The biggest hand calculation in a century! [Pi Day 2024]
Vložit
- čas přidán 12. 03. 2024
- Please note down the new value of pi: 3.14159265358979323846264338327950288419716939937510582097494459230781640628620899862803482534211706798214808651328230664709384460955058223176
If you'd like to sign up to hear about future ridiculous maths projects that require volunteers: forms.gle/w44THpNJ3jWUPqHy6
Play list of all my previous calculating pi day videos: • Matt calculates π
This calculation was paid-for with surplus funding from a previous Stand-up Maths kick-starter. If you'd like to support future projects, you can join me on Patreon. My Patreon supporters keep me irrational. / standupmaths
Huge thanks to the Pi-by-hand Committee who spent two years working on this project:
Alex Genn-Bash
Ayliean McDonald - / @ayliean
Ben Sparks - / @sparksmaths
Christian Perfect
Deanna Judd
James Grime - / @singingbanana
Katie Steckles - / @katiesteckles
Matthew Scroggs - / @chalkdustmag
Max Hughes
Nicole Jacobus
Sophie Maclean - bit.ly/Sophie_Numberphile
Thanks to Steve Mould for making sure everything was legit, and only mentioning tau six or so times.
William Shanks was voiced by the excellent Ben Moor: www.spesh.com/ben/
The amazing voice over was provided by the fantastic Gemma Arrowsmith: www.gemmaarrowsmith.com/
CORRECTIONS
- None yet, let me know if you spot anything (other than the final 6).
- Yes, this video was released the day before Pi Day. It's so teachers have a chance to watch it and prepare their Pi Day lessons for the big day.
Filming and editing by Alex Genn-Bash
Animations by William Marler
Written and 'performed' by Matt Parker
Produced by Nicole Jacobus
Music by Howard Carter
Design by Simon Wright and Adam Robinson
MATT PARKER: Stand-up Mathematician
Website: standupmaths.com/
NEW BOOK: mathsgear.co.uk/products/love... - Zábava
"Long division, not wrong division" spoken like a true line manager
So much so that I can hear Rhys Darby saying it
"We are dividing but we are not deviated" -matt later on in the video
@@skarrambo1 ...not swearwolves
I was actually thinking it would be a good exercise for IT middle managers to do so they get to understand IT and its dependencies.
The skill set is not that broad. They could do it.
Meanwhile Japanese folks be like "long division, rong division" same thing!
posted at 3:14pm GMT, well played Matt
15 seconds after 3:14 pm I presume? 🤔
If only he'd (or presumably his editor'd) managed to edit 5:07 off the run time.
And 926 milliseconds of course @berend_dijk
I noticed that!
but one day before 03.14 😣
23:24 The madlads built a branch predictor into their human GPU.
underrated comment!
Absolutely underrated-I think this is one of my top 20 or so sentences ever said in the English language
This is actually incredible. If they do this again, I wonder if they could get a cpu architecture engineer to help design the system.
@@MIKAEL212345 unfortunately the huge error rate and the time scales involved are incomparable with semiconductors. My guess would be that we'd have to have an architecture very very different from a processor.
But I like the overall idea of engineering a better process with those constraints in mind.
The insane thing is that it just, like, evolved naturally. This is divergent evolution.
"we cant cheer for every digit"
proceeds to cheer for every subsequent digit
Until the last one
"Ridiculous Maths Person" is the greatest complement I can recall ever hearing for an introduction.
But what is the complement of a Ridiculous Maths Person?
@@richardfarrer5616 touché, foiled by misspellings again
And I'm going to say solar physicist is the complement to ridiculous maths person
That had me laughing so hard, I had to pause the video and recuperate.
A physics professor and solar physicist of course
Quality delivery throughout. Is the narrator credited?
0:48 "I don't think we are going to do this..."
0:52 "We're going to do this.."
Give your editor a raise for that cut.
I think Matt mastered the art of thinking in multiple timelines at once at this point.
Yeah, but then the editor needs a pay cut for the terrible audio levelling throughout the video, so it'll cancel out.
@SgtSupaman they'll get it. If you know anything about it, you know it takes time and repetition to get good at. I'm grateful for their work regardless and can't wait to hear them improve
There were some more of these too, like at the end of day 3: "I love watching this human machine going to smoothly" *jump cut to DAY 4* _All hell is breaking loose and new jobs are springing up by the hour._
@@kindlin I was going to mention that too
A hundred years ago, we would call these people computers.
yup, it was a job description, even well into the 20th century
There were human computers all the way to the 1960s….
Technically speaking, they still are computers
How did the early computers do calculations?
