August video: Are simulations experiments?
This was a really fun video to make. Thanks so much for your suggestions!
I will post the September poll in a week or so.
Link to PhET simulations for fun: https://phet.colorado.edu/en/simulations/filter?subjects=physics&type=html,prototype
Link to Fast Pass video: https://www.youtube.com/watch?v=9yjZpBq1XBE
Comments (46)
When you did your simulation of galaxy formation with a starting DM halo that had angular momentum, does that mean you specified a number of massive objects with some classical angular momentum or was this like a cloud of bosons with integral spin? I presume the former. In that vein, what are the speculations on the nature of a dark matter particle; boson, fermion or hadron? I’m a chemist many years away from quantum chemistry and statistical thermodynamics, so sorry if my questions aren’t great.
Solution: have hair with a slight wave. Ah! An early Feynman bros reference! That video is the one that led me to subscribe here. By the way, the ads are thick & fast even on those unlisted videos: like, every few minutes, 20-30 seconds long (& of course way louder). Are they set for monetisation? (Is that even possible with unlisted ones?) Maybe hosting the videos elsewhere would be a good option for the Patreon offerings.
Just finished watching 2:30 hours of Stephen Wolfram talking to Brian.. Lots of speculation going on all over the place. Best of luck to them. I am glad that I had the framework of this video before watching that one. - Cheers.
My entire PhD thesis is on simulating nanoscale interactions of radiation with macromolecules. It's entirely unverifiable experimentally. What does an 18 MeV photon do if it hits 1 nm of gadolinium? Well, I know... probably.
This was really interesting, thanks for sharing your views on this subject. I'm a software engineer, and I like to write little toy simulations of things to help me make sense of the world around me. That said, even if they were made to be as close to 1-to-1 to reality as possible, I would not trust them as experiments. By virtue of being a simulation within the system it's simulating, it will be lacking. In general, I think, you cannot know you have complete system of a system from within the system, so a simulation will necessarily be missing the unknown unknowns that we hope experimentation to eventually sus out.
wait dr collier is real? I thought she was just a theoretical physicist
I'm a software engineer with basically no background in physics except youtube videos, but I'd love to write simulations for physicists, that sounds way better than any of my actual jobs have been...
Regarding code of physicists vs software engineer, in my experience as a software engineer most 3rd party packages I work with for engineering things is very sensible with good documentation, but whenever I go into "science" categories I quickly lose my mind. Even the most popular and relatively better documented spaces will still have a lot of ??? from me while I try to figure it out. It's anecdata I guess but it happens often enough. I think it's interesting as I wonder if I just need the right people teaching me through it or if it really is that bad and a whole population of coders are being held back.
(note: this is my take as a computational physics grad student) Part of it is that most scientists don't have formal training in software but another part of it is the incentives in academia. As a grad student or postdoc, your incentives are mostly to publish papers and so the code needs to be good enough for one project. There's not a lot of incentives to keep it nice for future researchers or integrate it into a larger package, even if that would benefit the research group and science as a whole. Maintaining good code in computational research is a bit like maintaining a lab in good working order, in that academia doesn't provide incentives to do it on a longer scale than a research project. To what extend it gets done, it's usually out of extreme need, a person with atypical altruism or standards, or hiring a specialist to focus on it.
I'm not sold on this idea that there exists, in principle, some collection of mathematics capable of predicting or explaining everything that ever has and ever will occur. Wolfram's bit on "is this a theory of everything?" made this claim. I guess Laplace was the most famous advocate? Recently I've seen it a lot from believers in "artificial general intelligence" and related ideas (what Timnit Gebru calls "TESCREAL"): if we could figure out all the math behind how brains work, we could re-create a human mind, complete with consciousness. And anyone who rejects this is rejecting the fact that the universe obeys the laws of physics, where "the laws of physics" refer to this hypothetical complete mathematical account that explains everything. This is totally circular, right? Everything obeys the laws of physics, and so therefore we can account for everything using the laws of physics? Why does every single thing that has ever and will ever happen need to have, in principle, some complete mathematical explanation? Is the argument just that math has gotten us really far in understanding reality and so therefore it can get us all the way if only we complete it? We use math to understand reality, and so therefore reality is obligated to conform to math? Seems greedy to me, like we think the universe owes us an explanation of itself,
Your intuition is correct: we know that mathematics is incomplete due to Gödel. You can prove that math can’t prove everything, using math! By the transitive property, you certainly can’t prove physics with only math! 😉 Maybe Wolfram aspires to be the next Bertrand Russell? 🤔
So about the Steve Jobs blurb- I think you're probably right regarding his music tastes growing up. But, long-haired classical music guys are definitely a type of guy. Or, at least, they used to be I imagine they still are.
