Simulating The Universe with Supercomputers

Supercomputer simulations of the Universe are a huge part of modern astronomy. In this interview with Dr Andrew Pontzen we discuss, how exactly are these simulations done, what's the future of this industry, how it will shape the James Webb era and is it possible that we live in a simulation after all.
📖 Andrew's book "The Universe In a Box": www.penguinrandomhouse.com/bo...
🦄 Support us on Patreon:
/ universetoday
📚 Suggest books in the book club:
/ universe-today-book-club
00:00 Intro
01:23 Supercomputers
07:02 History of simulations
10:30 How to simulate the Universe
21:50 "Impossible" galaxies by JWST
30:28 Large simulations of the Universe
36:30 How will the JWST era go
40:33 How tech will change simulations
43:43 Can we live in a simulation
49:40 Outro
📰 EMAIL NEWSLETTER
Read by 60,000 people every Friday. Written by Fraser. No ads.
Subscribe Free: universetoday.com/newsletter
🎧 PODCASTS
Universe Today: universetoday.fireside.fm/
Astronomy Cast: www.astronomycast.com/
🤳 OTHER SOCIAL MEDIA
Twitter: / fcain
Twitter: / universetoday
Facebook: / universetoday
Instagram: / universetoday
📩 CONTACT FRASER
[e-mail]
⚖️ LICENSE
Creative Commons Attribution 4.0 International (CC BY 4.0)
You are free to use my work for any purpose you like, just mention me as the source and link back to this video.

[@lucashouse9117]

Best part of my night is watching these interviews. So informative and well done. Helps take my mind off the cancer so thank you!

[@sadderwhiskeymann]

I wish you a quick recovery dude

[@lucashouse9117]

@@sadderwhiskeymann thank you. I appreciate it!

[@bensimonjoules4402]

Hope you get better. Keeping the mind occupied is the best healer.

[@davecgriffith]

Fraser is such a good interviewer!!

[@ronigbzjr]

Yeah he's one of the best in the field of science news.

[@ianshepherd2861]

Great interview...wonderful to have a conversation with someone who actually knows what they are talking about. I have a much better understanding of simulations now than I did yesterday. Thanks!

[@qfman2]

Great interview! Thank you both

[@aspiratedaloha2946]

Great guest! really enjoyed the interview. Cheers

[@diegoalejandrosanchezherre4788]

Personally i think that fundamental questions about the Nature of reality and existence are the most important ones, they are the core and the drive for the human quest for knowledge, therefore, those questions are the core and the drive of thinking, of math, of filósofy, of science itself... Are the most wonder's type of questions for me....

[@KnittedFox]

This was such a great interview!

[@ricksspeedshop]

That was a fantastic subject thank you very much!

[@Chip_in]

Thank you Frazer and Dr Pontzen l enjoyed this interview very interesting ⛳️

[@PetraKann]

Numerical analysis and finite element analysis is used because the differential equations, integrals etc cannot solved analytically. This is very common in science and engineering. The computers run numerical techniques that solve the equations of motion, heat and mass transfer equations etc. Climate predictions as well as weather forecasts and aircraft, bridge, building designs use computer numerical techniques to predict behaviour. The computer crunches the numbers you instruct it crunch. The scientist needs to use the right equations and make the right assumptions and derive the right conclusions.

[@mshepard2264]

When I worked for remote imaging at Natgeo. We were designing an under water vehicle. We were running simulations of control systems. And like the guest said some things are not easy to make run in parallel. So we built monster dual socket EPYC with 64 cores and a ton of ecc ram. It worked better than a cluster or the cloud because it was running in a single memory space.

[@formarosastudio]

Thanks so much for doing this interview! I know I was requesting it a lot. In my own work I am constantly making computer models for architecture/ design etc and something insightful that you realize very quickly is what is missing and what you need more information on. I think in some ways this is why you build these models, and he touched on this a bit in the interview, when he talked about how its hard to simulate magnetic fields and blackholes because we just dont know enough about them. Those are the insights that I think are really productive, also he mentioned how everything interacts in the universe and how galaxies, quasars etc effect eachother and “chaotic” properties can emerge. I think for a future interview it would be interesting to discuss this topic with Santa Fe Institute of Complexity, where they really dive into computation and emergent properties of nature and data.. Anyway, great interview, wish it went a bit deeper, but hopefully this series can continue and we can get there eventually! So much to uncover and so much happening in this field right now. Thanks again Fraser

[@frasercain]

You can always read his book. :-)

[@brucehansensc]

I love the sound of the interview and discussion on complexity. Fascinating, important and highly misunderstood in the general public.

[@triskeliand]

@49:904, don't bother you are speaking to the converted. we are not a hostile audience, we just love to learn. Cheers Dr Pontzen from NZ

[@imetr8r]

I agree that weather forecasts are much better than when I was a child. In the 1950's and 1960's a weather forecast for the following day had perhaps a 50-50 chance of being correct. By the 1970's, three day forecasts became prevalent and the 48 hour forecast was rather accurate. Today, the 10 day forecast is surprising accurate for 6 or 7 days ahead. So today, when I hear people say one can't trust the weather forecast I think they are "living in the past" or not very preceptive (i.e., smart).

