1
00:00:00,216 --> 00:00:05,930
Okay, well, hello, here I am with Igor
[FOREIGN] and Sam Shaw at LinkedIn

2
00:00:05,930 --> 00:00:11,138
headquarters at Mountain View.
[laugh] And, what I would like to ask them

3
00:00:11,138 --> 00:00:14,827
is how they first got into social network
analysis.

4
00:00:14,827 --> 00:00:18,877
Igor already started this story which
predates LinkedIn.

5
00:00:18,877 --> 00:00:23,000
So, I will let him continue.
So, I'll just tell it again.

6
00:00:23,000 --> 00:00:26,182
So in the beginning when I started within
social network was 2003.

7
00:00:26,182 --> 00:00:31,752
So that is really like in the dark ages.
At that time the field was more driven by

8
00:00:31,752 --> 00:00:35,680
consultants.
And you had these views of small networks.

9
00:00:35,680 --> 00:00:39,752
I was playing in the role of Enterprise
knowledge management and Enterprise

10
00:00:39,752 --> 00:00:44,092
knowledge management was the search for
document at manner for something that it

11
00:00:44,092 --> 00:00:48,594
has I need to do and for expense to help
me out and the koffing that we had is that

12
00:00:48,594 --> 00:00:52,720
within each document we could create a
symetic math that you should come the

13
00:00:52,720 --> 00:00:56,524
document back that I give a account that
counts these concept that let.

14
00:00:56,524 --> 00:01:00,490
So its a nice reservation right.
On the other side we have names of people.

15
00:01:00,490 --> 00:01:03,407
There are experts, and we felt like that
these are kind of static.

16
00:01:03,407 --> 00:01:06,683
I mean, I don't know who they are,
because, otherwise, I would have directly

17
00:01:06,683 --> 00:01:09,780
reached out directly to them.
I don't know how they're related to one.

18
00:01:09,780 --> 00:01:12,438
So we took that and put it in Cindy's
plate.

19
00:01:12,438 --> 00:01:15,157
Because we had interactions around
documents.

20
00:01:15,157 --> 00:01:20,052
We have the shares that we have today.
So we could figure out why these two girls

21
00:01:20,052 --> 00:01:24,161
or these two ladies were part of the
expert that we will treat back.

22
00:01:24,161 --> 00:01:27,907
And then it created a map.
It created one of these social maps.

23
00:01:27,907 --> 00:01:31,774
So that, for me, was my a-ha moment.
Well, because, small start up,

24
00:01:31,774 --> 00:01:35,813
40, 50 people.
You read the map and say hey that's right,

25
00:01:35,813 --> 00:01:38,633
that's exactly how we work.
It did work.

26
00:01:38,850 --> 00:01:42,609
Mm-hm.
I mean that was a small group set up 40-50

27
00:01:42,609 --> 00:01:48,899
the company make it so we stop doing this.
[laugh] Now it's way different, now its

28
00:01:48,899 --> 00:01:54,828
maps of hundreds of millions of users but
there the difference is you get to play

29
00:01:54,828 --> 00:01:58,791
with real society.
And you can actually see how it's alive,

30
00:01:58,791 --> 00:02:01,188
how it changes.
That is phenomenal.

31
00:02:01,379 --> 00:02:03,923
Mm-hm.
So, so one thing I've been telling

32
00:02:03,923 --> 00:02:07,865
students is how it's different when you
slice a small network.

33
00:02:07,865 --> 00:02:12,888
And people can easily recognize oh yes
these are actually the groups that work

34
00:02:12,888 --> 00:02:16,831
together and so on.
And we'd imagine that with a large network

35
00:02:16,831 --> 00:02:21,155
you have a lot of overlaps from
individuals changing companies, that

36
00:02:21,155 --> 00:02:26,242
doesn't make the same amount of sense.
Or do you need to slice it before you can

37
00:02:26,242 --> 00:02:28,976
really work with it?
I think Sam and I can talk at length about

38
00:02:28,976 --> 00:02:31,520
that. How many pixels are there on the
screen?

39
00:02:31,520 --> 00:02:36,024
Sort of less than one hundred eighty
million depending on the resolution.

40
00:02:36,024 --> 00:02:40,096
See if I put a dot to remember.
I put a dot remember on Twitter, on

41
00:02:40,096 --> 00:02:43,921
Facebook and on LinkedIn.
Well we have this screen that's full.