I'm curious, how feasible would it be for you to do the work you do now or previously in academia if you weren't employed by an organization? For instance, if you were a full-time YouTube person, would you have the ability to create simulations and still do interesting science? Are there barriers (prohibitive cost, availability of powerful enough computers, access to models, etc.) that would stop an expert like you from continuing to contribute to the field?
Stephen Wolfram came to my high school sometime around 2019. (For some reason. We don't bring people around to give speeches, like, ever?) The thing me and my nerd friends noticed the most was that he spent the entire time talking himself up - like he was giving a speech for, like, book journalists (?) instead of a high school. Which to be fair he was probably doing a lot more of the former than the latter. Then we looked him up on Wikipedia and apparently it's a battleground (and IMO he's definitely paid for Wikipedia editors). The "A New Kind of Science" article in particular has a great talk page. --Don't let envy ruin this article-- "The schools opening at top universities for the sole reason of studying computational science more than vindicates Wolfram. The 3rd branch of science has basically been established." -Unsigned comment from 2013.
I felt very similar when I watched him give a talk.
I was glad to hear someone explain what Stephen Wolfram was about. Now I'm not concerned that I don't understand what he is doing.
Great video, I know it takes a while to edit these, perhaps you should think about posting these to your main channel after a month or two. Patreons still get early access and you get more content for the main, Win Win
Yeah, I kinda agree that if you are going to make content you may as well get a larger audience.
"I was so tired of thinking about [The Lord of the Rings]" *Cocks head in confusion* I know those words, but I can't understand that sentence... Good video! I definitely agree that simulations are not experiments. The use of simulation that I'm most familiar with is using tools like FLUKA and GEANT to model the interactions of particles in your detector. If I'm going to interpret Wolfram's whole thing charitably, I'd say that the idea is to come up with an alternative model that reproduces the empirical success of quantum mechanics, QFT, etc., but starting from a cellular-automoton-like ontology (in a similar vein to something like Bohmian mechanics vs. orthodox QM). Which would actually be quite interesting if it could be done! But, as you say, none of (at least what I've seen of) what he has published seems to do that.
Forgive me as I am not a physicist nor am I a doctor of any kind; I’m a mere mortal data engineer. BUT the way I have come to understand the process is: theory > model > simulation > observation repeat. That’s the pipeline. Every time. I don’t care what the domain is, it’s never different. Any one of those steps on their own are kind of pointless in a way? Well maybe not observation, that’s always a good time. I was sort of hoping to hear a bit more about phenomenology because that is kind of where I would have categorized simulations; neither theory nor experiment. The topic selected felt like a false choice in a way? Like it ignored a seemingly valid categorical option in my opinion. But idk I could just be misconstruing what all this is anyway.
What do you mean by phenomenology in this context?
That was a great video! I'm just a regular person, I really enjoyed the exposure to simulations. It was also great to hear more about the work you do. Sadly I am a male and of course failed your test of keeping my eyes on your face at all times. Sorry!
This is strange comment :)
Finished reading this a moment ago and think it applies “Computer-aided design technologies enable existing modeling and simulation capabilities to evolve into digital twins. Digital twins represent actual products, real processes, or both in a virtual state that can be functionally analyzed for performance.” Speicher, Terrance L and DeFranco, Joanna F. ‘Industry 4.0 and Digital Twins’. Computer. June, 2023. DOI: 10.1109/MC.2023.3243450
Also in regard to Richard. He is probably good faith and sees a trend in structures drawn by the simulations and believes these structures represent loose forms of quantum structures. Specifically, hes looking to apply the structures to possible quantum phenomenon. Which is interesting, but if it is his idea the first breadcrumb is on him to determine and not other people.
