Press "Enter" to skip to content

What is Knowledge in the Age of Big Data? | Timandra Harkness | TEDxSquareMile


what is knowledge in the age of big data
it sounds like rather big ponderous
question just now so I’m going to start
with something a bit more tangible what
do you know do you know what this is I’m
going to give you a clue it’s the same
thing photographed from both sides any
thought
it’s the shin bone of a young wolf so
don’t feel bad if you didn’t recognize
that and if you were thinking it’s a
tally stick you are also right because
about 30,000 years ago in a cave in what
is now in the Czech Republic in the Ice
Age somebody scratched 57 notches into
this wolf bone 57 notches in groups of
five making it therefore the oldest
surviving example of digital data data
you could count on your fingers now
today of course we think of digital data
as being something you could put into a
computer in a form that a computer can
store and process and give back so it
would be in binary so 57 in binary is 1
1 1 0 0 1 satisfyingly that’s just under
a byte making a byte roughly equivalent
to a wolf’s bone thank you that’s a pond
that only I enjoy it also means that the
one terabyte hard drive on which I back
up my entire life is equivalent to a
million million wolf bones so when
people say big data that’s what they’re
talking about a million million wolf
bones so much for data what about facts
which of these facts do you know which
these questions could you answer where
is Toronto very good excellent I’m glad
you are joining in how old is Harvard
University for older than the US very
good you’re going to get top marks in
this quiz
it is indeed older than the country in
which it now sits what is the French
word for werewolf excellent
do you know why you know that ah cuz
he’s French okay well that’s a head
start I it’s a fact that I know and
never knew I knew until somebody asked
me in a bar so good these are facts
there and if you didn’t know them what
you you do now and and also you could
easily find them out without this
question though how many cats are there
in the world now this question last
Tuesday just gone I was hosting an event
called maths inspiration which is talks
about maths for school students and one
of the speakers Rob east away was
talking about estimation and he used the
example of Fermi questions where you
don’t know the answer and you can’t
really find that exact answer but you
can make a good estimate from what you
do know and he invited the audience to
contribute questions that you could use
as a fermi question so this appeared in
the question box in green ink how many
cats are there in the world and Rob
started from something he did no he
thought well I know how many people
there are in the world was about seven
billion people so if I estimate the
average number of cats corresponding to
each person the the mean number would be
one he decided because some people don’t
have a cat some people have several cats
so he estimated there are seven billion
cats in the world but before he’s even
finished a hand went up on the second
row so I said to the two school student
yes what’s your question and he said I
don’t have a question I have the answer
six hundred million I’ve just googled it
on my phone so they do say that in the
Internet age all human knowledge is just
three clicks away but where does that
figure come from not that figure 600
million comes off the internet and I
haven’t been able to find any reason why
it’s more convincing than the seven
billion estimate that Rob came up with
the figure of three clicks away I can
tell you where that comes from
that comes from a 2012 Stanford research
paper by West and less kvetch in which
they got some human beings to play a
game called Wikipedia to see how quickly
they could get from one piece of
information to another on Wikipedia and
while the human beings
doing this West and less kovitch
calculated mathematically the shortest
possible route the minimum number of
clicks that you would need to do the
same road now I realize not everybody
spends as much time looking at graphs
with as much love as I do so let me talk
you through this a little the number of
clicks along the lot the x-axis here is
a logarithmic scale which means that the
numbers get closer together as you come
this way so it starts off with one in
the middle is 10 over this side is a
hundred so you can see that the the
colored lines which are the human roots
the roots the human beings found are
sometimes as long as 70 or 80 clicks
whereas the shortest possible route that
mathematically optimum route that a
perfect machine would get are never more
than ten clicks and you can also see I
think the the left-hand scale is the
percentage so roughly how many how many
how many roots were that long that the
frequency so the most popular length of
route is three and this is a bit clearer
actually if I show you that the table of
results you can see the shortest
possible paths the mode the median that
are both three and the mean is 2.9 I’m
obviously using a working assumption
here that you know what mode median and
mean that mean they need average so you
could say that this that all human
knowledge in the Internet age is only
three clicks away on average if you’re a
machine if you’re a human being it’s
four clicks away but nevertheless it’s
out there so does this mean that if you
have an internet connection you know all
