August 2024: How realistic are simulations?

You're right, those costumes from "ST: Picard" look really great. I think they changed between the seasons, I think the example was from the 2nd? Idk, but looking at them, what popped up again for me was: Why is it always pajamas? I mean, in the 60s, 70s, I get it. But ... we're in the 2020s and "The Future (TM)" is still supposed to look like Barbarella, just with a few triangles and stripes here or there? I'd love to listen to those discussions in the writers room. "No, no, Carmen, that's not possible. Your proposal for the colour scheme, that's great, but it can't be like, you know .. it can't be clothes. It has to be pajamas. Can you redo that for tomorrow's meeting and then we'll see?"

Wouldn't it be more appropriate to say "I'm running an approximation of X"? Because you're saying "actually, *really* simulating the thing is a) unnecessary and b) most of the time not feasible because the computing hardware doesn't exist yet or is too expensive". I know the answer. That's just what it's called, and you don't make the rules. Just saying ...

This is an issue that comes up in F1 car design a lot. They use a lot of simulations for aerodynamics but they still can't beat a wind tunnel, and even a physical simulation like a wind tunnel isn't as accurate as sensors and flowvis paint on the car going around the track. See https://www.formula1.com/en/latest/article/testing-explained-rob-smedley-on-flow-vis.7nU2VePGlVrhIGa8wgCoLE if you're curious. Also for what its worth, watching this many æons later, I'm getting ads.

on epistimology(sp?) I had a great idea for a youtube video the other day on truthiness, the underrated value of wikipedia and its major flaws, the spread of information via social media, something about books. and I had all these great ideas and insights. and I didnt write any of it down which is why I sound like a stoner right now. anyway dont blane you if you beat me to the punch. Ive been editing a youtube video forever now. also great video on simulations. when I did a stint at the nrc (national research council of canada) there was a guy doing sims on like... backpropagating energy from cars into the system/grid. and he was modelling at the electron level. and his sims "took forever" and I always wondered if he was purposefully slowing down his research to keep his job. also had some people at nrc doing some atomic clock physics stuff which I loved. where is this comment going? anyway good video

I'm guessing you buried the lede intentionally.

The way I explain the art of implementing the exact right amount of physics is that they have Rube Goldberg competitions for middle schoolers... The most annoying question I get (as an engineer who does computational physics) is, "Can you see the pistons/flames/gears/whatnot moving? How good are the graphics?" Hmm, I might get a shitty 1Hz slideshow if I'm using a commercial solver, otherwise I'm just looking at squiggly lines or perhaps wiggly surfaces if I'm feeling spicy. I really like numerical methods, there are a lot of fun tidbits like how CFD stability criteria are just making sure you don't lose information or break causality - CFL conditions just check the speed of sound and your discretization length and timestep to make sure a pressure wave can't pass all the way through a cell inside of a single timestep, and upwind schemes to solve convection problems are unconditionally unstable if you have a sign error and make it a downwind scheme, for examples. I use numerical methods, reduced-order modeling, empirical correlations to cut out steps too expensive for their fidelity benefit, and whatever computer science I can learn off Google to make physics models run in real-time on embedded MCU hardware. and the hardware never changes (conservative industry) but my code always does and I've been sneaking a little more dynamics and a little more fidelity in my code for years. I don't think the software will ever be perfect because there are optimizations for every possible use case and every solver only covers a subset.

i watched your picard video. then i watched a few other random ones. then i subscribed and went back to the start and watched everything in order up to this video and i just got to the part where you say maybe nobody will watch your picard video. 👍

Its funny how you thought "no one will watch a video about Star Trek: Picard" because that's how I found your channel haha. I wonder how many others discovered you through that video. Been really enjoying all the great content! Its making me want to re-learn calculus and find some physics courses to take :)

Would love to see the epistemology video! Philosophy gets too complex and theoretical for me as a layman so a thoughtful review of it from a scientist would be great :)

Love the video, to your question about the ADs, I do not get ADs on your Patreon videos =)

The weather prediction was really good but got worse the last couple of years. It rains for 20 minutes before the weather app on my phone changes from cloudy conditions to reflect reality.

The YT Player is soooo much better, so i prefer that, but I have premium as well... YT puts ads on all videos because bandwidth is far from free and they can't store and host things for free however, infinite growth and all that

You probably don't know the answer to this question but as a computer scientist who is really quite depressed with my current job, I would love to know what keywords to search for in order to get a job optimizing physics simulations for people like you. That sounds like a job that brings so much more value than like... 80% of what the tech industry is doing right now.