[@frasercain]

Yep, 100% agree. Even if the long-range forecast was inaccurate, you can see it shifting in slow motion.

[@YousufAhmad0]

❓ Dr Pontzen said their's no indication of "cutting corners" but some people have suggested that the fact that we have a discretely defined Planck length might potentially be one such indicator. Any comments?

[@triskeliand]

It's one of many knobs... Seems to work quite well where it's set, If you twiddle too much with that constant Things tend to go all out of whack. Right?

[@YousufAhmad0]

@@triskeliand the question is why is there a discrete/quantized minimal limit to any dimension of spacetime to begin with; rather than about what it's exact value is. Shouldn't it really be continuous? Quantization is required in order to be able to simulate continuous real-world phenomena using discrete calculations.

[@Yezpahr]

Delay in CPU to CPU bandwidth sounds like it could be shimmied into the simulation directly as the finite speed of light/gravity/causality (instead of waiting for that data). What if Dark Matter is just a pocket of strong gravitational effects that got ripped off its source during the rapid expansion moment, when the universe suddenly blew up in size? In other words, imagine a source of gravity. These effects are strong at the start but will spread outwards from the source. Boom, rapid inflation happens, the source is now so far away that light will never reach it again, including any and all effects of causality. This pocket of gravitational effects would experience no decay, because causality itself broke.

[@ekeredtv]

I've never laughed so hard while listening to a Universe Today podcast episode. At 26:12 Andrew catches himself from saying, "twiddle our knobs." You can see a small grin appear on his face. Super fun and interesting interview!

[@bmobert]

Its amazing how often nobs are involved is simulation! Doesnt matter if it's astronomy or meteorology...Gotta twiddle those nobs! 😂

[@jmacd8817]

Very cool! One thing I want to know about these simulations is exactly what are they similulating? Particles? Density clouds? Something else entirely? What/how are the interactions modeled? Multiple 3 body problems? A more complex PV=nrT? What language C++? I love the top level discussion, but I'd be thrilled to learn more of the nits and bolts of the actual software... and how they "twiddle the knobs".

[@diegoalejandrosanchezherre4788]

Minute 42: absolutely no!!! 🤦🏽‍♂️You don't need the all universe computational resources as You might think to simulate the universe. Actually the speed of ligth is the time to renderize the causal diamond for observers, and the information of the volume of space can be storage in a two-dimensional surface vía the holographic principle

[@YousufAhmad0]

❓ What granularity do these simulations go down to? Do you simulate down to the atomic interactions that trigger fusion reactions that ignite stars? If not, how do you simulate star formation? Are they just coarse granular approximations?

[@tonywells6990]

Simulating clumps of gas light years across.

[@YousufAhmad0]

❓ What is the holographic principle? How is it possible that all the information held within some volume can fit entirely on its surface area? Isn't the volume of say a sphere much larger than its surface area?

[@PhysicsPolice]

44:00 It's possible to have mixed resolution in a simulation using clever algorithms e.g. the Barnes-Hut algorithm. Why wouldn't this technique be extensible down to simulated creatures who invent particle colliders?

[@frasercain]

I'm sure there are short-cuts that could never be detected. But my general perspective is to reject assumption that each level of a simulation is an accurate version of the next layer of reality up. The Sims might think they're living in an amazing reality, but it has very little in common with our reality.

[@triskeliand]

@@frasercain Put the peeps in boxes. lower the simulation costs. I am more in line with the Bayesian principle But as cream rises to the top, So does the minimum free energy principle apply. The same way RNA will only fold in a form so as to minimise energy, So too will a simulation. Cutting corners is OK. Nobody can observe Everything all at once.

[@KenMathis1]

I'm not saying we live in a simulation, but I am unimpressed with the arguments given against it. You don't have to simulate our entire universe to make us think we live in that universe. As already pointed out, you can do statistical estimations of a vast majority of it, and/or some beings could use observations of their own universe to provide data directly into ours. You could also dynamically change the resolution of the simulation depending on the changing needs of different parts of the simulations at different times. For example, unless someone is doing an experiment at the quantum level, you can probably get by just fine with higher level approximations for most things. The results at the macro scale are just probabilistic outcomes anyway. In fact, and I might be getting way out over my skies on this one, isn't that exactly what we see with the measurement problem in quantum mechanics.

[@Starman_67]

So, Fraser just made a comment regarding the communications bottleneck being a point against us living in a simulation. I would place a counter arguement that this COULD be discounted by bringing quantum entanglement into the equation in ways and scales we've possibly not considered.

[@frasercain]

Sure, if someone figures out new physics then all bets are off.

