.
        
             | nyssos wrote:
             | > What can't be fit by declaring the amount of dark matter
             | that must be present fits the data?
             | 
             | Tons of things - just like there are tons of things that
             | can't be fit by declaring the amount of
             | electromagnetically-interacting matter that must be present
             | fits the data.
             | 
             | You can fit anything you like by positing new and more
             | complicated laws of physics, but that's not what's going on
             | here. Dark matter is ordinary mass gravitating in an
             | ordinary way: the observed gravitational lensing needs to
             | match up with the rotation curves needs to match up with
             | the velocity distributions of galaxies in clusters; you
             | don't strictly _need_ large scale homogeneity and isotropy
             | but you really really want it, etc. Lambda-CDM doesn 't
             | handle everything perfectly (which in itself demonstrates
             | that it's not mindless overfitting) but neither does
             | anything else.
        
               | XorNot wrote:
               | You also have to do other things like not break General
               | Relativity.
               | 
               | Which MOND does: it creates huge problems fitting into
               | GR.
               | 
               | Whereas dark matter as just regular mass that interacts
               | poorly by other means does not.
        
               | jampekka wrote:
               | There are modified gravity theories that are
               | compatible/extensions to GR, e.g the f(R) gravity
               | theories.
               | 
               | Nobody probably believes MOND as such is some fundamental
               | theory, rather as a "theory" it's sort of a stepping
               | stone. Also MOND is used often interchangeably (and
               | confusingly) with modified gravity theories in general.
        
               | karmakaze wrote:
               | > Dark matter is ordinary mass gravitating in an ordinary
               | way: the observed gravitational lensing needs to match up
               | with the rotation curves needs to match up with the
               | velocity distributions of galaxies in clusters
               | 
               | Those are all the same thing, the shape of spacetime. The
               | only thing DM adds is a backstory that this shaping comes
               | from hypothetical undiscovered particles with properties
               | that match observations.
        
             | andrewflnr wrote:
             | Dark matter is constrained by, among other things,
             | dynamical simulations. For instance, here's an example of
             | reproducing real world observations, that previously didn't
             | have great explanations, using simulations with dark
             | matter:
             | https://www.youtube.com/live/8rok8E_tz8k?si=Q7vmQYpZr_6K7--
             | m. And that's not even getting into the cosmology that has
             | to (and mostly does) fit together.
        
               | karmakaze wrote:
               | Interesting that you should link that video. Its title
               | card says "Angela Collier". Here's a more recent video by
               | the physicist[0].
               | 
               | Re: where it says "using simulations with dark matter",
               | we can't _simulate_ DM because it doesn 't _have_ any
               | properties beyond our observations. All we do is
               | distribute amounts of _it_ to match observations. _It_
               | could be  "Dyson spheres with EM shields" and the results
               | would be the same.
               | 
               | [0] https://www.youtube.com/watch?v=PbmJkMhmrVI
        
               | andrewflnr wrote:
               | Yes, and I think that video is stupid. She doesn't use
               | the term that way in her own talk, and neither does any
               | scientist I've ever heard. I think she's trying to make
               | some abstract point about science in general and muddying
               | the water in the process. Her takes on terminology are
               | often bad IMO.
               | 
               | That doesn't take away the fact that when you work with
               | the slightly more specific theory of "particle dark
               | matter" it produces real results. And I believe there's a
               | lot more work over the years in similar areas. It doesn't
               | get talked about because it's not sexy, so people who
               | only follow cosmology when there's drama don't hear about
               | it. That was just the example at the top of my mind
               | because I'd seen it recently, and the result is really
               | quite spectacular. Did you watch it through?
        