I'm looking forward to the Picard video (what a weird series it was). Those uniforms also got an appearance at the end of Prodigy's season 2 so they're not just from Picard now

Sorry I’m getting to this so late, but I know (as you said, data) that weather predictions are so much more accurate. When someone disagrees that weather prediction is not valuable (and they aren’t a crank) I remind them that even if it’s wrong it is valuable. If they predict a 2 day snowstorm and you go and buy 2 days worth of food, what’s the harm? Versus you don’t buy additional food and there is a storm.

Unless you have premium, the video isn’t going higher than like 720 anyways and looks like garbage most of the time in HQ screens. And I do have premium so it’s not an issue for me personallyyyyy.

I have yt premium, worth it for me even aside from skipping ads. Even if the patreon video player is okay, yt is so convenient for collecting different vids. When you try, I I hope you'd keep uploading to yt in parallel.

How is that video about Nobel prize coming along???😭

Weather forecasts are SO much better these days, it's honestly pretty amazing.

i don't have YT Premium and i'm not running ad blocker, and i didn't see any ads, even when watching on Youtube. So i think what you did works :) also, great video! Given your general and specific expertise i'd love to hear your take (new video idea?) on general climate models and the extent to which they can effectively explore the phase space of climate change w/r/t more extreme outcomes (especially with very specific intervention ideas like deliberate aerosol release or other solar radiation management proposals thrown on top). As in, is there a set of really bad probably feedback-driven outcomes that are (i hope) long tail but which a typical GCM can't really end up in, which would bias its distributional report towards some false but more comfortable mean? Something i worry about a lot without having enough math/physics background to really assess...

I think the predictions have gotten better, but the weather is also becoming more unpredictable, if that makes any sense. There are a lot more "freak" extreme occurrences that can be predicted in general but not in specifics. Or something. No ad so far. You also haven't said your catchphrase! 😞

i run adblocker so im not seeing them on any site (though some want it turned off if you want to surf them).

I'm not too bothered by the ads - and nice to be able to watch on my TV, which I don't think I can do easily if they're only on Patreon.

I started watching Picard. it was something really different to the OG TNG then I started seeing the folks at Red Letter Media pretty much destroying them and I literally stopped. though they actually started to recomend the 3rd season and I loved it. I wish so badly that something like TNG returned or even better a series with actual HArd sci-fi. The only thing that was close was the expanse. I wish somebody like Andy Weird had a series with super detailed math in it. The Martian movie sucked though.

It's totally worth going to a Star Trek con. I was a fan for years and never went to one until a friend in the DC area convinced me to go and it was fun. Been to a couple since. They weren't huge ones, big enough to take over a whole hotel for a weekend, but small enough that you can talk to people and see the same people multiple times through the con.

- I didn't see any ads, but it seems like most of the channels I like are on Nebula now, so maybe that's worth looking into? - weather prediction is so much better now, then even the early 2000's - Picard and Discovery tried so hard but were such missed opportunities. Strange New Worlds, Lower Decks, and Prodigy are great, though! - so good to hear someone talk about how AI has way better applications than large language models!

I'd love to hear your thoughts on Picard. It didn't ruin Star Trek for me, but it did crystalize for me what it is that I like about Star Trek and it's not big serialized adventures, like in Picard. It's a focus on individual episodes where they explore an interesting idea/premise. I got Paramount Plus to watch Picard, but I didn't really enjoy it. But then I tried Lower Decks and I really dug it.

You could look into spinning up a PeerTube instance to host your own videos. I set up a test instance on Elestio for my son and his friends to watch silly YouTube videos without YouTube access, and it works great, albeit with very limited traffic. I have no idea what it would cost to run at whatever scale your channel is currently at, but I'd be happy to help out with the set up if you're interested. I like to think about the distinction between "model" vs "simulation" when it comes to the popular notion of the simulation hypothesis. I often hear the idea shut down by people who consider the cost factor, and it makes me wonder if these same people might take the idea more seriously if it was branded as a model hypothesis. What's to say that reality itself is not merely a model of some sort? Do we have any way of verifying that there really is so much going on around us? That spacetime is really as big and lengthy as it appears? That matter and energy are really so abundant? Perhaps the truth is really much simpler. Perhaps our world is neither flat nor round, but something more akin to the soup of the day, which may vary depending on what side of the bed we wake up on. I like to keep those ideas in the realm of imagination though. Seems a little pointless to go too hard in either direction. Ultimately, we have to make at least a few default assumptions, and to me it just seems kind of stupid for smart people to argue so much about whose assumptions are closer to truth, as if anybody could ever know. Science seems to be working out pretty well for explaining things and getting things done, except it's also brought us to the brink of global destruction, so does it really matter if it's closer to truth? I wish more smart people took AI seriously. I think both the people who are actively working on it and the people who poopoo it are kind of missing the point, particularly the dangers, and just how close we seem to be to wrecking the entire global economy with this power hungry technology, and the power hungry people behind it.