42
00:02:43,921 --> 00:02:46,882
So you have to slice and dice it to some
extent.

43
00:02:46,882 --> 00:02:51,018
And through that, you start getting
actually the nice thing. huh Yeah.

44
00:02:51,018 --> 00:02:54,397
Many, a lot of challenges comes in how you
are going to slice it too.

45
00:02:54,397 --> 00:02:57,975
Because, and as we have done, run into
these problems where, slice it in a

46
00:02:57,975 --> 00:03:01,454
certain manner then you lose the
information that you want to look at.

47
00:03:01,454 --> 00:03:05,380
And you slice it in a different manner.
So, you look at LinkedIn in the Bay Area

48
00:03:05,380 --> 00:03:08,263
performs very differently than it does say
in easily.

49
00:03:08,263 --> 00:03:12,238
And, and looking at that is, is And
figuring out how you are going to slice it

50
00:03:12,238 --> 00:03:14,337
is quite a functional option.
Uh-huh.

51
00:03:14,534 --> 00:03:17,560
It's also leaked to me, he's back on his
steps.

52
00:03:17,560 --> 00:03:21,111
And in the old days it was like the myth
of the average man.

53
00:03:21,111 --> 00:03:23,893
There, there's an average person as an
ever did.

54
00:03:23,893 --> 00:03:26,793
Well, the average person is half female,
half male.

55
00:03:26,793 --> 00:03:29,752
The average family is 1.5 kids.
That doesn't exist.

56
00:03:29,930 --> 00:03:33,600
The slice and dicing makes me reminded is
not an average user.

57
00:03:33,762 --> 00:03:36,477
Mm-hm.
There's not an average figure to look at.

58
00:03:36,477 --> 00:03:39,464
The different aspect is.
There's, there are the students.

59
00:03:39,464 --> 00:03:41,962
There are job seekers.
There are professionals.

60
00:03:41,962 --> 00:03:44,677
There are this, there are that, there are
that, that.

61
00:03:44,677 --> 00:03:46,957
Facebook, you can have there are
teenagers.

62
00:03:46,957 --> 00:03:49,292
There are early teens.
There are late teens.

63
00:03:49,292 --> 00:03:52,387
They're all different.
[laugh] No, no, no.

64
00:03:52,387 --> 00:03:54,016
Definitely not.
Definitely not.

65
00:03:54,016 --> 00:03:59,603
Actually one thing that I noticed is that
it's drastically changing.

66
00:03:59,603 --> 00:04:06,440
I don't know if, did we meet in 2005 at
that Microsoft meeting on social

67
00:04:06,440 --> 00:04:08,136
nzetworks?
I'm not sure.

68
00:04:08,317 --> 00:04:11,941
They brought bunch of research, Research
material for their dance. And they brought

69
00:04:11,941 --> 00:04:16,229
kids, 13-14 year olds.
And they start saying, I do this on my IM,

70
00:04:16,229 --> 00:04:18,887
I do this on MySpace, I do this on the
phone.

71
00:04:18,887 --> 00:04:21,846
I do this and that.
And we were all flabbergasted.

72
00:04:21,846 --> 00:04:23,356
What?
You do that on this?

73
00:04:23,356 --> 00:04:27,595
And it goes like, you don't break.
Don't break up with someone on IM.

74
00:04:27,595 --> 00:04:31,028
You called him up or you don't do this on
the phone.

75
00:04:31,028 --> 00:04:34,528
I forgot what it was.
These individuals are our users.

76
00:04:34,528 --> 00:04:39,282
And, we all benefit from the perception of
prevalence of social networks.

77
00:04:39,282 --> 00:04:42,980
That it's not something, I don't know what
to do with it.

78
00:04:43,193 --> 00:04:48,456
It's almost inbred into their DNA.
Where as my generation, we going to look

79
00:04:48,456 --> 00:04:50,875
at my sister.
Why this, why that.

80
00:04:50,875 --> 00:04:55,000
But my dad, pffft.
But that changes drastically, so it's not

81
00:04:55,000 --> 00:04:58,504
a,
They, have the whole ecosystem sort of

82
00:04:58,504 --> 00:04:59,791
adapted to it.
Mm-hm.

83
00:04:59,958 --> 00:05:03,652
What we will be in five years.
I don't know, but it becomes so much

84
00:05:03,652 --> 00:05:06,170
easier to do on it, that what we actually
see.