I had a bunch of guys (yes) arguing in a discord call over if scientific proof was deductive or inductive. It's literally this exact same phenomenon. I don't get why people pick a fence when experimentalist proof work for theorist to operate off of 😭 or theory is how we think of experiments to do.
you dropped "curly girl problem" so nonchalantly that for a minute i actually believed that was a common idiom
Yeah i think it is even regular dudes know or heard of this discussion or complain before, I'm on the opposite side i think she didn't need to explain it at all
I think that the reason you gave for why physicists are bad programmers was actually good programming advice (ie. doing the fastest thing that solves the problem) whereas the real advice you got from a tutor was rather bad (you're already automating trillions of things, why do you need a rule of thumb for automating 10?) :)
Thanks for this. I'm fully guilty of copy / pasting chunks of code and then manually editing little bits of each copy, rather than figure out how to do it elegantly. The inelegant approach usually takes an extra couple minutes of coding; the elegant approach takes an unknown amount of time if it requires googling or looking through documentation. But, I'll still feel embarassed if anyone else sees my copy / paste / edit job.
One of the many interesting things about water is that when doing electronic structure theory calculations, higher level treatments of quantum effects are needed to accurately predict important properties than many materials.
See for instance (this is after a very quick and dirty search): https://pubs.acs.org/doi/abs/10.1021/acs.jpclett.7b00391# https://pubs.rsc.org/en/content/articlehtml/2017/sc/c7sc04205a https://journals.aps.org/prl/abstract/10.1103/PhysRevLett.129.226001 https://journals.aps.org/prl/abstract/10.1103/PhysRevLett.117.186401
Now that commercial space flight is here the true believers in experimental simulations can show their commitment. If simulations are sufficient, they can be a part of the team conducting a mission to Mars experiment. That team will build their mission solely on sims without any rover, orbital, or other data. They’ll all fly there and if they succeed then sims are experiments. Is that a scientific experiment?
You laugh but the current nasa interns were trained on Kerbal Space Program and honestly gives you an intuition for orbital mechanics.
"It's not [physicists] job to check the arxiv for anything that's posted and then offer feedback" I would have said true a week ago but I guess if you claim to have a 400 K Tc ambient pressure superconductor they'll be like a dozen papers in reply within the week.
The idea of Wolfram going McAfee sent me
This video is great! I've mostly understood the power of simulations in astronomy and other areas of science, but your explanations and examples bring the point home, especially about the inherent limitations. I think that "pure" mathematicians use simulations in number theory, but that may not be the same thing.
Great video! I'd almost put simulations between theory and experiments. Simulations are the predictions made by the theory maybe? Like they can help you figure out some unknown value by varying that value in the simulation and see which one most closely matches observations of the real world.
I think Wolfram's behavior is best explained as a combination of narcissism and religion. The narcissism is obvious, but aside from that, I think he truly believes in the religion of Discrete Mathematics. The universe *must* be discrete. It simply must. As a discrete mathematician [1], I find the idea of a discrete universe highly tasteful, but I don't understand his level of certainty. I do think it might be possible to create some kind of elegant discrete model for physics. But he clearly hasn't actually done that. Surely he must be smart enough to know that he's, at the very least, stretching the truth in claiming success. I can only make sense of him as a religious zealot willing to do anything, even besmirch his own reputation, in order to kickstart the research program that will eventually reveal the Truth and therefore justify everything. [1] I demand the right to continue to call myself a mathematician despite leaving academia and now doing software development all day. One day I will publish another paper to prove it. Hmph.
Great video Dr C! I'm not good enough at anything to understand the critique against Wolfram, though I know that's how many scientists feel. Is his stuff plain wrong, or is his system somehow like that The Library of Babel website where it contains absolutely everything because it's produced by an algorithm, so of course it will contain X. I enjoyed the pen and paper guy anecdote.