weather prediction: idk, but i like weather dot gov for checking detailed forecasts if im feeling industrious. love that website. weather's not political right? for some reason i found it really amusing to hear you estimate the number of particles/molecules in objects of different sizes

If you'd need to delete the videos on patreon then I'd rather they just be left up on youtube so there's the back catalog of videos. I've seen other people host them on dropbox or google drive or whatever but those are always janky as hell. I'm not sure there's a better option than youtube, all things considered.

My experience over the last few years viewing YouTube videos via Patreon is that the video do not contain ads when you view them from a web browser logged in to Patreon (no youtube premium required). The ads appear only when you open a separate window to view the video (or open the video in an app).

As for ads ... you could try to demonitize the video by saying certin words, right? 😆 I don't see ads anyway cause you are getting that sweet Premium cash from my views! A vote for democracy is a vote for weather!

1) I didn't see any ads. 2) Weather forecasts have definitely gotten better. I'm older(!) and I remember when the 3-day forecast wasn't very accurate. Now, I look at the 10-day forecast, and it's usually (barring unpredictable hurricanes) quite accurate. And 3) I'm a statistician who did a lot of economic modeling of businesses. I totally agree with your take on what modeling is for. Once, 20 years ago, I defended a co-worker who was tasked with modeling a business and his boss was forcing him to get so detailed, he was creating an accounting system for this theoretical business, detail down to the individual transaction, kinda like your example of modeling a building at the molecular level. It really happens(!) and is the wrong approach to modeling. Thanks!

I would watch a Picard video for sure!

Oh yeah and don't bother with the Patreon video player it is crap

What do you mean Picard is crap?!!? My hot take here: it was fan service. And it was great (overall). About simulations: I liked your take on this, I think it makes sense. Yes yes, there's integration errors and step sizes and floating point inaccuracies. But you can always make them as small as needed. Though it kinda is weird when you're trying to simulate a solar system for 1Bi years and then you have your planets flying off at 500Mi years and then you ask yourself if that really happens or if it's your math that sucks I try to understand epistemology a bit, then I remember that philosophy spent a lot of time debating if a vacuum exists. Anyway, I'm not so keen on Popper and I could write a lot about it I guess, but I'll sum up with "knowledge is what lets you do something with it".

My understanding is the ads were not present within patreon app/website only if you took the link to youtube site/app would u get ads. Also epistemology episode sounds cool dunking on one random comment is cool and based lol. *edit okay just tested yes there still is a ad if u watch vid in youtube app/site instead of patreon but i dont rlly care tbh

Imagine tornado warnings being behind a paywall

Like others, I watch on multiple devices with my YT Premium account. So I strongly prefer YouTube videos for acollieralso over Patreon videos. But I will bow to the majority in case most want ad free videos in Patreon.

Love this topic!

Look if there's anything Nickelodeon taught my generation it's that you could be slimed at any time and therefore it is completely feasible that I could actually end up needing all of these clothes

Also love the videos on stuff you watch/read, I think I initially subscribed because of the LotR video lol

Graphics programmers are absolute wizards when it comes to making complex things go super fast. One might expect they are all CS PhDs but the vast majority just wanted to make silly games as a kid and just got really good at all of the things to make a fun game, I think thats really neat. Its one of the few science adjacent disciplines that largely grew out of self taught people.

Ads were off for me* *This person pays for YouTube premium

I dunno about ads since I have youtube premium but I watch these videos on my tv on the youtube app, so I strongly prefer youtube over native Patreon video.

Project 2025 defunding free access to weather data... Good! Because that is probably the singular biggest threat to our country. And It gets rid of all that pesky data that shows global warming is real. Now on to more important things, like the new Captain Disillusion video

I am really looking forward to your Star Trek video. I love that Delta you have behind you. My wife and I went on Star Trek The Cruise this February. It was fun, but clearly we are Star Trek nerds. But we here on Patreon are your super-fans so we will probably watch any video you put out on any topic. I didn't think Picard was that bad. I hope your "Star Trek ruined" thing doesn't last forever. Lower Decks, Prodigy and Strange New Worlds are all great. I totally agree that weather forecasts are much, much better than decades ago. I have YouTube Premium so I never see ads, but you should try uploading something to Patreon just to see what the player, bandwidth, etc. are like.

To answer the weather question (seeing as this is a special interest of mine), I find that the accuracy is very location dependent. For instance, weather modeling in Kansas seems to have been pretty good 15 years ago, modeling in Colorado was okay 5-10 years ago, and modeling in the Central Kentucky to East Tennessee area was garbage 20 year ago and...ok now. It is frustrating to constantly get tornado watches for storms that don't end up having any tornadoes, especially because there are no sirens in this state, apparently. Side note - what is the job title of the programmers that optimize physics code, because that job sounds fascinating.

"stop shitting on the straw man AI tech bro i've invented" is responsible for a compulsory and loud audible reaction from my person.

And about the YouTube/ad thing. I prefer a YouTube video as that makes for a smoother experience on my tablet!

Works great for me. No ads :) By the way isn't philosophy a simulation with a whole load of people arguing about what the rules are? Just a thought.

I think we are a lot of people that would really enjoy your epistemologi video! I can understand why you wouldn't wanna do it and you can't make every video but a low effort Patreon only or an "also" video would still be fun and interesting to watch! But prioraries need to be made I guess...

dunno about anyone else but I am hyped for the star trek picard video. I haven't really seen anyone talk about it aside from the redlettermedia stuff, and I think their commentary tends to be pretty shallow.

I'm excited to hear your take on Star Trek: Picard. I was in high school when TNG came out, and have a special attachment to it, so I appreciated the blatant fan service in Season Three. It turned the series from destroying my interest in Star Trek to a more neutral stance. I'm interested to see what a younger, smarter viewer thinks of all this.

I would love to see the Picard video. I saw the show and yeah it was dogshit, but it did have some good things too buried in there. I liked some of the new cast, like Rios and Soji. Jeri Ryan became a much better actor since Voyager. The return of Q was fun, too bad they fumbled the plot in S2. And of course the return of the OG cast was great. The first few episodes of S3 were even pretty good IMO but I didn't like the Jack plot and then things got dumb again. But my main complaint of the series would be that Patrick Stewart (who was an executive producer and had input on the direction) doesn't seem to understand Jean-Luc Picard or what made TNG a great show and that makes me a little sad.

Software engineer with a physics PhD here. One thing that annoys me to no end when I hear tech bros make wild speculations about the simulation hypothesis is that they imagine that getting a more realistic simulation is just a question of having more computing power, as if that wasn't limited by the laws of physics of the world where the simulation runs. No, to run a realistic simulation, you need the amount of information and the kind of computer that would collapse into a black hole if we could even build it. You could object to that that you don't need to simulate to infinite precision, you can "compress" your representation of the world, so it needs less information, but that doesn't work either, because the world we observe shows no sign of that sort of compression. Quite the opposite in fact, where some physical phenomena that we can and do probe into essentially require huge information amounts. Imagine if you could take a microscope to textures in a video game. Would you see pixels or would you see an infinitely intricate interaction of infinite numbers of particles interacting with the rest of the universe in an impossible dance that happens to correspond perfectly to the macroscopic state? Cause we clearly see the latter.

Here's what I love about this video: you talk about the "waste of energy" problem, which is so frequently dismissed: in academia, but even more in the private and public sector. There's this horrible attitude everywhere of "well, we've got the funding, and it's someone else's money anyway, so might as well burn it." Like, even setting aside the fiscal recklessness of spinning up a cluster of x1.32xlarges, you're killing the planet saving yourself a few hours of code optimization or getting 1-2 extra decimal places on a sim that's not even setting assumptions to near that order of precision. Thanks for treating it seriously while also leveling the appropriate level of mockery on the "guy who checks extra luggage for a four-day trip."

Honestly. I've never felt like physicists(at least at the school I went to, as a whole in my impression) were bad at coding at all. Like. A quarter of computer scientist majors struggle with the basics and a quarter more+ struggle with more nuanced stuff like optimization. Mathematicians can mostly make stuff work but usually had the weirdest idiosyncratic methods and ideas. Physics majors did their own thing and where fine. And the worst were Petro and Mechanical engineers. Of course this was just the likely incorrect stereotypes I gathered from my experience in one school.