85
00:05:06,354 --> 00:05:10,291
What they want from it.
Much easier than us telling them what it

86
00:05:10,291 --> 00:05:11,952
should be look at.
Mm-hm.

87
00:05:11,952 --> 00:05:14,843
Mm-hm.
And I, I guess there's kind of a, a two,

88
00:05:14,843 --> 00:05:18,288
two part question.
To what extent is individual speaking

89
00:05:18,288 --> 00:05:22,902
here, when they use a network such as
lifting different from holding, or the

90
00:05:22,902 --> 00:05:27,454
same as what they might be doing when
they're doing regular networking or

91
00:05:27,454 --> 00:05:30,530
hanging out?
And also to what extent you know, such

92
00:05:30,530 --> 00:05:35,451
behavior might inform the algorithms that
you use in trying to facilitate those

93
00:05:35,451 --> 00:05:38,773
interactions.
Do you, do you try to simulate what might

94
00:05:38,773 --> 00:05:41,020
happen, and?
And off a.

95
00:05:41,506 --> 00:05:42,424
Mm hm.
Offline.

96
00:05:42,586 --> 00:05:43,720
Mm hm.
Is actually.

97
00:05:44,840 --> 00:05:47,494
[laugh] Yeah, I mean, I think there's
quite a...

98
00:05:47,494 --> 00:05:51,880
I mean LinkedIn is a little bit different
in that it's, you know, I mean your

99
00:05:51,880 --> 00:05:56,670
information is inherently public which is
like, you know, you're putting your CV out

100
00:05:56,670 --> 00:06:01,171
there; you want to be found, is what our
core kind of principles of having this

101
00:06:01,171 --> 00:06:04,691
identity on the web.
So, like you know, when you have that kind

102
00:06:04,691 --> 00:06:09,308
of dynamic things change: like I think
people are much more conscious about what

103
00:06:09,308 --> 00:06:13,405
they want to present and then they change
their behavior, I think, and...

104
00:06:13,405 --> 00:06:18,080
To kind of reflect that, which you would
do differently, say in an offline setting.

105
00:06:19,420 --> 00:06:24,889
And we've, I-, it's, it's, I know, it's
hard to actually understand what would we

106
00:06:24,889 --> 00:06:28,034
do.
Like the differences between that and kind

107
00:06:28,034 --> 00:06:30,390
of.
Of whatever that we've actually.

108
00:06:30,647 --> 00:06:37,609
Have a really true chris butter sandal.
But what about, not right now as we are

109
00:06:37,609 --> 00:06:43,024
recording, but in two weeks.
The students will be learning about

110
00:06:43,024 --> 00:06:47,408
Milgrams experiment and passing messages
six times.

111
00:06:47,408 --> 00:06:50,218
And to.
You know, it, I, I think it doesn't may be

112
00:06:50,218 --> 00:06:53,865
happen naturally, but, it, it, but it
could, would be place where you might

113
00:06:53,865 --> 00:06:56,177
facilitate, people actually navigate.
Mm, mm.

114
00:06:56,331 --> 00:06:56,998
The.
Mm-hm.

115
00:06:57,152 --> 00:06:59,669
The network.
To what extend do you, do you see, and

116
00:06:59,669 --> 00:07:02,546
what kinds of smarts do you?
Things just kind of blow up.

117
00:07:02,546 --> 00:07:04,703
Right?
The number of friends of friends of

118
00:07:04,703 --> 00:07:05,576
friends.
Mm-hm.

119
00:07:05,730 --> 00:07:06,450
The.
Can we?

120
00:07:06,450 --> 00:07:10,360
The.
Again, 2003 there were three start ups in

121
00:07:10,360 --> 00:07:12,986
social networks.
Friendster, Spoke and LinkedIn.

122
00:07:12,986 --> 00:07:16,298
If you look at Spoke.
Spoke was a city dealing in business

123
00:07:16,298 --> 00:07:19,152
transactions.
I am a salesperson, I want to sell to

124
00:07:19,152 --> 00:07:21,664
company X, I don't know anybody in company
X.

125
00:07:21,664 --> 00:07:25,718
But who do I know that knows someone?
It's kind of similar to Stanley's

126
00:07:25,718 --> 00:07:28,858
experiment, like, how do I get my message
to that person.

