WS MoreOrLess: Big Numbers

Summary

---

More or Less: “Big Numbers”

Host: Tim Harford
Guest: Chris Baranjuk (Science & Technology Journalist)
Date: May 16, 2015
Podcast: BBC World Service – More or Less

---

Episode Overview

This episode of More or Less delves into how big numbers—and computers’ struggles to handle them—can have dramatic real-world consequences. Host Tim Harford and guest Chris Baranjuk explore high-profile cases from exploding rockets to airplane glitches and viral YouTube videos. They explain the concept of “integer overflow,” the technical limitations of computers with numbers, and what this means for everything from technology disasters to daily life.

---

Key Discussion Points & Insights

1. Ariane 5 Rocket Disaster – When Maths Goes Wrong

• [00:42] Story Introduction: Harford recounts the 1996 Ariane 5 rocket explosion, set off by a computer error rather than human or mechanical fault. The mission (the “cluster project”) failed only 37 seconds after launch.

  • Quote [00:51]:
    “The failure wasn't caused by human error or a mechanical problem, but simply because the rocket's onboard computer got its maths wrong.” — Tim Harford

• [01:53] Explanation: Baranjuk describes the software problem: a number too big for its allocated memory space “switched [the computer] into this shutdown mode and self-destruct sequence.”

• [02:25] Analogy: Harford likens it to a cricket scoreboard not designed to display four-digit scores.

  • Quote [02:25]:
    “It’s like trying to count above 100 and you’ve only got two digits... and it’s sort of like that, only in computer language.” — Tim Harford

2. Integer Overflow: Not Just for Rockets

• [02:48] Everyday Examples: Baranjuk explains that assumptions about limits are common in programming—that “no one’s ever expected” to exceed certain values.
  • Today’s real-world consequences include an FAA advisory for Boeing 787s: if left running for 248 days, a counter rolls over, risking a shutdown in flight.
  • Quote [03:42]:
    “If the counter does reach that maximum, even if the plane is in mid air, it'll enter a fail safe mode, shut down part of the engines, and so, you know, the potential for catastrophe is certainly there.” — Chris Baranjuk

3. Why Can’t Computers Cope with Big Numbers?

• [03:56] Technical Limitations: Baranjuk explains that efficient computer design means limiting number size to what’s “assumed” necessary, not infinite values.
  • YouTube’s Gangnam Style ([04:04]): The song’s view count in 2014 famously hit YouTube’s previous max (2,147,483,647), forcing a software update.
  • Quote [04:16]:
    “But YouTube then of course updated its software to carry a larger number, and the new maximum is now well over 9 quintillion.” — Chris Baranjuk

4. The Millennium Bug – Lessons from Y2K

• [05:26] The Panic: Many computers only recorded years using two digits. As 2000 approached, this caused fears that systems would malfunction.
  • Baranjuk notes that much behind-the-scenes work by programmers averted catastrophe, but bugs still appeared: train station signs showing the year as 1900, or odd five-digit year readouts on websites.
  • Quote [05:48]:
    “A lot of computer programmers would be very upset if you said nothing seemed to happen because of two things. One, they worked very, very hard... And the second thing is that, well, yes, these bugs did actually cause some errors in the early days of the current millennium.” — Chris Baranjuk

5. The Next Big Bug: 2038 Problem

• [06:33] Upcoming Challenge: Harford and Baranjuk discuss how computers count in binary, with 32-bit date systems hitting their max on January 19, 2038, at 03:14:07 GMT.
  • No one really knows how systems will behave once this limit is reached—paralleling Y2K concerns.
  • Quote [06:56]:
    “There is a theory that in 2038, the same sort of problems as we had with the millennium bug might happen again... at this particular moment in time.” — Chris Baranjuk

6. Listener Follow-Up: Due Dates Coincidence

• [07:45] Listener Story: A statistical curiosity from a previous episode: listener Penelope and her friend Eleanor expecting babies on the same day—a 1 in 30 chance.
  • Eleanor reports both babies arrived within 24 hours of each other, but not on the same day.
  • Quote [08:49]:
    “I think that's still close enough for them to have joint birthday parties.” — Tim Harford

