Tip:
Highlight text to annotate it
X
(Applause)
Ok, now, let's start thinking about our age, which is the information age.
Every minute tetrabites of data are saved in the cloud.
A company like Google
needs one million servers, three million computers
to handle these data.
This is because global communication and networking
shape our modern society.
Cheaper and smaller computer power
has triggered the digital revolution.
Some author says: "We are living the equivalent
of the industrial revolution of the information."
Companies are shifting to the most measured age in our history
but for us this is a new massive flap on human dynamics.
Despite that, most of the efforts to understand these data
have been directed to marketing, but when we have the ability
to capture and understand massive amounts of information
we can revolutionize science, technology, business and so on.
Then, this phenomenon is directly related
with human growth.
Nowadays half of the 6.6 billion people on Earth
live in a city, but, by 2030 5 billion people is expected
to be in the city. This is because every day
one hundred thirty thousand people move into an ancient city.
This is going to bring huge amount of problems
in terms of crime, polution, disease, infrastructure and congestion.
Scientist, Marshal McCluhan we have to - the medium is a message -
then, we have to think in technological innovations
not by their content but by what are their means
to change society and here comes my question:
Can we use this massive amount of information to plan for better cities?
Then there is a paradigm shift,
that is that the Internet is becoming mobile
as Google revolutionized the Internet
mobile is the next frontier.
And this is because mobile phones are after all sensors,
everytime you make a call, receive a text message
or send a text message,
your location is pin-point by a mobile phone tower.
This is generating massive amount of information
in real space and time.
Marketers who want to understand that, to understand their customers,
financers to predict the price of their stock
but hopefully we're going to have governments to use this to craft
their policy to control an epidemic outbrake
and other public hazzards.
This phenomenon I've been talking about
is by no means a thing of only rich countries,
we have almost 5 billion mobile phones users world-wide
the median penetration is 67% and in those places
that is less than that, there is a growing market.
Then, as Sandy Pentland said, mobiles phones are the neurons
of a truly global and emergent neural system,
but, comes the bad news, the data alone do not bring any magic,
then, we all hope that the magic is in the statistical tools that allow us
to crunch and understand the laws in the large number.
We like to believe that we are unique and with a lot of free will
but when we analyze data
we see there is a lot of predictability
and we can be grouped in few groups.
Then, let's go to a city,
this is Boston and Cambridge
across the river, where MIT and our lab is
How will it look like if we would know how many people
are accessing the Wifi access point?
Nowadays in this city we have more than one million people
with Smartphones. When you have access
to the WiFi you need an accuracy of 25 meters,
then, there is a lot of physical resolution.
Then, if we see the relativity
it looks like this, the bars is these 25 meters
boxes that tell you how many people are there,
and the colors mean, in blue those are most used at night
and in red, those that are most used during the day.
Here we have real time information of land used
we can know with this accuracy how many people we have in parks,
in commercial areas, in residential areas and so on.
And then, it's great we knew it that visualizations
can be fantastic.
How do we use it? How do we fufill our dream?
Well, it doesn't look so fancy but, then you have to do -
I'm gonna give you some tips I've been working on -
then you have to count and you have to group
as you maybe are already thinking,
not everyone uses the phone the same; then, there are some people
that use the phone only
only once a day, or some others
that use one call per hour and then we start grouping,
but the thing is, no matter if you don't use your phone very often
if we follow you long enough for a year, we really learn
about these flexible and fixed activities of everyone.
Then we can construct trajectories with the movement of everyone
and not only that, but we can know
where those locations in a city in which each person
is most likely to be found.
Imagine everyone and trace some circles
that enclose those trajectories.
Then we have one million people in the city,
with their circles and inside there
the locations where they go; and by a location we mean
a visit that lasts more than an hour,
a lot of information to play with.
Then you might be thinking how is my radius different to your radius?
what it looks like, how it changes with cultures or age.
Then we have - and this is the same data
that is saved in every city we have a mobile phone and
we have measured this over and over and it looks like this:
Many people have - this is the powerlock
we have seen today I put it in this scale,
it is like - most of the people have smaller radius
and some others have bigger radius those are the ones that connect
one city with the other.
Then you might be thinking, ok, that's nice
but what happen with those radius I used it in many locations in the city
then you loose predictability.
Then comes the impressive part, we can group mobile phone users
and it happens that there are some people that in six months
they stay, visit only five locations,
some others visit ten or fourteen,
but the impressive part is that we see the first ten,
they're 60% of their time then, we have their radius,
we have the amount of people and they're most of the time
in few locations.
But if we want to do plannning we need to learn something else,
that is, what is it that they're doing?
We want to have - as someone said:
it is not particle it is people
we need to learn which kind of activities, they're doing,
and for that, we need to go to the demographics in the city
to real space, to see what are their characteristics, then,
we saw the travel surveys we analyzed in Chicago
of 1% of the population, that is 30,000 people,
10,000 households.
And here comes the challenge: this is a 2009 survey
they are asked if they're registered full time in a school
or they have a full-time job.
We have 54% of the people that is in none of them
for a modelling of the city if you don't know how to model
how the people is moving, that is a big problem.
But let's see the powers of locations of the devices,
this is the same people, but we're just creating the movie
with the data of this survey.
And then here is 30,000 people here in the morning,
you see in the bar, here to the right, the activities they're doing,
that we see in blue, people that is working,
in yellow, people that is at school, red, shopping
light blue, recreation we're in the noon,
is a better distribution over the city,
after 3:00 PM no more yellow, everyone left the school,
then we have more people in recreation;
some other have stayed at work, like us,
and now is white, white is being at home,
and the city went to sleep.
The impressive thing is that when we took
the trajectories of everyone and then, my trajectory,
my itinerary is my vector, and we can make distance
between each other and cluster the people,
those that are closed are the similar itinerary,
we are like this.
Not so much complexity,
this give us a lot, and it was supposed to be
a not predictable city but there is a lot of things to do
just seeing our itinerary.
Then - ok,
we have learned we travel a lot
we're predictable, let use it.
Let me show you now in the time I have left,
to show you how we can use this.
And we move now to San Francisco,
those points are the mobile phone towers,
then from my mobile phone data like this location of our data
counting we can measure what is this house,
likely to be the house of people and we can track trips
that occur in a particular time windows of the day.
As you're familiar with Google map, this kind of tower,
you have town of origin, town of destination,
the route is given, we play like that with a mobile phone
my students wouldn't say play, right?
Here is how it looks.
We have every four hours, because we need to wait enough
to have enough movement with the phone and
here is what we see and then we're able to capture
different usage of the streets with the phone for the first time
and to know that we're doing correct we can validate it with taxi travel time
then, all this work and the taxis have given me the same
then you might say: Why do we do all this?
The thing is that taxi travel time cameras, look detectors
tell you volume, but don't tell you origin and destination,
then here comes the tower, the streets are colored
not just pure by the volume but by how many
major vehicles sources they atract
this is the kind of popularity index of the street
captured by the mobile phone.
Then we have for example here our red segment with a volume
of 988 vehicles in the four-hours morning period,
and then we're marking in yellow
where are the vehicles sources that produce 80% of its traffic.
We have others that are like the red ones,
that are more popular streets, similar volume but they're used globally
and all those blue streets are the ones that might be congested
but are my neighborhood street used locally.
Then we can correlate
the congestion of a street with the neighbourhood
that are generated with the power of the data
without the data, it wouldn't have been possible
and how can we use this?
We use it like that, we check what is the congestion
in a particular street and then we map it with
the neighbourhoods that are more affected by that congestion.
In our neighbourhood each traveller uses their own path,
but we have this connection between congestion and affected
compared to free-travel time.
And then that allow us to reduce such target
and what we have
is reduced for drivers every one-thousand users
we can reduce 16% of the total travel time
thanks to - this is a targeted decongestion.
If we have had the same reduction of trips
but not knowing where to target that would be 1% of total reduction,
that means we're gaining a two minutes travel
versus 32 minutes for the same route.
This is, I hope, an example how can we put data into work for better cities
and then that brings me to my conclusions,
cities are exploding with people, they're going to be exploding with people
but they're also exploding with information.
We have given up our privacy in order to get access
to instant information, but we are data generators,
then, I invite you to think in the ways we can use this data
the ways we can teach new students,
to start working with this,
because here, maybe a lot of new jobs,
and hopefully, I believe that it is now that we can really learn
about new clues of human dynamics
just now because the data is available
in he information age.
Thanks.
(Applause)