127
00:07:28,858 --> 00:07:33,368
Well, I don't know who that person is.
In this case, I don't know who that person

128
00:07:33,368 --> 00:07:35,880
is.
You have to set up who you are trying to

129
00:07:35,880 --> 00:07:38,924
target.
I always felt that I would be the most

130
00:07:38,924 --> 00:07:42,514
successful start up.
In that domain, of course, you haven't.

131
00:07:42,514 --> 00:07:45,033
[laugh] I was wrong.
I'm fine with it.

132
00:07:45,411 --> 00:07:48,411
We.
We have a, a couple of parts that we're

133
00:07:48,411 --> 00:07:51,451
looking into.
One the pathfinder which is actually

134
00:07:51,451 --> 00:07:55,830
trying to surkee, that this is how we are
interconnected between a and b.

135
00:07:55,830 --> 00:08:00,069
So, making that path explicit and making
you the choices of which path do you want

136
00:08:00,069 --> 00:08:03,791
to go down in order to convey a message or
do anything across the world.

137
00:08:03,791 --> 00:08:07,669
And it's, it's slightly tricky, because
you have multiple hops to go through,

138
00:08:07,669 --> 00:08:11,339
multiple types of connections.
Whether you're directly connected to the

139
00:08:11,339 --> 00:08:15,424
person, whether you share a group with
that person, whether you live in the same

140
00:08:15,424 --> 00:08:17,905
area.
All of these sort of interplay differently

141
00:08:17,905 --> 00:08:21,938
within the algorithm, and we're still
learning how to actually make that think

142
00:08:21,938 --> 00:08:23,519
better.
Huh, huh.

143
00:08:23,519 --> 00:08:27,322
So, so any, any previews.
[laugh] Nope unfortunately.

144
00:08:27,537 --> 00:08:31,484
[laugh] No, no?
Okay That's one of these things, these

145
00:08:31,484 --> 00:08:36,650
cool things when you work in places like
that, did with something before.

146
00:08:36,650 --> 00:08:41,960
We do have the data to try things out.
Most of our ideas don't make it out.

147
00:08:42,136 --> 00:08:46,956
Because it's, either it's, it's extremely
hard, or the, the signal is not there, or

148
00:08:46,956 --> 00:08:50,072
the iterations.
And we, it's gonna take longer than we

149
00:08:50,072 --> 00:08:53,423
thought, and some do.
And when those do, sometimes they are

150
00:08:53,423 --> 00:08:55,540
extremely surprising to us.
Mm-hm.

151
00:08:55,885 --> 00:08:57,956
We just launched two product.
Yes.

152
00:08:58,301 --> 00:09:03,249
Which we, we laugh about it every day now.
Which are extremely viral, and they behave

153
00:09:03,249 --> 00:09:04,917
virally differently.
Mm-hm.

154
00:09:05,090 --> 00:09:09,520
And to understand why is actually a very
fascinating question to think about.

155
00:09:09,755 --> 00:09:14,074
So tell me more.
It sounds like you had something to do

156
00:09:14,074 --> 00:09:15,723
with this.
Oh yeah.

157
00:09:15,959 --> 00:09:21,927
So we did this thing called endorsements
where you can provide kind of a

158
00:09:21,927 --> 00:09:28,523
lightweight, thumbs-up sort to basically
on your colleagues' skills and expertise.

159
00:09:28,523 --> 00:09:32,471
And we really under.
It is really interesting to see that kind

160
00:09:32,471 --> 00:09:36,396
a social dynamic in that future it can
happen as people slowly get, get to

161
00:09:36,396 --> 00:09:38,894
normal, go to quality figure out skills
they have.

162
00:09:38,894 --> 00:09:42,718
It actually go and see this kind or
egoistically got start, so I'm going to

163
00:09:42,718 --> 00:09:46,694
give them that and what is that actually
change ego's perception, how would it

164
00:09:46,694 --> 00:09:49,040
actually spreads that thing in entire
network.

165
00:09:49,200 --> 00:09:53,202
And I, that, that, that thing has been
growing immensely, much more than we

166
00:09:53,202 --> 00:09:58,758
actually even anticipated.
And people are really in tune to that kind

167
00:09:58,758 --> 00:10:02,926
of thing.
So where do you get the expertise?

168
00:10:02,926 --> 00:10:05,446
Oh yeah, yeah.
Only for you.

