Google – The Code
Exercise
Watch the video, answer the questions on a piece of paper, then check your answers.
- How many web searches does Google get?
Over two billion a day
- What subjects did he search for that day?
Cities in Mexico and films in Hackney
- What was Google’s hunch?
That they could use all our searches to make predictions about our lives
- What did Google first try to predict?
Outbreaks of flu
- How many search terms were there in the database?
Over fifty million
- Google examined the data for what period?
The past five years
- Which search terms does she mention whose popularity matched the pattern of flu cases?
Symptoms, medications, sore throat, complications
- Was the popularity of these search terms an approximate guide to actual flu cases?
No, it was an accurate indicator of flu activity
- How did Google staff react to this discovery?
They were amazed
- How quickly can Google often predict a flu outbreak?
Before people have even gone to the doctor
Gap fill
Watch the video, complete the gaps on a piece of paper, then check your answers.
access
wisdom
forces
query
match
events
seasonal
trend
related
popularity
mirrors
finding
extraordinary
Transcript
With access to over two billion web searches a day, Google have found a way of tapping into the wisdom of the biggest crowd on Earth and by doing so they’ve been able to reveal the forces that control our lives and harness them to make predictions about us.
Think of all the things that people might search for on a daily basis. Think of the things that you might search for on a daily basis.
Yeah, well, I’ve searched for a couple of cities in Mexico and films in Hackney today.
Lots of people may be searching for a similar thing: movies in Hackney, for example. And you can see, if you looked at that query over the past three years, what the pattern of searches for that term looked like.
Google had a hunch they could use all our searches to make predictions about our lives. They wanted to see if they could match the pattern of certain searches with events in the real world. Google began by seeing if they could predict outbreaks of flu.
So flu has a nice seasonal pattern and because it has that pattern every year over many years, we’re able to take that trend and say which search queries match that pattern. So we built a database that included over fifty million different search terms...
Fifty million?
Yes. Well, we didn’t only include things that may be related to flu. We included things like "Britney Spears" or, I mean, everything that people search for would have been included.
When Google looked back at the past five years of data, there were certain search terms whose popularity exactly matched the pattern of flu cases.
So people were searching for things like "symptoms" or "medications" or "sore throat”. There were other things like "complications”.
So you’re saying that the number of search terms for flu-related things almost exactly mirrors the actual cases of flu that we see in the population?
That’s true. It is an accurate indicator of flu activity just based on lots of people searching for these terms. We were amazed by this finding.
As soon as they see this pattern of search terms, Google can predict there’ll be an outbreak of flu often before people have even gone to the doctor. This is the extraordinary power of the code.