human knowledge of course not because
you only have a finite amount of time in
your life to click links on the internet
and if you remember nothing else from my
talk today please take away that one
fact you only have a finite amount of
time in your life to click links on the
internet if however you are a machine
that can click links at superhuman speed
then you can amass a lot of knowledge if
you are for example IBM what
an algorithm developed to understand
human language which was let loose onto
the internet to acquire masses and
masses of facts and information and then
entered for the u.s. game show Jeopardy
where it beats the to human champions at
a quiz show there it is in the middle
although it did get some things wrong
it did answer a category of questions US
cities onion it in the category US
cities it answered one of the questions
with Toronto which to us seems like a
really basic mistake but to IBM Watson
it can collect all this information but
it doesn’t have the context that we have
it doesn’t have the real-world
experience that we have sometimes data
and facts and not enough so going back
to our wolf bone for example we can
count the notches we know there are 57
notches but we don’t know what the
notches were counting we don’t know who
scratched them or what for we don’t know
what they mean I like watching the TV
programs where they dig up human
skeletons and then they examine them
forensically and tell us something about
what they can tell about the person who
lived so they can say oh we know that
this woman lived with chronic toothache
for 50 years or we know that this man
was attacked with a sword but he
survived and he lived on with a scarred
hand and face for the rest of his life I
like it partly because it reminds me how
much medical care has come on in the
last few hundred years but it doesn’t
really tell us what’s important about
those people it doesn’t tell us that
this woman with a toothache was actually
Queen Elizabeth the first of England it
doesn’t tell us that this man who
survived the sword attack was President
Andrew Jackson of the United States of
America you could dig my skeleton up in
500 years and you could work out that I
had a wisdom tooth removed by a surgeon
who was particularly skilled by the low
standards of the 21st century but it
wouldn’t really tell you anything
important about me or what I think or
what it’s like to be me it wouldn’t tell
you anything about my life you could
capture my data now
you could fit me with a heart rate
monitor and put that together with other
data about me like my location or what I
just used my credit card for or even
what other people were in the same
location as me at that time and you
could find some associations even if you
were a machine you could find
associations between my heart beating a
little faster and where I was and who
with and what I was doing but it
wouldn’t actually tell you what I was
thinking and feeling at that point an
algorithm couldn’t tell that because it
doesn’t have any subjective experience
an algorithm could look at a scan of my
brain and this is a scan of my brain not
actual size
now now Gregor would look at that it
could categorize it as a female brain it
could possibly go through and find some
disorder in my brain that I’m unaware of
and that even a trained neurologist
hadn’t spotted yet which could be handy
but the algorithm would have no idea
what that meant for me what impact it
would have on me what choices it would
face me with in my life for that I would
need a human neurologist who might be
assisted by a very clever algorithm but
who could talk to me human to human and
say well these are the implications I
understand how you’re feeling I can
empathize with you what do you want to
do when we say we know a person it’s not
the same as knowing a fact because we
are thinking not only about the outside
things we observe the data we’ve
collected on that person what we
observed about their behavior we’re also
using our subjective experience to
empathize to project ourselves and
imagine what it’s like to be them in
French it’s not even the same word
savoir to know a fact Canet to know a
person and to be familiar with a person
is perhaps a better translation in those
words connects were familiar there are
associations of kinship of
fellow-feeling so if we say a machine
knows a person that
that makes no sense a machine has no
subjective experience I may once have
been guilty of writing big data knows
you better than you know yourself that
was hyperbole I had a book to sell it’s
not actually true
big data cannot know a person so meaning
who knows what this means l mg TF why no
I get this sometimes if I’m having a
discussion on the Internet
and I asked somebody what what do you
mean by the age of reason and this they
respond with this l mg TF why let me
google that for you its sarcastic is not
the rudest thing they could say they
could say Jo G I and if you don’t know
what that means
just flippin google it now this annoys
me because I don’t want to know what a
machine thinks this means I don’t want
to know what Google thinks this means I
want to know what they mean by it in
this context if you ask a machine what a
word means a machine or the Internet is
probably using something called word
embedding where they model words in 200
dimensional vector space hold on to your
seats do the demonstration so when it’s