---

Memorable Moments & Notable Quotes

• Explaining integer overflow:
  “This is like trying to count above 100 and you’ve only got two digits.” — Tim Harford [02:25]
• On avoided plane disasters:
  “Ideally not in mid air.” — Tim Harford [03:39]
• On programmers’ invisible work:
  “A lot of computer programmers would be very upset if you said nothing seemed to happen…” — Chris Baranjuk [05:48]
• On YouTube breaking its own view counter:
  “Gangnam Style… had actually maxed out YouTube's view counter…” — Chris Baranjuk [04:16]

---

Timestamps for Key Segments

• [00:42] Introduction to Ariane 5’s integer overflow failure
• [01:53] Technical explanation of the rocket error
• [02:48] Real-world consequences: Boeing 787 and integer overflow
• [04:04] Why computer memory is limited; Gangnam Style’s YouTube counter
• [05:26] Millennium Bug (Y2K) and subsequent fixes
• [06:33] The 2038 Problem – Next “millennium bug”
• [07:45] Listener update: due date statistical story

---

Tone & Style

Tim Harford keeps the show light, witty, and accessible, mixing analogies with nerdy humor (“It blew itself up like in the Bond movies”). Chris Baranjuk provides technical clarity without jargon, making the audience feel part of the decoding process. Anecdotes—like babies nearly sharing a birthday—round out the episode with warmth and everyday relevance.

---

Summary prepared for listeners who missed the episode or want a refresher on tech catastrophes caused by big numbers.

Transcript

[Tim Harford]

This is the short edition of More or Less, first broadcast on the BBC World Service.

[BBC Announcer]

Thank you for downloading from the BBC. For details of our complete range of podcasts and our terms of use, go to bbcworldservice.com podcasts

[Tim Harford]

welcome to More or Less on the BBC World Service. We're your weekly guide to the numbers all around us. And I'm Tim Harford.

[Chris Baranjuk]

This nees sign cat Trois de unite.

[BBC Announcer]

Feu.

[Chris Baranjuk]

Alimage.

[Tim Harford]

In June 1996, the first Ariane 5 rocket took off from the European Space Agency's launch site in French Guiana. The launch marked almost a decade of hard work and billions of dollars of investment. The scientists involved were working on a mission to study how the solar wind affects the Earth's magnetic field. It was called the cluster project. I'm sure there's a joke in there somewhere. Because the rocket veered off course and only 37 seconds into its flight exploded. The failure wasn't caused by human error or a mechanical problem, but simply because the rocket's onboard computer computer got its maths wrong. It's an example of what's called integer overflow, when a number's too big for a computer to handle, and it affects all sorts of things, from planes to pop music, as the freelance science and technology journalist Chris Baranjuk has been telling me. So, Chris, what went wrong with Ariane 5's computers?

[Chris Baranjuk]

Ariane 5 was a very advanced spaceship that's still actually used today as a rocket. In 1996, it was its first flight. There was a number in the software that was trying to be converted from a large space in the computer's memory into a much smaller one. And consequently that number didn't fit. So it switched into this shutdown mode and self destruct sequence was initiated.

[Tim Harford]

So it blew itself up like in the Bond movies.

[Chris Baranjuk]

And of course, the European Space Agency had spent billions developing it in the first place, so it was a very, very bad start.

[Tim Harford]

So just to understand this idea of putting a large number in a small space, this is like trying to count above 100 and you've only got two digits. You've got a cricket scoreboard and no one's ever expected to score a century, or maybe no one's ever expected to score 1000 runs. So you've got three slots on your cricket scoreboard, but no one's ever expected to score 1000. And so you don't have that fourth digit. And it's sort of like that, only in computer language.

[Chris Baranjuk]

That's a good way of thinking about it. And the key part of your explanation there is that there's an assumption at the heart of that. No one's ever expected to score that many points. So this is what computers do. In many cases, this counter will simply roll over and go back to zero. And actually there are lots and lots of examples of. Only 2 weeks ago it was announced by the Federal Aviation Administration in the states that Boeing 787 aircraft suffer from exactly the same sort of problem. There's a counter in software on the control unit which drives power to the plane's engines. And that counter will max out if the plane has been left on for more than 248 days. So the advice from the FAA was simply to switch the plane off and on again, because then you'll restart the

[Tim Harford]

counter, but ideally not in midair, ideally

[Chris Baranjuk]

not in mid air. But this is a big problem because if the counter does reach that maximum, even if the plane is in mid air, it'll enter a fail safe mode, shut down part of the engines, and so, you know, the potential for catastrophe is certainly there.

[Tim Harford]

How is it possible that computers actually cannot cope with large numbers? You would have thought it was the most straightforward thing in the world for a computer.

[Chris Baranjuk]

It sounds really weird when you think about, first of all, the fact that computers seem so capable of doing certain tasks, and secondly that numbers are infinite. But the problem is, if you're building a computer system, you want to make sure that it's going to run as efficiently as possible. So you're going to choose storage spaces for the numbers that you store, which have limitations based on the assumptions that you'll make about what that computer will actually do. There's no point building a computer that can store an infinitely large number if you only need to do something like in our example, count the score of a player in a game. Another famous example that came up quite recently in December actually was Gangnam Style, a very popular song on YouTube had actually maxed out YouTube's view counter when it reached a very specific number of views, which was 2,147,483,647. It actually maxed out the number of views that YouTube can record at that time. But YouTube then of course updated its software to carry a larger number, and the new maximum is now well over 9 quintillion. So it'll be quite a while before Gangnam Style reaches that minute.

[Tim Harford]

So the classic example of this is clearly the millennium bug, which hopefully some loyal listeners will be too young to remember, but I remember very clearly the panic as we reached the end of the 20th century and suddenly realized all these computers were set, or some computers were set just to count the years in terms of the last two digits. I used the word panic because people were very concerned about this and nothing really seemed to happen, or did it.

[Chris Baranjuk]

A lot of computer programmers would be very upset if you said nothing seemed to happen because of two things. One, they worked very, very hard in the years preceding the year 2000 to actually patch software and introduce fixes for these things. You can go back to the mid-90s, and people are already talking about what they're going to have to do in the advent of the year 2000. And the second thing is that, well, yes, these bugs did actually cause some errors in the early days of the current millennium. So you can go online, you can find plentiful examples of this. There were signs at train stations that said the year was 1900. There were websites which gave the series of five digits 1-9-100-100 years after 1900, which is certainly correct, but it's not the way we would write that year.

[Tim Harford]

Now, computers don't actually count in decimal like we do, they count in binary. And the computers we're using today have a limit on the date and time they can count to. That means we could experience another millennium bug in the not too distant future. And as Chris Baranjuk explains, programmers think they've already worked out exactly when that will be.

[Chris Baranjuk]

There is a theory that in 2038, the same sort of problems as we had with the millennium bug might happen again. Because on the 19th of January, 2038, at 14 minutes past 3 in the morning, there's going to be a Limit reached on 32 bit architecture, which is counting seconds. So at that point, people expect that a similar reaction by software to what happened with the millennium bug could occur again at this particular moment in time. Computers which are counting seconds from a point where they start their count in 1970, they're going to reach the limit of their counter. And no one really knows how that will affect computer systems once they do reach that limit.

[Tim Harford]

To find out more, you can read Chris article about this on BBC Future. The link is on our website, BBC.com more or less. That's almost all we have time for. But before we go, last series, we were feeling broody when loyal listener Penelope Chaney got in touch to say, I'm expecting a baby in April and discovered that one of my close friends is also expecting a baby. On the same day, we calculated that there was a 1 in 30 chance of both Penelope and her friend Eleanor having their babies on the same day. Not too bad. But there was a much, much smaller chance of that day actually being the due date, April 4th. We were wondering what had happened when this message came through on the more or less answer phone.

[Eleanor]

Hi, this is Eleanor. Just to let you know, our daughter Juliet Imogen was born on Saturday 11 April, which is about 24 hours after Penelope had her little boy. So sorry we didn't quite manage to coordinate for you.

[Tim Harford]

So Penelope's baby boy arrived six days after his due date and Eleanor's baby arrived one day later. I think that's still close enough for them to have joint birthday parties. Thank you for letting us know, Eleanor, and we wish both families all the best. If you want to tell us about any births, deaths or marriages in your life, by all means get in touch via more or lessbc.co.uk until next week. Goodbye.

[BBC Announcer]

There are dozens of different podcasts now available from the BBC, including news, documentaries, science, business, arts and sport. For details of them all, go to bbcworldservice.com podcasts.