Randall Munroe: Comics that ask "what if?"

Recorded atMarch 20, 2014
Duration (min:sec)09:12
Video TypeTED Stage Talk
Words per minute217.56 very fast
Readability (FK)58.98 easy
SpeakerRandall Munroe
CountryUnited States of America
Occupationprogrammer, writer, physicist, engineer, geohasher
DescriptionAmerican cartoonist, author and engineer

Official TED page for this talk


Web cartoonist Randall Munroe answers simple what-if questions ("what if you hit a baseball moving at the speed of light?") using math, physics, logic and deadpan humor. In this charming talk, a reader's question about Google's data warehouse leads Munroe down a circuitous path to a hilariously over-detailed answer — in which, shhh, you might actually learn something.

Text Highlight (experimental)
100:12 So, I have a feature on my website where every week
200:15 people submit hypothetical questions
300:17 for me to answer,
400:18 and I try to answer them using math, science
500:21 and comics.
600:22 So for example, one person asked,
700:25 what would happen if you tried to hit a baseball
800:26 pitched at 90 percent of the speed of light?
900:29 So I did some calculations.
1000:32 Now, normally, when an object flies through the air,
1100:34 the air will flow around the object,
1200:36 but in this case, the ball would be going so fast
1300:37 that the air molecules wouldn't have time
1400:39 to move out of the way.
1500:41 The ball would smash right into and through them,
1600:44 and the collisions with these air molecules
1700:45 would knock away the nitrogen,
1800:47 carbon and hydrogen from the ball,
1900:49 fragmenting it off into tiny particles,
2000:51 and also triggering waves of thermonuclear fusion
2100:53 in the air around it.
2200:55 This would result in a flood of x-rays
2300:57 that would spread out in a bubble
2400:59 along with exotic particles,
2501:00 plasma inside, centered on the pitcher's mound,
2601:03 and that would move away from the pitcher's mound
2701:07 slightly faster than the ball.
2801:09 Now at this point, about 30 nanoseconds in,
2901:12 the home plate is far enough away
3001:14 that light hasn't had time to reach it,
3101:16 which means the batter
3201:17 still sees the pitcher about to throw
3301:20 and has no idea that anything is wrong.
3401:22 (Laughter)
3501:24 Now, after 70 nanoseconds,
3601:27 the ball will reach home plate,
3701:28 or at least the cloud of expanding plasma
3801:31 that used to be the ball,
3901:33 and it will engulf the bat and the batter
4001:37 and the plate and the catcher and the umpire
4101:40 and start disintegrating them all
4201:43 as it also starts to carry them backward
4301:46 through the backstop, which also starts to disintegrate.
4401:50 So if you were watching this whole thing
4501:51 from a hill,
4601:53 ideally, far away,
4701:56 what you'd see is a bright flash of light
4801:57 that would fade over a few seconds,
4901:59 followed by a blast wave spreading out,
5002:01 shredding trees and houses
5102:04 as it moves away from the stadium,
5202:06 and then eventually a mushroom cloud
5302:09 rising up over the ruined city. (Laughter)
5402:12 So the Major League Baseball rules
5502:13 are a little bit hazy,
5602:16 but — (Laughter) — under rule 6.02 and 5.09,
5702:20 I think that in this situation,
5802:22 the batter would be considered hit by pitch
5902:25 and would be eligible to take first base,
6002:27 if it still existed.
6102:30 So this is the kind of question I answer,
6202:32 and I get people writing in with
6302:34 a lot of other strange questions.
6402:36 I've had someone write and say,
6502:39 scientifically speaking, what is the best
6602:41 and fastest way to hide a body?
6702:43 Can you do this one soon?
6802:45 And I had someone write in,
6902:47 I've had people write in about,
7002:49 can you prove whether or not you can find love again
7102:51 after your heart's broken?
7202:53 And I've had people send in
7302:54 what are clearly homework questions
7402:56 they're trying to get me to do for them.
7503:00 But one week, a couple months ago,
7603:03 I got a question that was actually about Google.
7703:06 If all digital data in the world were stored on punch cards,
7803:09 how big would Google's data warehouse be?
7903:12 Now, Google's pretty secretive about their operations,
8003:15 so no one really knows how much data Google has,
8103:18 and in fact, no one really knows how many data centers Google has,
8203:20 except people at Google itself.
8303:23 And I've tried, I've met them a few times,
8403:24 tried asking them, and they aren't revealing anything.
8503:29 So I decided to try to figure this out myself.
8603:32 There are a few things that I looked at here.
8703:34 I started with money.
8803:36 Google has to reveal how much they spend,
8903:38 in general, and that lets you put some caps
9003:40 on how many data centers could they be building,
9103:44 because a big data center costs a certain amount of money.
9203:46 And you can also then put a cap on
9303:49 how much of the world hard drive market are they taking up,
9403:51 which turns out, it's pretty sizable.
9503:53 I read a calculation at one point,
9603:55 I think Google has a drive failure
9703:56 about every minute or two,
9804:00 and they just throw out the hard drive
9904:01 and swap in a new one.
10004:02 So they go through a huge number of them.
10104:05 And so by looking at money,
10204:06 you can get an idea of how many of these centers they have.
10304:08 You can also look at power.
10404:10 You can look at how much electricity they need,
10504:14 because you need a certain amount of electricity to run the servers,
10604:16 and Google is more efficient than most,
10704:18 but they still have some basic requirements,
10804:21 and that lets you put a limit
10904:23 on the number of servers that they have.
11004:25 You can also look at square footage and see
11104:29 of the data centers that you know,