[@Starman_67]

@@frasercain 😆

[@Starman_67]

@frasercain how about this as a thought experiment, which I feel could be testable one day if the idea is actually sound - I'm just some random dude after all - just suppose that one reason for the simulation is to test the hypothesise of General relativity and the laws of gravity in a universe where the quantum nature of the underlying fabric of reality does appear to correlate? That one of the purposes of senteint species that came about in the simulation was to solve whether the 2 could be united in and overriding theory of everything? 🤔 😁

[@mshepard2264]

The other thing that is irritating about GPUs is most of them don’t have hardware support for double precision floating point. I do like how you can get CPUs with 64 REAL cores now on one processor.

[@chuckmccaslandsr.8859]

In your opinion, how many unknown elements do you believe may be possible in the universe

[@slo3337]

A very large number is possible . After a neutron star collision for example, there briefly are large chunks of nutronium that could be counted as new elements. I'm not an expert

[@yourguard4]

I don't think, there are stable elements, we don't already know.

[@sspoonless]

Wish I had more feedback options than just thumbs up & down. I like this, so it deserves thumbs up. But I am retired computer science major & am familiar with the subject matter, so I give thumbs down to discourage the algorithm from recommending more like it.

[@chuckmccaslandsr.8859]

How many terabytes does this system require to make this program function?

[@Yezpahr]

Terabytes of what? Don't worry, I do grant multiple choice: Ram / Virtual Ram / Storage units / External Storage / CPU Cache / LAN bandwidth (and subsequent bandwidth of aforementioned properties) ...

[@slo3337]

The answer is allot!

[@PhonicallyPsychotic]

Wouldn't we have to create a simulation that's "Just good enough" and not perfect ? Is it not part of the human condition to fill in the details and convince ourselves ? For resources, as a player of MMO's resources are rendered where they are needed, why build the city of Paris if you are standing in New York.

[@xitheris1758]

Their connection was laggy and made for some awkward moments of stumbling over each other. Great interview tho.

[@bozo5632]

If this is a sim, then it's probably an ancestor sim in which our hyperevolved descendents keep trying to figure out how Trump won in 2024. Or not. Either way, we may only have 17 months to live.

[@agentdarkboote]

I wish I had access to this kind of compute... But without paying for it...

[@jamesmoore4023]

Some of them have applications open to get time on them.

[@janklaas6885]

4📍36:30

[@louithrottler]

With the expansion of the universe, the digital archive that humanity has built up of our observations (of the universe) - do we have a 'map' of stars/galaxies that are now 'lost' to us, having now gone past our limitations of observation?

[@tonywells6990]

We cannot observe any of them.

[@qfman2]

Wow at 38 it seems likely that gravitational wave observatories will be the first to spot black holes that were outside the big bang that we are in

[@bobinthewest8559]

Excuse me, but… You can’t take forever to get your simulations right. You have to have it finished before the people inside discover it 😏

[@bravo_01]

❓Could an imploding submarine form into a black hole?

[@jamesmoore4023]

I see more and more evidence pointing to yes in my own life. Dropped a rope dog toy on a stone floor and it sounded like wood on wood. Tried recreating with a witness to it and neither of us could reproduce the sound. I'm excited and terrified of the convergence of AI, AGI, Quantum computing, space exploration, fusion, drug discovery, decoding animal language, SETI, autonomous cars, XR,...

[@edenb329]

schizophrenia...

[@booradley4237]

Wood!

[@amonynous9041]

That's how I simulate this universe with my supercomputer, trolling scientists with a press of a button, leave them head-scratching.

[@christopherfields6704]

🤤 *PromoSM*

[@bozo5632]

IT latency as a Fermi Paradox solution.

[@TheCosmicGuy0111]

Ooo

[@skarphld]

At first glance I thought the thumbnail title read "... Science Simulates Actual Work". Seemed rather rude and judgmental.

[@Brian2]

-_- read the title as "stimulating the universe" which had me thinking something entirely different before rereading.

[@masi416]

The base-reality where a super computer runs the simulation of our reality has no atoms. The matter there is infinitely divisible and has there for no quantum mechanics. Our matter made up from atoms is just their way of simplification so they can run their astronomical simulations. The simulation is not perfect, like superconductivity is a bug resulting from having atoms, but it is small enough bug to not bother them. Simulating life was never the goal of that simulation and is even considered impossible because of the limitations of atoms. The goal was to simulate the evolution of the universe, in such a resolution we can only dream of. Now it is on us to find other ‘bugs’ to actively rewrite the simulation-code from inside the simulation. By achieving this we can create infinite energy, travel faster than light or watch the inside of a black hole. It’s of cause total nonsense, but is there a book covering something like this?

[@slo3337]

I just simulate God, and let him figure it out

[@RoyallyChiefd]

Lol idk if that's a joke? Well played. 👏

[@GadZookz]

Sounds like a good reason for everyone to eat a lighter breakfast. 🤔

[@mshepard2264]

the idea that we are unwittingly in a simulation is useless. It explains nothing. Like what is simulating the simulation that is simulating us. It just like saying “ its turtles all the way down. “

[@triskeliand]

This is a great interview..... It has conjured up all sorts of extra questions, no thanks to Thomas Gleick. But I do reccommend en.wikipedia.org/wiki/The_Ambidextrous_Universe and to be quite Francis with you, adding a bit of noise to a simulation would only make it more realistic.