@@fredrickcampbell8198 Pencil and paper, adding machines, log tables... 🙃
I noticed that at 3:14 , the names of the 7 arc tans are each 1 of 7 of their own group of 7; day of the week, samurai, continent, sin, ocean, pyramid(edit: wonder not pyramid), and dwarf!
I love that that occurred at pi minutes into the video too.
Ancient wonder, not pyramid.
The great pyramid of Giza is one of the 7 wonders of the world!
I thought the names were so random, obviously that wasn't the case. Thanks for filling me in on this!
That's so clever, lol! Thanks for pointing it out
8:22 "Chunks of 20."
8:30 "60 - 100"
Ah, the Parker Chunk.
😂👏
I was going to note that as well. However, they did correct it at 9:06
Well seeing as the previous range was 60-80 it was still a range of 20
Also wouldnt a chunk of 20 be 0-19, 20-39, etc.? (Or 1-20, 21-40, etc.)
@@k0pstl939that’s true. I wonder what would be digit 0. If that’s 3 then that can be omitted and the range would be 1-20,21-40etc
You know you're in the right show when the crowd goes crazy over a string of digits.
Like sportsball fans? 😁
You'd think it was the lottery announcement.
I literally got goosebumps with the crowd hahaha
@@ailaGI love sportsball! (cue relevant xkcd).
It's a great thing to work with normal people for six days.
For anyone curious, the 140th digit of Pi is "2".
Thank you for that comment
Spoilers!
3.14159...282725352848111 is all I remember (the 7 is the 139th digit)
To me what is amazing is how closely you emulate a modern bureaucracy. You have created a controlled experiment in administrative science that most researchers could only dream of. It's even realistic in that some agencies have linear tasks that administratively resemble calculating pi. I wonder if you thought about creating incentive structures.
Being part of something unique and crazy and where everyone there has a shared passion is kind of its own incentive tbh.
@@alexpotts6520 true! In public administration it is called public service motivation. There seems to be a lot of that here--but they also manage to cultivate it among themselves (e.g. the presentation of the results).
@@alexpotts6520 True, but imagine if there were _actual_ incentives, as well? Get 10c for each (total) digit, each day you're there. Calculate a lot of digits and be their most days and maybe get 100$ or something the time around. That is probably too much, but maybe it's X dollar gift cards, which could conceivably be sponsored by the gift card providers themselves (as gift cards are amazing marketing).
"...most of the people on it were unhappy for pretty much of the time. Many solutions were suggested for this problem, but most of these were largely concerned with the movements of small green pieces of paper, which is odd because on the whole it wasn't the small green pieces of paper that were unhappy."
Nice Douglass Adams reference there.
That came to my mind, as well!! :D
What
@@gollossalkittyfrom The Hitchhikers Guide to Galaxy.
But doesn't that mean pi = 42?
35:25 'I came here for pi, and all I got was sandwiches'
Legend
underrated quote
To be fair, one of the participants baked and brought a pi but it wasn't big enough for everyone to get a slice. I didn't 😢
Those students at the end were fantastic. I especially like the guy who rated the event a 3.1415 out of 5.
@@julienroy3355one of the participants was baked, and i suspect it was the dude with the sandwiches
Alternate title: "Matt builds a human GPU (again)"
*CPU
@@MrNikolidas *APU
no GPU was right
@@HamStar_In that analogy, at a stretch Matt himself is the GPU. He built a CPU that gave him the data necessary to display it on the paper.
@@MrNikolidasGPUs are excellent at massively parallel computation. The main thing they do is run small, similar, independent calculations in parallel, similar to what the human calculators are doing.
Something that should be noted about why we were so surprised by the last few digits,
the 2 numbers, each 140 digits long that were subtracted to give us pi, were found by adding 7 different 140 digit long numbers each.
when adding 7 numbers together you can get carry over that goes several digits over. So to know what a digit is, it isn't enough to calculate that many digits, you need to calculate several more digits to be able to get the carry over as well.
the last 5 digits all could have been wrong without us making any mistakes what so ever. the 5th and 4th digit from the end were probably correct if we hadn't made a mistake. the 3rd digit from the end was somewhat lucky that it was correct. and the 2nd to last digit was actually 100% a coincidence.
That's why we were so excited when it was correct.
Very helpful. I should have known this stuff, but I didn't quite understand my class on numerical analysis, or whatever it was called. All I remember was ULPs, heh. I was aware of IEEE double precision guard digits, which serve a similar purpose perhaps.
@@michaelbauers8800 It all relates back to the very concept that it is even _possible_ to split up the digits in a calculation like this. Each individual digit only depends on so many of the factors in that long arctan expression, so you only need those nearby factors to get at each specific digit, and they just run the whole scheme in 20 digit increments to optimize for the human ability to work through long division while not getting bogged down in too many digits at once. At the end, you necessarily start losing those last factors that you just haven't calculated to ensure those specific digits, so at any point, if one of those uncalculated factors affects the digit you're investigating, each subsequent digit would likely also be wrong.
@@kindlin Yes, that makes a lot of sense. I was not thinking very clearly about the specifics earlier. There's worse case scenarios for PI, such as those glacially converging series. Like the Leibniz series
A) When adding two numbers, you can get a thousand digit carry or more - write down a thousand nines and then add one to that number, and that one will carry through a thousand digits.
B) What matters is the range of the potential error. Adding seven numbers with the same precision gives an error from zero to seven in the last digit, and subtracting off another seven numbers makes the error between negative seven and seven. For a calculated value ending 176, that means if your only errors come from truncating the fourteen numbers, the true value would have to be in the range 169 to 183, so the penultimate digit would have a decent chance of being correct (5/7 or ~71%) while the last digit is basically random.
C) Of course, if there was additional uncertainty in the fourteen numbers beyond their having been truncated, that would automatically create corresponding uncertainty in the final result, but isn't explained above.
@@rmsgrey theoretically yes, practically no. an n digit carry is only possible with a series of 9s, which is exponentially unlikely (proportionate to 10^-n). It's also something we would have known about since one of the sums would have resulted in trailing 9s. If your result isn't trailing 9s, then you know that you can't have an arbitrarily large carry over error. So practically, no.
This is an excellent demonstration of how productivity doesn't scale linearly with manpower. The final digit countdown was incredible. What a lovely event
The good old Vertical scaling vs Horizontal scaling conundrum.
Also the importance of multithreading
Software engineers like me, usually understand this. Overhead is a real problem. This was great!
Notably 7 days of 200 is only 1400 people-days, or about 4 people-years. Cuz long division is O(n^2), 139 digits is about 4/49s of the way to 700. Which seems almost reasonable based on the whole few decades thing Shanks had. Course the few decades thing is between publications, he published 530 digits, then went back a couple times over a few decades. I have no idea how long the first 530 took (and can't find anyone who does!)
This isn't a video about maths, it's a video about management.
I like how Matt's name badge has a pi symbol instead of the two T's
I'm here for this.
Mapi :з
Map
Short for (Rediculous) Ma(ths) Pi(rson)
Pronounced, of course, as Map Parker
Every year I look forwards to Matt getting pi wrong. Nice to see you guys actually nail it, well done.
Actually, I wouldn't mind them getting it wrong. Matt would always turn it into a fun experience 😎
Technically they'll always get it wrong every year. They could get 1000 digits and still be wrong :)
Nail it? They're only 0 percent of the way there!
He did get it wrong again. Just way less wrong than previous tries. 😀But it looks like everyone had fun doing it.
@@erlandodk goes just to show that they weren't dividing hard enough 😆
I feel like a curse has finally been lifted this year. Here's all of Matt's previous pi calculation attempts:
2013: 3.13834
2015 (part 1): 3.1512
2015 (part 2): 3.128
2016: 3.0418399789...
2017: 3.0523384783...
2018: 3.1415927
2019: 3.11791
2020: 3.1415916785...
2021: 3.875
2022: 3.14159265358868298...
2023: 3.11712
2024: 3.14159265358979323846264338327950288419716939937510582097494459230781640628620899862803482534211706798214808651328230664709384460955058223176
He's approaching the answer, but it doesn't look like he's doing so mathematically (e.g., in the manner of a Taylor sequence)-the oscillations are pretty random.
Classic "off by 1" error
Edit: referring to the video being uploaded the day before Pi Day
they are missing the carry so probably not off by just one.
off by 1/10^140
He posts a day early so teachers can use the video on pi day if they would like!
@@quehabloWell, you’re not wrong. 😂
Classic Parker's calculation
22:45 "They're dividING, but they're not divided." Beautiful Matt, just beautiful haha.
diversity... er.. division is our strength!
22:40
This really was Deus Ex: Mankind Dividing.
United we Divide, Divided we Fall
Matt reading off the result at the end might be my favorite "oner" ever. A single camera shot of a guy reading some numbers for five minutes was far more enjoyable and emotional than anyone could predict. Thanks to the podcast I've been looking forward to this for months, but it exceeded every expectation.
I'm also terrified of what the next attempt in two years' time will look like. "Seventeen thousand people spent a month and three million Post-It notes to calculate pi!"
Wait, that was seriously five minutes? Feels like simultaneously six seconds and an entire movie.
Now that you mention it... yeah. It's up there with "Once In a Lifetime" in Stop Making Sense.
It's nice for the pi day videos to release on March 13 so that teachers have it available to show their students on pi day!
Unless they live REALLY close to the international dateline😅
I wouldn't be worried about Steve multiplying things by 2. I'd be worried that he would cut things in half and cover them with clear acrylic! 😂❤❤
Ah, years ago you would be worried he'd pour it out of a beaker.
@@DukeBG why not both?!
And incrementally shaving people's facial hair throughout the day
I wonder if there is a way to perform fast divisions with water or some clever machinery.
@lolilollolilol7773 There is and it is quite easy. Fill a bucket with the required amount of water, X. Pour it into a trough. Then insert dividers evenly spaced to create Y divisions. The amount of water in each bucket is X/Y
We can't let these darn number enthusiast get away with another one of these projects of pure silliness! We've been working so hard in education to create a system that gets that disgusting joy out of those kids as early as possible and now this guy comes and keeps ruining everything. Won't somebody please think of the children?
I've got nothing against this kind of silliness -- I'm totally here for it -- but it's a shame when educators think this is the way to make math interesting as if what us mathematicians do is sit around doing long calculations.
@@petergerdes1094 If a kid saw this and saw all the people having fun, they may ask why they're doing it in the first place. That line of questioning could lead down a road of "what even is pi?", "why does that formula produce pi?", etc. And that's how a kid could maybe get sucked into the beauty of math
@@mana3109 They could...it could also convince them it's nothing but more stupid calculation which is the primary reason they hate math and don't get anything out of it.
It mostly depends on their teacher, if they have a good teacher who actually understands what's going on and is incentivized to convey math as being something with interesting questions and creativity of course they can use something like this but they could also use a 1000 other more common things too.
It's the teacher who understands why math is interesting and fun that's the commodity in short supply.
Unfortunately, even though most teachers are well meaning they were usually only taught math as a matter of rote calculation as well and when that's true I agree this kind of thing is better than nothing but it doesn't replace actually conveying that math is so much more.
Not to mention the incentivizes in the classroom encourage them not to do that.
Hacker: "Education in this country is a disaster. We're supposed to be preparing children for a working life. Three quarters of the time they're bored stiff."
Sir Humphrey: "Well I should have thought being bored stiff for three-quarters of the time was an excellent preparation for a working life"
(Yes Prime Minister)
The problem is that our education system wasn't designed to actually educate children, it was created to take the children of working class people and turn them into compliant but effective cogs for our corporate overlords. If you want a better education system, we need to get rid of capitalism.
one of the volunteers here, such a great time, i didn't do much calculating but i was there as archiving went from a box to a desk, to a desk in the corner to desk in the corner with people with clipboards, i was on that desk for a time, crazy complicated as things changed around us all, somehow it came together in the end.
I really enjoyed my time here and I'd also like to say, now that i have met the man, that Matt Parker is legend!
Was there a bad bit filtered section? If you returned more than x incorrect digits you were assigned a different station?
@@TheDrewjustforyou no, the day was more informal than that,
At the beginning, one of the council of pi would say 'we need people to do x', i (and a few others) volunteered for those things which were logistical in nature like archive team or copying or doing the mod check
Something about phrases like "this is the times table for Pacific squared" and "I just finished dividing by Doc squared" just absolutely tickles me
I actually misunderstood what was said, thanks for clarifying that.
How did they pick these names? There's no theme!
@@liz4v As noted elsewhere, the theme is "things that are part of a group of seven."
@@alexpotts6520 thanks! Any idea what Doc refers to? And is Giza about the pyramids?
@@liz4v Doc is one of the seven dwarfs.
*The ghost of William Shanks* : Is everyone having fun and a good time?
*Matt* : You know, I think everyone is!
*Shanks* : Well, then they're not dividing hard enough!
🤣
This is my favourite bit so far!
You know, I just realised that its poetoc that 139 is a prime number! I was bit bummed we didnt reach 140 but pi was hand calculated upto prime number of digits which makes it even better
I'm pretty sure the leading 3 isn't included in the number of digits, so arguably 140 total digits was reached, but only 139 after the decimal point.
141 would also be nice because, y'know 3.141
@@jamasa0074Twin prime with 137, 1/137 being one of the most important numbers in the universe.
I think they properly calculated 138 digits (not counting the leading 3.) and got lucky on the next one, that didn't have a carry-in from the next bunch down that did not get computed.
@@JohnDlugosz on my reply here, i have arranged the numbers (in group of 10 digits ) so knowing the last didgit was wrong even you can see that excluding the 3, its correct till 139th decimal
With the number of digits this team calculated this year...
you could construct a circle,
1 million times the size of the known universe,
28 million billion light years in diameter,
to the precision of 1 hydrogen atom.
That deserves the applause.
Suspicious.
@@deltalima6703this is just 10^139.
@@deltalima6703 Mathematicians have estimated that an approximation of pi to 39 digits is sufficient for most cosmological calculations - accurate enough to calculate the circumference of the observable universe to within the diameter of a single hydrogen atom.
For Matt Parker, this is a genuine acheivement. 139 digits. Good luck getting further next time Matt!
Maybe we can try harder than six 😤
Like 10 times the previous best? He had a lot more help of course. The error checking was critical of course
I think this system is a pretty good baseline-also, as people have pointed out, if you scale the number of man-hours up to what Shanks did, you'd far surpass his record. Maybe two years from now he'll have found a Taylor series that distributes the workload even more, or simply scaled up the operation with the current error-checking scheme. Honestly, if I were to have the resources to spend a week in London, I'd love to come out in 2026 and help them break the record.
That’s outstanding! Last time you had a great open-loop process but no error detection and correction. Adding in the validation stages _dramatically_ improved your productivity. Now the challenge is to optimise the process without sacrificing reliability, and go for it again in 2025 🤗
Right on! I'm confident they can figure out a clever way to be substantially better next time ✌
No, 2025 is for another ridiculous way of doing it. It's just the even numbered years they do it by hand.
2026. It's ever two years.
If they had that archive group working well, maybe could they reuse the sheets for next time?
I'd expect more resources to volunteer next time. They may hit some novel scaling problems unrelated to the calculation aspect from just having so many people concentrated for a week.
JAMES GRIME CAMEO!!!!!!
We love James Grime!
Steve Mould with his 2Pi cameo!
Ayliean was there too!😊
singing banana!
A veritable who's who of the people the CZcams Maths Algorithm has shown me.
Atleast it makes sense that the last digit was the one that was off, there was probably a carry over from the next digit you didnt have
Yeah, and they expected more digits to be off (as you can probably grasp from their suspense in the video) as the carry-overs usually come in as multiple of digits (in this kind of calculation), not just one.
@@uNiels_Heart well yeah, I'm aware. What I'm saying is the one digit that WAS off is (probably) not their fault
Had they continued 2 more digits, the current final digit would probably be right, and the new final digit would be wrong
(In theory)
(Is what I was talking about)
Yeah exaclty. Since there are more then 50 numbers or so added up (We needed to go up to arctan order 40 or so for the Monday arctan, which means already 20 odd terms for this one arctan) there could have been a two digit carry over. Therefore we didn't expect the 139th digit to be correct as well
going from 1 person 6 digits to 5 people 5 digits really shows how "if you wanna go fast go alone. If you wanna go far, go together"
This is the first time in 40 years that a computer was comprised of a bunch of wired together logic in one giant building.
Isn't a server room, or a supercomputer, a bunch of wired together logic in one giant building?
really puts into perspective the sheer difference computers made.
a single computer (and a low end one too) could do all of this in probably less than a second.
Yea actually low end computers are so fast; the new ones only feel slower because software developers are lazy and waste the extra speed. Lots of bloat in modern software, both in the operating systems and in the software we run!
@@0iaIt's not about laziness, it's about market timing... A slowish software that gets released beats the incredibly fast software that's still in development
@@KleyguerthI suppose my idea of "laziness" is an optimistic idea that people would want to only release fast software, and you're probably right.
@@0ia bloat doesn't come from laziness, it comes from a lot of hard work adding features that you'll never know exist, or just devour because it makes for a better experience. We hit "fast enough", then optimize for things other than raw speed nobody can notice.
@@0ia some stats:
I have an intel I3 1115G4, which has 844 GFLOPS of GPU power
a high end RTX has about 20x or more power
the most powerful computers are millions of times faster than mine
the most powerful computer can do 1 100 000 000 000 000 000 32 bit operations a second.
there are 8 billion humans on earth
if all of us pooled together solely to do mathematics, we would be comparable to the most powerful computer *40 years ago*
19:17 Mat saw Shanks' redemption.
Is that a pun on Shawshank Redemption? 😆
@@uNiels_Heart Took me a while of puzzling to make a good one, but yes it was.
To the punitentiary with you!
Congratulations to matt parker for breaking 2018's record of 14 digits of pi correct( that pi day 2018 pi record held for 6 YEARS)
The next time you do this you're gonna blow the record out of the water. That optimization, checking in parallel with the next set of calculations starting is big.
No, that was a scheduling optimizing for completion of 140 digits with a given amount of resources not a throughput optimization. Had they not given out that speculative work as the parts wound down then people would have been idle waiting for the calculations to be confirmed.
Unless they get a bunch more people to participate next time. In that case they may need to do this to have enough work for all of those people.
I'm intrigued by the lack of mention of how "resource allocation" was done.
It's a fact that dividing by those very large values is a very specialized task; especially when alternative tasks of dividing by the small odd numbers, performing the modularity checks, and adding are also required.
I'd wager that it is **difficult** to accurately predict in advance who performs best at the most challenging long divisions. Introducing a task-orientation for resource allocation, based on observed error rates, I believe has huge potential for reducing error rate. That's the real bottleneck. Every incorrect return now ties up multiple more capable (for that class of calculation) resources AND delays subsequent work.
A detailed write up of the plan and execution would really interesting.
I never thought I'd get genuinely emotional watching this video lol but at the end I almost had tears in my eyes. Theres just something so beautiful about when humanity can collaborate on something just to prove to themselves they can do it. Why can't the whole world just be like the group of people in that auditorium?
Happy Early Pi Day!
I cried during the final readout. No shame. We love humans here
This is the first time I got emotional because someone was just reading numbers ...
Great things can happen when people put their minds to it. Silly things can happen too! Hooray!
@@Pouckie90I actually did get emotional once before someone was reading out numbers!
Dad was not pleased with my marks.
Because the world needs houses and highways, not a bunch of sweaty nerds with zero muscle mass and blue hair
So you practically turned Pi into a big social happening that brings people together and makes everyone happy.
Nice work, man! 🙂
"Awfully drawn 8"
"Yep. It's alright on here"
That was gold.
Please note down the new value of pi:
3.13
*13.3
This makes me suspect a legislature was involved!
Parker Pi
@@nathenielthe less known cousin of the Parker square, the Parker circle!
google is giving me a pi animation whenever I use the calculator, I was so confused if maybe pi day had changed or sth
23:17 to continue Emma's computer analogy, this is where they added speculative execution and increased the performance.
Was great fun to participate in this event. Thanks for organising!
I thought it was a great analogy. I started remembering the work that went into my toy CPU ( built in a logic gate simulator.) Back in the old days, there were no pipelines in CPUs. Now they rely on pipelines. And of course super computers have relied on pipelines since the beginning for all I know; Cray 1 used "vector processing" which was pipelining. I once parallelized an algorithm and it was slower. So the comment made about it not linearizing was of course very relevant. At some point, throwing more resources at a problem might even hurt. With lessons learned here though, I am sure they could scale to more people. Funding might be an issue.
I absolutely loved being part of this Tau / 2 calculation!
that guy giving rating of 3.14 gets me...
I knew it was coming as soon as he said "0 to 5" and it still made me laugh, the cut off was great too because you can tell he went on and said more digits lol
Comedic timing was perfect. I actually didn't see it coming, somehow.
0:49 that cut, 10/10 funny
I was apart of the calculation and it was some of the most fun I've had in years. So many really fun people to talk to. I definitely recommend joining the next ridiculous maths project Matt does next
I've never been so excited by someone reading digits of pi 😮
19:28 this looks like a Monty Python skit
Almost certainly inspired by The Holy Grail. I would be surprised to find out it wasn't.
I like how the 7 arctan equations each have a name which is a member of a group of 7. But there are a few that don't make as much sense to me, hopefully they can be explained:
1. Monday (obviously 7 days in a week)
2. Shimada (I'm not sure on this one, but it's a city in Japan)
3. Asia (one of 7 continents)
4. Greed (one of the 7 deadly sins)
5. Pacific (another puzzling one, there aren't 7 oceans as far as I'm aware)
6. Giza (a city in Egypt, known for its pyramids, but I don't see a connection to 7 here, either)
7 Doc (the odd-named dwarf of the 7 dwarfs from Snow White)
If there are explanations for Shimada, Pacific, and Giza, I'd love to hear them. (Maybe it's in the video and I haven't seen it yet, I paused to think about the names)
shimada is one of the “seven samurai”, pacific is one of the “seven seas”, and giza is one of the seven wonders of the world!
The Pacific one is probably in reference to the "seven seas", which is a term dating back thousands of years. The actual list is the Arctic, North Atlantic, South Atlantic, North Pacific, South Pacific, Indian, and Southern oceans apparently. Honestly it's more a narrative thing than a terminology thing so i wouldn't stress too much about it
5: sailing the 7 seas is the stock thing a pirate does
6: pyramids of Giza are one of the 7 wonders of the ancient world
2. Kanbei Shimada is the leader of the Seven Samurai
5. The Seven Seas is certainly a concept, even if out of date.
6. Giza is one of the Seven Wonders of the World
Giza is one of the Seven Wonders of the World.
That one guy is right, comparing this scheme to modern computer architecture is neat! The final day of sending out dependent sheets before finishing checking is similar in concept to speculative execution (related to the meltdown bug, sending out future instructions before finishing exception checking); lookup tables are similar to caches; and having figured out the dependencies between different calculations achieves the same goal as register renaming. I genuinely don't think there's any big ideas that apply from computer architecture that got missed!
27:13
That person in the crowd with an emotional support Blahaj😂
They've calculated pi to the maximum accuracy you could fit into a tweet back in the day
200 people over 7 days still can't beat 1 person over 40 years. That one person was phenomenal!
On a per digit rate the new group was twice as fast, 10 person-days per digit, vs the original 27 person-days.
to be fair
200 people *7 days = 1400 peopledays
which is less than 5 years for a single person
The problem is not linear in digits. O(n^2) at least.
@@kjdude8765I assume that the original guy wasn’t spending 8? Hours per day on this. While we don’t really know how many man hours/digit he spent, this year’s efforts can provide a benchmark for future techniques
Well, I mean, 8 billion people over 100 years can't beat 1 person over 40 years without computers!
FINALLY, Matt you have posted a PI day video on PI day here in NZ
❤️
Much appreciated
"Honestly, just felt kind of nerdy." A good cap-off.
I love how in the process of doing this, Matt has independently discovered one of the reasons why writing multithreaded computer programs is so difficult!
Please have *Pi Fisher* 36:09 back for the next attempt. I'm sure their efforts were invaluable
We need to know more about this person: Are you actually named Pi? Are you a mathematician? Are your parents mathematicians? Do you have three kids born in march, january and april? Are your siblings named e and i?
@@malterichert2927It’s their evil twin named Tau that you need to watch out for.
Unless Pi Fisher answers himself, I can say that he is a very nice person! I was on the π-brary team (I'm talking at 7:00) and he was the head π-brarian. His name indeed is Pi (I do believe it is greek) and he helped us fish for pi. You can see him with brown hair and beard wearing a face mask in the π-brary. We definitely could not have done it without his organizational skills 🫡
@@BRORIGIN A wonderful case of nominative determinism.
Agreed!
It's been far too long since I've seen James Grimes
singingbanana ftw
This team is chock full of big names: Sophie Maclean, Katie Steckles, James Grime, Matt Scroggs, Ayliean McDonald... These people are my celebrities.
was so much fun to help out, Matt!! Thanks so much to you and all the team for organising this mathematical madness, can't wait to have another crack in a couple years time!! - Freddie :)
amazing!! There's no reason next year you couldn't pick up at 140 and continue! MISTER Shanks didn't do it all in one go! So you guys can do the next 139 next year!
Before watching I was like: "Yeah maybe they'll get 12 digits this time"
lol. Matt's past results were no where near as close. But he had a lot of help this time, and a lot of error checking.
I was there and it was every bit as fun as it shows in the video. Thanks for the opportunity!
I love the explanation of 16:40 because the parallel between those people and a processor was also my first thought!
I love how this calculation is XKCD 1417 compliant.
I was not expecting the emotional swings.
Fantastic editing.
when I was a kid and I heard about people or computers figuring out what the digits of pi were via calculations, I always assumed that just meant measuring a circle's diameter and circumference and then doing a single long division problem of c/d. I didn't really understand back then that you wouldn't be able to accurately measure the circle to enough significant digits to get even close to a hundred digits of pi.
WOW!
I genuinely did not expect this video to bing a tear to my eye, but I was taken at the end past the 100th digit.
Just brilliant!
The last digit being off was the icing on the cake because, to me, it showed the optimal amount of error correction was in place.
Well done to everyone involved.
The last digit was not a mistake. It could only have been right by pure luck, because it's the result of adding 7 terms together, and there was likely a carry/borrow from the 141st and 142nd digits, which they didn't have time to calculate.
3:12 xkcd 1417
oh my god it is
Well spotted. I missed it.
Here's what each collection of seven:
1. Days of the week
2. Seven Samurai (I had to Google this one)
3. Continents
4. Seven deadly sins
5. Seven seas
6. Seven wonders of the ancient world
7. Seven dwarves
Edit: The Pacific one is a bit suspect. The expression "the seven seas" is ambiguous, and the Pacific is not on all lists and on some is divided into the North and South Pacific.
@@siener It's actually remarkable how culturally widespread the notion of "the seven seas" is despite a *total* lack of consensus as to what those seas are:
en.wikipedia.org/wiki/Seven_Seas
I can’t believe I cried at the accomplishment. Hard work and good attitude as well as a little bit of luck made it so incredible.
This video needs to be mandatory viewing for any self proclaimed developer that demands your software be using more multi threading to make it faster.