11204:31 how big are they?
11304:32 How much room is that?
11404:33 How many server racks could you fit in there?
11504:35 And for some data centers,
11604:37 you might get two of these pieces of information.
11704:39 You know how much they spent,
11804:41 and they also, say, because they had to contract
11904:43 with the local government
12004:44 to get the power provided,
12104:46 you might know what they made a deal to buy,
12204:49 so you know how much power it takes.
12304:50 Then you can look at the ratios of those numbers,
12404:53 and figure out for a data center
12504:54 where you don't have that information,
12604:56 you can figure out,
12704:57 but maybe you only have one of those,
12804:59 you know the square footage, then you could figure out
12905:01 well, maybe the power is proportional.
13005:03 And you can do this same thing with a lot of different quantities,
13105:05 you know, with guesses about the total amount of storage,
13205:08 the number of servers, the number of drives per server,
13305:10 and in each case using what you know
13405:13 to come up with a model that narrows down
13505:16 your guesses for the things that you don't know.
13605:18 It's sort of circling around the number you're trying to get.
13705:20 And this is a lot of fun.
13805:23 The math is not all that advanced,
13905:25 and really it's like nothing more than
14005:28 solving a sudoku puzzle.
14105:30 So what I did, I went through all of this information,
14205:35 spent a day or two researching.
14305:37 And there are some things I didn't look at.
14405:39 You could always look at the Google
14505:42 recruitment messages that they post.
14605:44 That gives you an idea of where they have people.
14705:46 Sometimes, when people visit a data center,
14805:47 they'll take a cell-cam photo and post it,
14905:49 and they aren't supposed to,
15005:51 but you can learn things about their hardware that way.
15105:54 And in fact, you can just look at pizza delivery drivers.
15205:56 Turns out, they know where all the Google data centers are,
15305:59 at least the ones that have people in them.
15406:02 But I came up with my estimate,
15506:04 which I felt pretty good about,
15606:06 that was about 10 exabytes of data
15706:09 across all of Google's operations,
15806:11 and then another maybe five exabytes or so
15906:15 of offline storage in tape drives,
16006:17 which it turns out Google is
16106:18 about the world's largest consumer of.
16206:21 So I came up with this estimate, and this is
16306:25 a staggering amount of data.
16406:26 It's quite a bit more than any other organization
16506:29 in the world has, as far as we know.
16606:30 There's a couple of other contenders,
16706:32 especially everyone always thinks of the NSA.
16806:35 But using some of these same methods,
16906:36 we can look at the NSA's data centers,
17006:38 and figure out, you know, we don't know what's going on there,
17106:40 but it's pretty clear that their operation
17206:43 is not the size of Google's.
17306:44 Adding all of this up, I came up with
17406:46 the other thing that we can answer, which is,
17506:48 how many punch cards would this take?
17606:50 And so a punch card can hold
17706:53 about 80 characters,
17806:55 and you can fit about 2,000 or so cards into a box,
17906:58 and you put them in, say,
18007:00 my home region of New England,
18107:02 it would cover the entire region
18207:04 up to a depth of a little less than five kilometers,
18307:08 which is about three times deeper
18407:09 than the glaciers during the last ice age
18507:11 about 20,000 years ago.
18607:14 So this is impractical, but I think
18707:16 that's about the best answer I could come up with.
18807:19 And I posted it on my website. I wrote it up.
18907:21 And I didn't expect to get an answer from Google,
19007:25 because of course they've been so secretive,
19107:26 they didn't answer of my questions,
19207:28 and so I just put it up and said,
19307:29 well, I guess we'll never know.
19407:31 But then a little while later
19507:33 I got a message, a couple weeks later, from Google,
19607:35 saying, hey, someone here has an envelope for you.
19707:39 So I go and get it, open it up,
19807:42 and it's punch cards. (Laughter)
19907:44 Google-branded punch cards.
20007:47 And on these punch cards, there are a bunch of holes,
20107:50 and I said, thank you, thank you,
20207:52 okay, so what's on here?
20307:53 So I get some software and start reading it,
20407:55 and scan them, and it turns out
20507:57 it's a puzzle.
20607:58 There's a bunch of code,
20708:00 and I get some friends to help,
20808:01 and we crack the code, and then inside that is another code,
20908:04 and then there are some equations,
21008:05 and then we solve those equations,
21108:06 and then finally out pops a message from Google
21208:10 which is their official answer to my article,
21308:13 and it said, "No comment."
21408:15 (Laughter) (Applause)
21508:26 And I love calculating these kinds of things,
21608:29 and it's not that I love doing the math.
21708:31 I do a lot of math,
21808:32 but I don't really like math for its own sake.
21908:35 What I love is that it lets you take
22008:37 some things that you know,
22108:39 and just by moving symbols around on a piece of paper,
22208:43 find out something that you didn't know
22308:45 that's very surprising.
22408:47 And I have a lot of stupid questions,
22508:49 and I love that math gives the power
22608:51 to answer them sometimes.
22708:54 And sometimes not.
22808:55 This is a question I got from a reader,
22908:58 an anonymous reader,
23008:59 and the subject line just said, "Urgent,"
23109:01 and this was the entire email:
23209:03 "If people had wheels and could fly,
23309:06 how would we differentiate them from airplanes?"
23409:09 Urgent. (Laughter)
23509:11 And I think there are some questions
23609:15 that math just cannot answer.
23709:17 Thank you.
23809:20 (Applause)