only three dimensions so it’s the same
as the room we’re sitting in with that
we’ve got an x axis y axis and the z
axis this way so you won’t get vertigo
but the way the way a machine would use
Wert word embedding to give you the
meaning of a word is it would put a word
into this mathematical space close to
words that are close to it in meaning so
the word data would be up here close to
words like knowledge and information the
word Toronto be down here near other
Canadian cities Watson and the word
werewolf might be up there close to the
word for werewolf in other languages for
example they could all be close together
but if you ask a human what data means
they might give all sorts of different
answers they might say well it’s not
just scratched on a wolf’s bone or it’s
it’s bites or bit
inside a computer or it’s the Latin word
for something that’s given and it’s a
plural word of datum those are the
particularly pedantic people that I do
work with sometimes I but it also people
we use the word in different ways you
might use the word data to close down a
conversation and people often do these
days they say well this is this is the
data or these are the data if they’re
that kind of person this is this is the
evidence that’s it there’s no more
discussion here I have finished the
conversation by giving you something
incontrovertible I prefer to use data to
open up a conversation to say well this
is a given but this means it’s the
beginning of a process of reasoning
given that given that there are seven
billion people in the world how many
cats are there let’s embark on a process
of reasoning reason itself has different
meanings in different contexts there is
the age of reason which is a historical
period but why that’s important I’m sure
historians are going to argue until the
end of whatever age were in now there’s
without good reason which to a lawyer
means I might be able to get you
compensation for losing your job and to
a philosopher means do we really have
free will or if you’re a Blaise Pascal
the heart has its reasons which reason
knows not although as the the French
person in the audience will know what
was he wrote it in French and the words
he chose were liquor I see those old
color plays on the conic one he used the
word for knowing a person for being
familiar with being acquainted with so
reason is not acquainted with the
reasons of the heart which has a rather
different meaning and I asked myself why
did Pascal choose those particular words
what did he mean what did he intend what
did he want me as the reader to get from
that particular choice of words meaning
is not something that is embodied in a
word like the color red in these
balloons meaning is something that goes
from the speaker or the right
to the listener or the reader so he has
an intention about what I should get but
in order to choose the words for that
intention he has to project himself
imaginatively into my mind his uses
empathy to think I’m a subjective human
being how is a subjective human being
reading this going to respond are they
going to get from it because the writer
knows that the meaning is created by the
person reading the meaning doesn’t exist
objectively in the words the meaning is
created in the process of reading and
thinking and understanding and that’s
why a machine cannot mean anything
because it has no subjective mind in
1636 when Harvard was founded you would
have had to go to Harvard to get the
benefit of all the knowledge that they
have there you’d have have to go and sit
in their library to read their books
you’d have to sit in their lecture
theatre to listen to their lecture and
speaking today you don’t have to you
don’t even have to get dressed if you
have an internet connection you can
download research papers from Harvard
you can sign up to a free online course
you can listen to their lecturers you
can even do quizzes to find out how much
you’ve understood well that is marvelous
it’s a wonderful thing that the internet
and data have opened up so much of the
field of human knowledge to so many more
people but it’s not a good thing if we
lose track of what knowledge is if we
start to think of knowledge as a
substance like porridge that you scoop
out with your facts ladle into the empty
bowls of ignorance come on Charlie you
all have a fact ladle ready to fill the
empty bowls of ignorance knowledge is
not something that exists in the bits of
paper it’s not existing in the ink on
the paper on the page of a book or on in
the ones and zeros inside a computer
knowledge is a field of human activity
knowing is a verb it’s a thing that
human beings do so yes it may be true
that in the Internet age any fact on the
Internet can be found in four clicks or
in three clicks if you’re a machine but
the machine won’t know
it means the kid that googled and found
the answer of 600 million cats had
completely the wrong idea the point was
never to learn how many cats there are
in the world the point must learn how to
estimate how to build your reasoning
muscle by starting with something you do
know a given there are seven billion
people in the world and seeing if your
reason can take you from that to
understanding something completely new
about the world
so in the age of big data knowledge is
as it has always been a field of human
activity and the human activity is
learning and reasoning and knowing thank
you
[Applause]
Please follow and like us: