back to article That's Numberwang! Google Cloud staffer breaks record for most accurate Pi calculation

Emma Haruka Iwao, a developer advocate at Google Cloud, has celebrated Pi Day (3/14) by setting a new Guinness World Record for calculations of the beautiful mathematical constant, reaching a number with more than 31.4 trillion (ha!) digits. Researchers have been competing to calculate the most digits of Pi for years – often …

  1. DaLo

    How I wish I could calculate PI, really tried but failed

    1. Timto

      Instead of believe what the "Scientists" tell us you should question and experiment yourself.

      Drawing a circle with chalk and using grid paper to work out the area I've found pi is exactly 3.

      1. boardslider

        I’ve often wondered in a Douglas Adams kind of way if that is where we all went wrongn when the aliens finally make contact they will be puzzled over why we invented such a contrary system of maths where Pi is not equal to 3...

        1. TheInnerPartSystem

          More likely the aliens would wonder why we use integers and base 10, instead of base Pi.

      2. FractalZ

        Indiana once almost declared Pi = 3.2 , which explains many things.

        The State Legislature of Indiana once spent some time debating whether or not Pi should be defined as 3.2 as a matter of law. See: https://en.wikipedia.org/wiki/Indiana_Pi_Bill .

  2. Snow Hill Island

    Pi day

    With apologies to the large part of the world who don't write dates the way we do in the UK:-

    "And now let's rotate the date format..."

    1. anthonyhegedus Silver badge

      Re: Pi day

      "And now let's rotate the date format..."

      I can't stop visualising Mitchell rotating his fingers when he says that!

    2. notamole

      Re: Pi day

      The vast, vast majority of the world uses day-month-year. Only America uses month-day-year.

      https://en.wikipedia.org/wiki/Date_format_by_country

      1. Tom 7

        Re: Pi day

        And 22/7 is a lot more accurate!

        1. Iain 14
          Thumb Up

          Re: Pi day

          Probably showing my age, but I always knew Pi Day to be the 22nd of July.

          It certainly worked for Mel Croucher, anyway...

        2. Arthur the cat Silver badge
          Boffin

          Re: Pi day

          And 22/7 is a lot more accurate!

          Try 355/113. Known to the Chinese 15 centuries ago, and accurate to single precision f.p.

        3. notamole

          Re: Pi day

          It's also an actual approximated value, whereas 3.14 is just truncated. Why not 3.1?

          1. gnasher729 Silver badge

            Re: Pi day

            22/7 and 335/113!are both abbreviations, not of the decimal representation but if the continued fraction.

          2. Aladdin Sane

            Re: Pi day

            Why not make it 3 and use that to sort letters?

      2. Jay Lenovo

        Re: Pi day

        To be fair, Pi is an irrational number, so an irrational date format seems apropos to celebrate the day.

        1. DontFeedTheTrolls
          Childcatcher

          Re: Pi day

          So 29/3? That's an irrational date

          (Shameless #Brexit #Omnishambles rant)

      3. This post has been deleted by its author

      4. Cardinal
        Holmes

        Re: Pi day

        @notamole

        "The vast, vast majority of the world uses day-month-year. Only America uses month-day-year."

        .

        And neither are correct according to the ISO.

        https://en.wikipedia.org/wiki/ISO_8601

        EDIT

        OOPS! - Hadn't read down to BlartVersenwaldIII's comment when I posted my answer.

      5. JeffyPoooh
        Pint

        Re: Pi day

        notamole mentioned, "Only America uses month-day-year."

        Canada uses all possible formats at random.

        We've standardized on: "All of the above."

      6. Frumious Bandersnatch

        Re: Pi day

        Japan uses month-day. But then, they have symbols 月 and 日 tacked onto the numbers to signify month and day, respectively. Of course, they don't use something as stupid as mm-dd-yy, as you mentioned.

        For the full three-part date, the only things that should be considered are big-endian (Japan, which also has 年 to go with 月日) or little-endian (pretty much everyone else) and not some idiotic random order.

    3. BlartVersenwaldIII
      Happy

      Re: Pi day

      Some of us like the nice ISO date format... yyyy-MM-dd; it has the advantages of both sorting lexically and being unambiguous to people used to either MM/dd/yy or dd/MM/yy. After dealing with Hilarious Misunderstandings thanks to left/rightpondian misinterpretations I put on my benevolent dictator hat and got all us techies to standardise on ISO format instead on pain of death (or at least making the tea for a week).

      As a bonus for those of us who prefer it and use linux, ls has inbuilt support for using it via the --time-style=long-iso option, add it to your alias file today!

  3. Chris Miller
    Boffin

    I wonder how much that would have cost commercially. Would it have been cheaper to buy a supercomputer?

    Iwao has been interested in the famous irrational number from an early age.

    Not merely irrational (like root 2), but transcendental, as proved by von Lindemann in 1882.

    1. Charlie Clark Silver badge

      I think the point is that, no it would not have been cheaper to buy a supercomputer. This is a single, highly specialised algorithm, so not necessarily directly comparable with the stuff running on supercomputers. But the speed and price are only, er, part of the equation: with a supercomputer you normally have extremely fast networking and storage for data going in and out.

      Still, you can see that we're probably only a couple of generations from being able to rent say a supercomputer rack complete with dedicated glass fibre link. And in the meantime you can develop and test algorithms for that spanking new machine while you're waiting for it to be built.

      1. Fading
        Terminator

        Yep and it took

        10,000 years to come up with 42.....

    2. JeffyPoooh
      Pint

      Normal numbers

      CM wisely noted, "Not merely irrational (like root 2), but transcendental..."

      Yes, but also: Is it Normal?

      1. Aquilus

        Re: Normal numbers

        Probably.

    3. TheVogon

      Google cloud has loads of unused capacity. So much so that they are hiring Oracle staff to run it. After all who has more experience of running an empty cloud no one wants than Oracle.

  4. anthonyhegedus Silver badge

    Google make a nod to the rest of the world by saying "today March 14 (represented as 3/14 in many parts of the world)" - many as in "all the parts that are the USA", so a few then. Can we celebrate this properly on 31/4?

    1. Loyal Commenter Silver badge

      22/7 is the (marginally more accurate) international pi day

      1. find users who cut cat tail

        It's much more accurate. Last time I checked 3/14 ≈ 0.2142857, which is quite far from π. Or have Americans ruined fractions too?

        In some countries dates are written with a dot (mine for instance), but the order is D.M, not M.D, so you can't get 3.14 anyway.

        1. Loyal Commenter Silver badge

          more accurate, because 22/7 (3.142857...) is a little closer to pi (3.141592...) than 3.14 is.

          1. Charlie Clark Silver badge

            Depends on your defintion of accuracy: they're both only accurate to three significant digits and after that are almost equally inaccurate. I'm not engineer but I think that they're both good enough for most mechanical situations but nothing like good enough for anything computational.

            1. Loyal Commenter Silver badge

              You're right of course - I meant precise, rather than accurate...

      2. Charlie Clark Silver badge
        Coat

        Not only is it more accurate, it's also actually feasible. Or has April got another day and nobody told me?

        It's a pity because otherwise we really could have our Pi and eat it!

        1. John Savard

          Oh, well. You could always celebrate Pi Day on January 3rd.

    2. TimR

      We could, if there was a 31/4.....

      1. Loyal Commenter Silver badge

        Since 31/4 is 1/5, you might as well go ahead an celebrate it on the 3rd Feb, but for the previous year.

    3. ratfox
      Happy

      Can we celebrate this properly on 31/4?

      By all means, please do celebrate Pi day every 31st of April... But it's going to be hard.

      Personally, I prefer to celebrate Pi day on 22/7. Not only the order is correct, but it's also more precise.

    4. e^iπ+1=0

      "Can we celebrate this properly on 31/4?"

      30 days hath September, APRIL, ...

      So, no, you can't.

    5. John Savard

      As a Canadian, whose tax deadline is April 30, the last day of April, I am uniquely qualified to notice quickly that April hath thirty days, and hence the thirty-first of April is not available for Pi Day celebrations.

  5. Little Mouse
    Headmaster

    "...staffer breaks record for most *precise* Pi calculation"

    FTFY

    1. The First Dave
      Facepalm

      Re: "...staffer breaks record for most *precise* Pi calculation"

      In what way does this calc fail to be more accurate than the last?

      1. Charlie Clark Silver badge

        Re: "...staffer breaks record for most *precise* Pi calculation"

        Because precision and accuracy have clear and distinct mathematical definitions.

        1. I am David Jones

          Re: "...staffer breaks record for most *precise* Pi calculation"

          Yes but given that pi has been calculated (and not measured) and is presumably correct, then an increase in precision goes hand-in-hand with an increase in accuracy.

          1. Charlie Clark Silver badge

            Re: "...staffer breaks record for most *precise* Pi calculation"

            Correct, ie. formally correct, to n significant digits, I don't see accuracy coming into it unless you're using one of the many approximations. It's all a bit pedantic but it does make sense.

  6. StuntMisanthrope

    Golden ratios.

    Just keep going round in circle and coming back to λ and optical frequency. #notation

  7. richardcox13
    FAIL

    Seems a lot of work when they could have just asked a colleague

    Jon Skeet can recite π. Backwards.

    1. Loyal Commenter Silver badge

      Re: Seems a lot of work when they could have just asked a colleague

      I like the fact that the top answer in that Jon Skeet Facts question is by Jon Skeet.

  8. Anonymous Coward
    Anonymous Coward

    Pi

    Three point one four one...