One of the greatest meetings of Nerds ever seen. Very close to a Star Trek convention. Love it.
Trek.
😰
If I understand the algorithm correctly the team should be able to use these results as a starting point for further calculations. If they can average 150 digits per year they will break the record in three more years.
Only if they kept intermediate results, and even then I doubt it.
I don't think we will, the challenge is really to start from scratch - and anyway, as @@JohnSmith-zq9mo says, there's a lot of intermediate results to carry over
Yeah I think everything was archived so that we could continue next time. However the effort grows with the number of digits squared. 300 digits would than take 4 years in total, and 600 digits 16 years in total.
But there are ways to improve the process since the limit were the chain of dependencies and not the calculating power
It was incredible to have been a part of this, thanks to all the team for organising it. See you all in 2026 for the next attempt when we'll get even further! 💜
Classic divide and "concur" algorithm! I would love to participate in one of these impossible events some year, geography permitting!
Of the very many funny moments in this video, calling Steve Mould an "Independant observer" was the one that made me laugh the most.
Astonishing work nonetheless - congratulations all around!
I love the tongue in cheek news story presentation, the ominous juxtaposition cuts to each day, and the final countdown gave me shivers 🎉 With this system figured out, I think you might have the world record in the bag next year! I'm absolutely signing up to help however I can
I just know that there's at least one person in there that knew pi to like 300 digits by heart and already knew that there's was an error and was probably dying inside in the final digits check
16:21 - i agree with Emma. Ironically, the most interesting part of the video for me was the logistics of coordinating everything and how each dedicated-function desk all fed-together to make a full system for calculating pi.
I think the really interesting thing is how the layout of the room and the tasks each table was assigned, if very reminiscent of the diagram outlining processing cores in a CPU.
Such an awesome experience, I was there for nearly the whole thing. Surrounded by so many nice people just having fun, no stress, just good vibes and meeting the youtubers I'd watched for so many years and making new friends. Thank you so much for organising this Matt, Katie, Sophie and the entire standup maths team!
The real accomplishment here is implementing a human computer with error correction and parallel processing. Badass
Very happy we got a Parker Quotient where the final value had one digit wrong. Congratulations Matt!
Not sure why but I have grown to look forward to Pi day because of how you come up with methods new to me. It is so interesting!
You’ve captured my attention ever since the domino computer. Conceptually beautiful to me even if there were hiccups with that video and I’ve been with you for the Pi ever since
Friendly reminder that 22/7 is a better Pi Day than 3.14. It can be observed in more countries' date format, and it's actually a closer approximation of Pi than 3.14.
YYYY-MM-DD gang disagrees
It was so fun to be a part of this process, was there for 2 days and thoroughly enjoyed both! Great to meet lots of different Maths nerds along the way. I shall certianly be joining the next attempt whereever amd whenever that may be!!! Cheers Matt for signing my calculator!
Would've been cool to have all the different teams wearing hats with bright LEDs of specific color, and then taking a very long exposure image that shows the literal flow of data around the room.
25:55-33:24
How it feels to give the final readout for Forget Me Not in Keep Talking And Nobody Explodes when you're only one mistake away from detonation
11:01 "When a complete sheet was given to the check-in desk, they would do some initial checks to make sure it had been filled incorrectly."
Well there's your problem! You should've made sure it had been filled correctly, not incorrectly.
Haha, I was on the check-in desk and did a very basic mod 10 check upon hand in, hopefully I helped rather than hindered!
The narrator said "it had been filled in correctly" you missed the space
@@Stroopwafe1 that's the joke
I am so happy and emotional at the same time. I always feel this way when I see human endeavor. One of the best pieces of CZcams. Thanks Matt. Thanks team.
watching the video a bit late, but the "it was not going well" in the first couple seconds elicited a verbal "oh no!" followed by a grin. I already know this is going to be another wonderful video
32:59 what a wonderful moment!
I'm still rounding it to 4
*Engineering has entered the chat*
This reminds me of The 3 Body Problem and the way Trisolaris calculated stuff in the beginning. :D
Exactly lol 😂
That was the first thing that came to my mind. I assume the Trisolarans were better organised and more motivated though. And even then it'd have been painfully slow work
The fact that the *last* digit was the only wrong one suggests that the entire calculation was, in the end, technically correct.
That sure was it, but the procedure has lots of interdependencies which would make it a hard-to-parallelize task even for a digital computer. Amazing feat for humans!
It shows me that not only did Matt finally learn from a decade of previous failures and institute error-checking methods, but that the methods were highly effective.
3:20 I had a hearty chuckle at those team names. 7 references to groups of 7. Well played.