Physicist: I have a simulation of the big old physics phenomenon but it’s way slow. I must be a dum dum. 😔 Computer scientist hears: I have a working prototype that I have confirmed is correct. You can use it for verification and validation of your optimization efforts. 🤩 Over here in computer science, we love our physics sisters. They do all the science words so we don’t have to! 😉

Angela book video: "is it okay that I bought 10 books?" Angela this video: "I'm totally buying a star trek Picard costume for this video" :D :D

The part related to epistemology was definitely a surprise to me. Philosophy is a thing I take less and less seriously as time goes on. I can name way too many people who got lost in philosophy in one way or in another. Either by developing ideas nobody ever should take seriously or by going into impossible abstractions even the author was not able to comprehend. And this is something that never happens with people like Fermi or Einstein. They are just like, okay, enough, too much philosophy, now I go for a boat trip, now I play on the violin, bye. So when I read that someone dedicates himself to study Buddhism or effective altruism or takes protestant ethics too seriously, there is some kind of sadness in the back of my mind. In philosophy there is no measure to know whether you are lost and people are very creative to use even things like candy and chess in abusive ways. Philosophy has just too many open ends, offers too much opportunities to develop bad habits.

I was about to start Picard, I’d love to see it

I would 100% like to watch a Picard video. One of the biggest draws for me is how optimistic the Start Trek universe is. DS9's _In the Pale Moonlight_ is about as "gritty" as I want Star Trek to ever get - and even then it only works (for me) because it was a one-off in being ok with end-justify-the-means.

I'd prefer a direct upload to Patreon, regardless of ads, because Getting from the embedded YouTube link in Patreon to the YouTube app is really annoying and the YouTube video player that's embedded is frustratingly bad and has ads

YouTube by default doesn't show ads on embedded videos because of "putting advertisements next to things the advertisers don't support" reasons (probably others as well, maybe revenue sharing concerns). EDIT -- Here's their documentation: https://support.google.com/youtube/answer/132596?hl=en Patreon is a pretty open platform it's very likely they don't want to risk putting advertising next to risquee content that is (or at least used to be) here.

I got no ads. I also have an ad blocker. Also, now I want a sweater that says "This is just a mockup".

Would love to see the Picard video. It didn't quite ruin Star Trek for me because Lower Decks and Discovery exist (... well... existed... looking forward to Starfleet Academy). The Patreon video player is pretty annoying, in my experience, particularly on mobile devices like iPad (using the webpage, not the awful app). Would definitely buy the sweater!

No ads on Patreon, but as soon as I click Watch on YouTube, I get an ad. I'm guessing I wouldn't see ads if I were paying for YouTube Premium.

Regarding ads, I don't see them but I have Ad Blocker. if people aren't already using Ad Blocker: https://adblockplus.org/ I highly recommend it. It's easy to use and configure and it makes web sites that would be unusable for me usable. and if you have a spare dollar or two, give them a donation. However, unfortunately, more and more sites are getting wise to Ad Blocker and won't let you see the site unless you disable it but it still works on many sites and if a site forces me to turn it off, I usually flip them the bird (not sure why, not like the site can see me) and just find another site.

No ads.

But then, I have YT Premium, so I may not be the best canary.

The cap on storage is around 2^50 bytes for one system. That gets you to about 10^15 bytes or around 10^12 simulated atoms. If you network a bunch of machines you could add another say 10^6. That only gets you to 10^18. Fudge it to 10^20 with like mass storage across continents or something. You're still very low... and it wouldn't be fast. This is harder than what we do to train AI models. And that's only 5g of H.

Using Firefox with UBlock Origin, DuckDuckGo, Privacy Badger, and Ghostery, I get no ads on YouTube at any time. YouTube is in a war with these companies to force us to let their ads through, but currently, when YouTube blocks you, you can look up how to adjust the settings in your browser to prevent this, or you can run YouTube in private mode and it works like a charm. This is why advertising is a bad way to fund a business.

CS person here - I think I could pretty easily build that simulation. Just need a six sided box and 5 grams of hydrogen. (sorry)

I got an ad in the YouTube app, fyi! 🤷‍♂️

Also, can we please spill the tea on the group who claims to simulate 10^24 particles? 👀

I usually click the "YouTube" link near the bottom right of the video to pop it out into a new tab. When I did that, I got an ad. I don't see any ads when playing the video on the patreon page. Not sure if that's because youtube "knows" that it already showed me the ad.

I think your previous video did a good job explaining! I was happy with it. But of course I want more detail :D

first! sorry.

Ads have I none.