           | Calavar wrote:
           | > > In 1953, Enrico Fermi criticized Dyson's model by quoting
           | Johnny von Neumann: "With four parameters I can fit an
           | elephant, and with five I can make him wiggle his trunk."
           | 
           | For those who are interested, you can watch Freeman Dyson
           | recount this conversation in his own words in an interview:
           | https://youtu.be/hV41QEKiMlM
        
         | dilawar wrote:
         | Hmm..
         | 
         | Hodgin and Huxley did ground-breaking work on squid's giant
         | axon and modelled neural activity. They had multiple parameters
         | extracted from 'curve fitting' of recorded potential and
         | injected currents which were much later mapped to sodium
         | channels. Similarly, another process to potassium channels.
         | 
         | I woudnt worry too much having multiple parameters -- even four
         | when 3 can't just explain the model.
        
           | nyssos wrote:
           | Neuron anatomy is the product of hundreds of millions of
           | years of brute contingency. There are reasons why it can't be
           | certain ways (organisms that were that way [would have] died
           | or failed to reproduce) but no reason whatsoever why it had
           | to be exactly _this_ way. It didn 't, there are plenty of
           | other ways that nerves _could_ have worked, this is just the
           | way they actually do.
           | 
           | The physics equivalent is something like eternal inflation as
           | an explanation for apparent fine-tuning - except that even if
           | it's correct it's still absolutely nowhere near as complex or
           | as contingent as biology.
        
         | will1am wrote:
         | The balance between empirical data fitting and genuine
         | understanding of the underlying reality
        
         | edflsafoiewq wrote:
         | Isn't the form of an equation really just another sort of
         | parameter?
        
           | qarl wrote:
           | Yes, it is.
           | 
           | Which makes the only truly zero parameter system the
           | collection of all systems, in all forms.
        
           | tobias2014 wrote:
           | This is why I think that modeling elementary physics is
           | nothing else than fitting data. We might end up with
           | something that we perceive as "simple", or not. But in any
           | case all the fitting has been hidden in the process of ruling
           | out models. It's just that a lot of the fitting process is
           | (implicitly) being done by theorists; we come up with new
           | models and that are then being falsified.
           | 
           | For example, how many parameters does the Standard Model
           | have? It's not clear what you count as a parameter. Do you
           | count the group structure, the other mathematical structure
           | that has been "fitted" through decades of comparisons with
           | experiments?
        
             | slashdave wrote:
             | You are using the word "fitting" rather loosely. We usually
             | "fit" models of fixed function form and fixed number of
             | parameters.
             | 
             | You are also glossing over centuries of precedent that
             | predate high-energy physics, namely quantum field theory,
             | special relativity, and foundational principles such as
             | conservation of energy and momentum.
        
           | rocqua wrote:
           | It tends to be a parameter that can be derived from rrasoning
           | and assumptions. This contrasts to free parameters where you
           | say "and we have no idea what this value should be, so we'll
           | measure it"
        
           | olalonde wrote:
           | Kolmogorov complexity[0] solves this loophole :)
           | 
           | [0] https://en.wikipedia.org/wiki/Kolmogorov_complexity
        
             | gowld wrote:
             | Kolmogorov complexity is an effort to wrangle it. It's
             | impossible to fully solve.
             | 
             | You can change measured complexity by altering the baseline
             | assumptions.
        
         | ahazred8ta wrote:
         | Notably done for the first time irl in "Least square fitting of
         | an elephant", James Wei (1975) Chemtech
        
           | Pamar wrote:
           | https://demonstrations.wolfram.com/FittingAnElephant/
        
       | bee_rider wrote:
       | Ya know, in academic writing I tend to struggle with making it
       | sound nice and formal. I try not to use the super-stilted
       | academic style, but it is still always a struggle to walk the
       | line between too loose and too jargony.
       | 
       | Maybe this sort of thing would be a really good tradition.
       | Everyone must write a very silly article with some mathematical
       | arguments in it. Then, we can all go forward with the comfort of
       | knowing that we aren't really at risk of breaking new grounds in
       | appearing unserious.
       | 
       | It is well written and very understandable!
        
       | xpe wrote:
       | One take away: Don't count parameters. Count bits.
        
         | Scene_Cast2 wrote:
         | Better yet, count entropy.
        
           | xpe wrote:
           | Why "better"? Entropy in the information theoretic sense is
           | usually quantified in bits.
        
       | xpe wrote:
       | Another take away (not directly stated in the article but
       | implied): Counting the information content of a model is more
       | than just the parameters; the structure of the model itself
       | conveys information.
        
         | will1am wrote:
         | I think often underappreciated insight
        
       | pharmacy7766 wrote:
       | One parameter is enough:
       | https://aip.scitation.org/doi/10.1063/1.5031956
        
         | dweinus wrote:
         | "This single parameter model provides a large improvement over
         | the prior state of the art in fitting an elephant"
         | 
         | Lol
        
         | Nition wrote:
         | Nice. This is like how you can achieve unlimited compression by
         | storing your data in a filename instead of in the file.
        
         | redox99 wrote:
         | That's like saying your entire hard drive is a single number.
        
           | comex wrote:
           | https://en.wikipedia.org/wiki/Illegal_number
        
           | gowld wrote:
           | https://github.com/philipl/pifs
           | 
           | > pfs: Never worry about data again!
           | 
           | > pfs is a revolutionary new file system that, instead of
           | wasting space storing your data on your hard drive, stores
           | your data in p! You'll never run out of space again - p holds
           | every file that could possibly exist! They said 100%
           | compression was impossible? You're looking at it!
        
         | danbruc wrote:
         | The number of parameters is just the wrong metric, it should be
         | the amount of information contained in the parameter values,
         | their entropy, Kolmogorov complexity or something along that
         | line.
        
         | zellyn wrote:
         | From the paper:
         | 
         | > Paintadosi [4] argues that one parameter is always enough. He
         | constructed a function that, through a single parameter, can
         | depict any shape. However, in essence, this work is a form of
         | encoding, mapping the shape into a real number with precision
         | extending to hundreds or even thousands of decimal places. For
         | our problem, this is meaningless, although the paper's theme is
         | that "parameter counting" fails as a measure of model
         | complexity
        
       | aqme28 wrote:
       | > It only satisfies a weaker condition, i.e., using four non-zero
       | parameters instead of four parameters.
       | 
       | Why would that be a harder problem? In the case that you get a
       | zero parameter, you could inflate it by some epsilon and the
       | solution would basically be the same.
        
         | Sesse__ wrote:
         | They also, effectively, fit information in the indexes of the
         | parameters. I.e., _which_ of the parameters are nonzero carries
         | real information.
         | 
         | In a sense, they have done their fitting using nine parameters,
         | of which five are zero.
        
           | aqme28 wrote:
           | I didn't read enough to catch that. How the heck did they
           | justify that?
        
         | nyssos wrote:
         | > In the case that you get a zero parameter, you could inflate
         | it by some epsilon and the solution would basically be the
         | same.
         | 
         | Not everything is continuous. Add an epsilon worth of torsion
         | to GR and you don't get almost-GR, you get a qualitatively
         | different theory in which potentially arbitrarily large
         | violations of the equivalence principle are possible.
        
           | aqme28 wrote:
           | That's not relevant here though, because their function _is_
           | continuous and they 're fitting to an arbitrary shape. It's
           | not a "perfect science," so there would be wiggle room.
        
       | ggm wrote:
       | No love for D'Arcy Thompson on growth and form? His parametric
       | models for organisms were quite nice (if very simplistic)
       | 
       | https://en.wikipedia.org/wiki/On_Growth_and_Form
        
       | classified wrote:
       | What is that horizontal bar above r0 in the last equation?
        
       | pietroppeter wrote:
       | Love how they misspelled Piantadosi as Paintadosi :)
        
       | foobarian wrote:
       | Reminds me of an old joke: "What is the difference between an
       | elephant and an aspirin?" - "There isn't any, except the elephant
       | is large, wrinkly and grey."
        
       | tagami wrote:
       | ATCG
        
       ___________________________________________________________________
       (page generated 2024-07-15 23:02 UTC)