I'm not a physicist so I don't know how physicists feel. But I have a guess. The papers promise that the "Wolfram model" will exhibit behaviors akin to relativity and quantum mechanics. I think most physicists lose interest when they get to the point in the papers where physics terms like "Lorentz transformation," "quantum mechanics," and so on start being introduced, and it becomes clear that the usage of those terms is quite distinct from the way any of those terms is used in physics. So it doesn't really matter whether the model is too general or not, because it's just... not physics. For example, if you know Conway's Game of Life, it has a universal speed limit. If you change some cells, then in the next step the neighboring cells could be affected. In the following time step, the cells next to those could be affected too, and so on. There's no way for any information to propagate faster than that, so the speed of causality in Life is 1 cell per time step. Every Life enthusiast calls this the speed of light, for obvious reasons. And it's a good analogy. You can say things about the causal separation of events in Life that are analogous to statements you can make about events in the universe. In real life, this is the subject of special relativity. Where things get derailed, in my opinion, is when you start to think that because Life has a "speed of light," and the universe has a "speed of light," that these are the same concept. If you start using the term "special relativity" to describe the study of causal relationships in Life, and your tongue isn't firmly planted in your cheek, you've lost me. Yet Jonathan Gorard (Wolfram's acolyte) does almost exactly this, unblushingly, in "Some Relativistic and Gravitational Properties of the Wolfram Model." (He's not talking about Life, of course, but about a fairly general family of hypergraph transformation models. However, if you understand how the speed of light works in Life, you also understand how it works in the hypergraph transformation models.) I'll admit that once I realized what he meant by "special relativity," I did not continue to read the paper to the section about "general relativity." Once I realized that the analogy with special relativity was so superficial, I just didn't care enough to see his next analogy. And that's not a good sign -- I love cellular automata, so if you're talking about cellular automata (or even hypergraph transformation models) and you're losing me, you're probably in trouble.
I'm a philosophy student and I, too, hate Aristotle...
Don't leave me hanging.
Thanks for the video! Can I share a couple of thoughts? The first one is a question. 1. In your example of simulating physical behaviour with different values of G, what about a case where you model, say, our solar system and you produce an ensemble of simulations using different G values and you select the one that looks closest to what we observe with telescopes? If such an approach worked, wouldn’t it be a way to arrive at a close approximation of the real value of G? And what if you then used a similar approach to get a value for some currently unknown physical quantity? Couldn’t you get new physics out of a simulation that way? (I think of it like a codebreaker trying different ciphers until one produces plaintext that’s in English.) 2. This is based on my admittedly very shaky grasp of Wolfram’s ideas. I think Julia/Mandelbrot sets make the point pretty well. Where the heck does that infinite complexity come from when all you’re doing is iteratively solving a fairly simple equation for different initial values? I think Wolfram thinks there’s something fundamental being revealed when very simple iterative calculations produce incredibly richly patterned outputs. Wolfram has invented all kinds of iterative algorithms that evolve incredible complexity apparently out of nowhere. And he thinks that physical matter is doing something similar: it’s iterating over its last output to produce its next output. Matter is in some sense computational, according to him (whatever that means). And I’m not sure he can explore any of this with conventional maths because (as far as I know) nobody has any suitable notation or any methods for the algebraic solution of iterative functions. Not sure if he’s on to something or not, but maybe you have to know the answer to the question “Why Is Mandelbrot Sets?” in order to figure out whether he’s following a red herring or not.
The problem with trying to determine G from the solar system is that we don't know the masses of the planets and sun very precisely. In fact, the standard planetary masses are derived from values of G measured in laboratories, originally by measuring the extremely tiny gravitational force between 30-cm balls of lead. Once you have an experimental value for G, you can assign the masses of the planets in a way that makes the solar system simulate correctly. If your value of G later changes because your original experiment was inaccurate, you'll end up changing your opinion on the masses of the planets. So there are enough degrees of freedom (G and all the planetary masses) that you can't work out the gravitational constant just from looking at the solar system, either in real life or in simulation.
I loved the section about making the galaxy model. That’s my favorite aspect of being a computer scientist: working with all sorts of different domain experts, learning what they need and helping them make it happen on the computer. And then doing it all again because they just had another great idea! 🙂