The Greatest Maths Mistakes | Matt Parker | Talks at Google

When math goes wrong, things can get expensive. Or absolutely hilarious. For this talk we invited YouTube personality (Numberphile, standupmaths), math communicator, comedian, and one third of the Festival of the Spoken Nerd, Matt Parker, to share his favorite math mistakes from his new UK #1 bestseller, "Humble Pi - A Comedy of Maths Errors".

Matt exposes errors on the Two Pound Coin, very specific rules for trains operating in Switzerland, and how simple unit conversion slip ups can cost billions of dollars. He also discusses the infamous 256th level of Pac-Man and answers audience questions about more hilarious mathematical failures.

[bbmikej]

A note about how aviation looks at mistakes in the US: we have a great program called the ASAP program where if you generally make a mistake (ie not intentionally breaking a rule or being negligent), you can self report to the FAA. They then look at it with representatives of the company and the workers union to determine if it was actually a mistake without a name attached to the report to keep it anonymous. If it was determined to be a mistake and the ASAP is accepted, the FAA cannot take action against the person who made the mistake. The number of people willing to admit to mistakes has increased since implementation and industry wide training has gotten better because of this.

[blackpenredpen]

What an honor to be mentioned by you, Matt! Thank you!

[Stubones]

What about the math error at the Savannah River nuclear plant that was regularly cracking the pressure vessels. Everyone kept checking the materials used, and engineering practices, but no one verified the actual operating pressures of the vessels, and it turns out that they ran for 40 years at 115% power levels because some engineer had a slide-rule error during its initial construction documentation.

[ChrisBigBad]

The train thing.. I call that "Fix by post-it" Like when you know that a program crashes if you click a button, but instead of fixing the program, the boss tells you to send an email to everyone, instructing them to put up post-its to their screens saying "don't press that button"

[Jkirek_]

"It was an orbiter, not a lander"

I think the craft itself disagrees with you there, Matt. It definitely landed, albeit at a high velocity downwards.

[samreid]

28:05 bit of a correction: that happened to the USS Yorktown, which at the time was a Ticonderoga class cruiser. The story changes a bit every time I hear it but essentially some sailor was working on the radar, typing in gates so that it doesn’t flag every seagull or cloud as a target. One of the gates didn’t apply in this situation so he put a zero in it. If he had left it blank or written null or anything else, the computer had a check to filter out non number inputs. However, the check read 0 as a valid number to enter into its equations, at which point it divides by zero and the system, running Windows NT, freaks out and shuts down. This computer not only ran the radar, but also propulsion and navigation. The ship ended up dead in the water for about three hours

I'm at peace when Matt's head line up with the monitors edge in the background.

[gejyspa]

In re: spacecraft -- About 35 years ago, while in college, I was working in a cooperative education job for eight months at RCA Astro-Electronics. The department I was working in was updating testing software created for the TIROS-N satellites for the Advanced Tiros N satellite, which had more instruments on it. The point of the testing software was to make sure that the instruments wouldn't shift too much during launch. To do this, they would measure with lasers the precise location of the instruments on the spacecraft put on a shaker table simulating launch, vibrate and then remeasure the positions, and put all the figures into this program to check if it was in specs. Well, on modifying the program (in FORTRAN, btw), I notice that it had a subroutine to calculate the difference between signed numbers. And here's how it did it: First it took the absolute value of the difference of the _absolute value_ of the two inputs, and then if the inputs had different signs, put a minus in front of the result. Essentially Z=ABS(ABS(X)-ABS(Y));If ((X<0) and (Y>0) ) or ((X>0) and (Y<0)) Z=-Z Well, hang on! If X and Y were different signs, those results were just wrong! And FORTRAN, like all computer languages, has no trouble with signed arithmetic. It just simply should have been Z=ABS(X-Y) I blinked a couple of times, showed it to my boss (cause I was just a college student, and wanted to make sure I wasn't misunderstanding something) And he agreed it was just totally wrong. So while it may never have happened that something crossed over an axis during the shake, I'm glad I was able to potentially save a multimillion dollar spacecraft from potentially being useless.

[hakarthemage]

imagine going to your tutor and saying "Sorry my PHD exploded"

One of the more funny long nerd talks I have seen in full in a long time. What disturbs me is that companies are reluctant to give away how mistakes were made. I think companies easily could obfuscate sensitive portions but still give away generic parts of the error. That would help not only the company itself but human beings as such. But sadly, often cause of errors are not even communicated within the company itself :-(

[HotelPapa]

"That's such a Switzerland solution to the problem"
So true it hurts.

[dhpbear]

30:45 - There's a more acute example than the 'millennium bug':


Loudspeaker driver manufacturers would rubber-stamp each of their drivers with the year and month of manufacture. They encoded the year using ONE digit! One couldn't tell if a certain speaker was made in 1955 or 1965!

[kidcrow]

13:46 the lone star joke he threw in there was brilliant.

[tommykarrick]

I love the long build up to the guy just giggling
“They killed an elephant with cocaine”

[kyoopihd]

"Pounds... bushel, or something..." xD

[thebonesaw..]

I absolutely love learning about unforeseen consequences. I'm not sure how varied a maths version of that would seem from the more general "maths mistakes", but it sounds like a book I'd love to read.

[santimonto]

If Matt does write a sequel to this book, it better be called "Humble Tau", or I'll be triggered.

[mailleweaver]

"...until you notice that one semicolon."
Spot on. Just a couple weeks ago I wrote a script to do something in the cabinet design software I work with, and I couldn't figure out why this one line didn't seem to work. Luckily (?) it is a very forgiving system and let the rest of the script run even with a problem in that line so I didn't notice it for a while. Once I did figure out the line wasn't working I spent hours poking at it: double checking the variables that fed into it, looking for holes in the logic of its formula, rewriting the same formula in different ways, and even trying completely different approaches to accomplish the same thing, but nothing worked. I eventually gave up on it and continued on with the rest of the code. Later on my eyes happened to pass over that line and notice that the comment at the end of the line didn't have the one semicolon in front of it that it was supposed to. Semicolon added, boom it works.

In a way it was good that happened, because one of the alternate approaches I developed trying to fix it was a lot more graceful and made it easier to add another feature later. It was still stupidly frustrating, though.

[lunasophia]

"I'll be around." No, Matt, you'll be a Square.

Not apologising for that.

[sagov]

"the plate's undone by a lone star", completely missed by the audience