    A server toils endlessly

    to bring a Pi-ku.

  9. hammarbtyp
    Joke

    31.4 trillion? Rubbish

    Trivial. I've calculated to 50 trillion digits, but got bored.

    Spoiler alert - the 50 trillionth digit is an 8

    1. Charlie Clark Silver badge

      Re: 31.4 trillion? Rubbish

      How are you defining your trillions, though? 10^12 or 10^18?

    2. PhilipN Silver badge

      Re: 31.4 trillion? Rubbish

      Agreed. I am going for the record the proper way using pencil and paper. Gotta go. In a rush to get this done before the End of Time.

      Oh - will pi and other universal constants still be relevant then?

      Bugger!

  10. hammarbtyp

    As a weird coincidence, the number of digits equals exactly Google's turnover for last year.

    Who knew that the when the company derived its name from the number Googolplex, it was not only a name but a financial aspiration

    1. DaLo
      Headmaster

      "Who knew that the when the company derived its name from the number Googolplex"

      They didn't they derived it from a googol, their HQ came from a googolplex.

      1. Anonymous Coward Silver badge

        Derived is an interesting way of saying "spelling mistake"

    2. jake Silver badge

      The name "google" is a craisis.

      It started life as "go ogle".

  11. Sgt_Oddball
    Coat

    Was the server that worked in this...

    Called Colosson by any chance?

    Also how big would the book be for the board game?

    Mines the one with the WordWang rule book in the (deep) pocket.

  12. dmjames0

    Pi on Pi

    I wonder how long this would take on a Raspberry Pi

    1. A.P. Veening Silver badge

      Re: Pi on Pi

      "I wonder how long this would take on a Raspberry Pi"

      The answer is "too" (long) ;)

    2. jake Silver badge

      Re: Pi on Pi

      Try it. Get back to us when you have the answer.

  13. Jay 2
    Pint

    Makes me wonder what's *not* Numberwang...

    Can't be bothered to RTA, but +1 and a virtual pint for the M&W ref!

  14. JimmyPage Silver badge
    Headmaster

    Beautifully meaningless ...

    on the basis no one - but no one - is going to be checking.

  15. gnasher729 Silver badge

    Once we get 10^18 digits in some base other than decimal , things might get interesting.

  16. illuminatus

    How many digits?

    Approx 10trillion * PI digits...

  17. Dwarf

    Output

    So where can we download the output - so we can start learning the sequence.

    1. Anonymous Coward Silver badge
      Facepalm

      Re: Output

      RTFA. Follow link. Done.

      Available as an API or a disk snapshot.

  18. JeffyPoooh
    Pint

    Puh!!

    I only needed a few minutes to work out the exact value of Pi last year on 'True Pi Day', the 22nd of July.

    It's equal to 3.142857... (repeating).

    Exactly.

  19. Grizzerly

    I've been retired from the industry for many years, but I love this story! The sheer joy of doing something like this just for the hell of it is great. As long as there are people who will do things like this, there is hope for humanity. Especially if there are publications which will head the story "That's Numberwang!".

    Keep up the good work!

  20. jake Silver badge

    All y'all do realize that ...

    ... this subject matter is completely irrational, right?

  21. Tomo

    Chinese inspired "numberwang" might be inappropriate as her name suggests she's Japanese-- I'd remove the "g" at the end of it.

    1. diodesign (Written by Reg staff) Silver badge

      Numberwang

      Ah, no - Numberwang is a reoccurring British comedy skit created by Mitchell and Webb

      https://thatmitchellandwebb.fandom.com/wiki/Numberwang

      It mocks math-based TV game-shows.

      C.

  22. SVV

    My quantum computer just calculated it to an infinite number of digits

    Instantly! Unfortunately the result got stored in the new universe that got created as a result of running the program.

  23. Benchops

    Eric Morecombe reciting digits of Pi to André Previn:

    "zero, one, two, three, four, five, six, seven, eight, nine, zero, one, two, three, four, five, six, seven, eight, nine, zero, one, two, three, four, five, six, seven, eight, nine, zero, one, two, three, four, five, six, seven, eight, nine, zero, one, two, three, four, five, six, seven, eight, nine, zero, one, two, three, four, five, six, seven, eight, nine, zero, one, two, three, four, five, six, seven, eight, nine, zero, one, two, three, four, five, six, seven, eight, nine, zero, one, two, three, four, five, six, seven, eight, nine, zero, one, two, three, four, five, six, seven, eight, nine, zero, one, two, three, four, five, six, seven, eight, nine, zero, one, two, three, four, five, six, seven, eight, nine, zero, one, two, three, four, five, six, seven, eight, nine, ..."

POST COMMENT House rules

Not a member of The Register? Create a new account here.

  • Enter your comment

  • Add an icon

Anonymous cowards cannot choose their icon