169
00:10:05,446 --> 00:10:11,747
[laugh] Off the record.
Only for people watching this, no one else

170
00:10:11,747 --> 00:10:17,175
is gonna know.
So you can do a few things, you can take a

171
00:10:17,175 --> 00:10:24,930
look at profile information, and be like
okay, you know, based on the descriptions

172
00:10:25,705 --> 00:10:29,194
to be like hey What do my friends know?
What do my connections know?

173
00:10:29,299 --> 00:10:30,001
Mm-hm.
That can...

174
00:10:30,106 --> 00:10:30,527
Mm-hm.
...

175
00:10:30,527 --> 00:10:33,930
Basically drive what I know.
And then you could also do the same thing

176
00:10:33,930 --> 00:10:37,157
about your company, your title.
And you bring all those pieces of

177
00:10:37,157 --> 00:10:40,233
information together.
And you could have a pretty, or a decent

178
00:10:40,233 --> 00:10:43,208
rep-, representation of what, some skills
someone might know.

179
00:10:43,208 --> 00:10:45,225
Right?
And it doesn't have to be perfect.

180
00:10:45,376 --> 00:10:47,544
But then, you know, you present is as a
user.

181
00:10:47,544 --> 00:10:51,326
And then they kind of define that.
And it's been wildly successful, using

182
00:10:51,326 --> 00:10:53,192
that kind of approach.
Mm hm, mm hm.

183
00:10:53,192 --> 00:10:55,461
Cool.
So kind of friend sourcing, expertise.

184
00:10:55,612 --> 00:10:58,184
Yeah, sort of...
It's sort of, like, playing on the

185
00:10:58,184 --> 00:10:59,344
Monopoly thing.
Yeah.

186
00:10:59,495 --> 00:11:02,420
Like, you've got people with that skill
things and sort of.

187
00:11:02,570 --> 00:11:06,539
Clustered together on top of it were
professional networks that you would have

188
00:11:06,539 --> 00:11:07,895
those connections.
Right.

189
00:11:08,046 --> 00:11:12,734
And you let them.
I just think that, that source of data.

190
00:11:13,296 --> 00:11:14,325
Huh.
Huh.

191
00:11:14,325 --> 00:11:21,994
So, I have done a little bit of research
of what people say publicly and what they

192
00:11:21,994 --> 00:11:24,052
say privately?
[laugh].

193
00:11:24,052 --> 00:11:30,880
Have you thought about that at all?
Publicly I say that Igor is an expert.

194
00:11:30,880 --> 00:11:37,241
[laugh] And privately?
[laugh] It goes back to one other question

195
00:11:37,241 --> 00:11:41,935
that you had earlier.
The, The difference is that everything

196
00:11:41,935 --> 00:11:46,197
that you do on Linkdin is, is tailored.
I mean, it, it has, it has your voice to

197
00:11:46,197 --> 00:11:49,394
it and you're identified through it.
You're not anonymous.

198
00:11:49,562 --> 00:11:52,311
Mm-hm.
You're not using a, a weird kind of name

199
00:11:52,311 --> 00:11:53,264
on that.
Mm-hm.

200
00:11:53,432 --> 00:11:58,144
It's really you with your name and that I
think effects what, what knowledge goes

201
00:11:58,144 --> 00:11:59,770
through the network.
Mm-hm.

202
00:11:59,938 --> 00:12:04,257
It effects how you express it and it
effects how you are not expressing some

203
00:12:04,257 --> 00:12:09,047
things.
Which makes it very easy to sort of in

204
00:12:09,047 --> 00:12:15,254
front of it. At least for you guys you
might be able to make some of princes when

205
00:12:15,254 --> 00:12:22,536
people don't endorse, [laugh], [laugh]
that would be a Yes, we cou, I would

206
00:12:22,536 --> 00:12:26,781
think, I would care, I would think about a
couple of odd ones.

207
00:12:26,993 --> 00:12:31,383
What?
The, the really big difference is that On

208
00:12:31,383 --> 00:12:35,081
your profile, you get full control about
what it is out there.

209
00:12:35,081 --> 00:12:39,627
So if I'm saying you're not good at
something well, you get control whether

210
00:12:39,627 --> 00:12:43,022
you want to show it or not which probably
we won't want.

211
00:12:43,022 --> 00:12:47,750
But then on the other hand a lot of
targeting that we have or recommendations

212
00:12:47,750 --> 00:12:50,539
could take that into consideration.
[laugh].

213
00:12:50,539 --> 00:12:54,479
Okay so this is part one.
You said that there are two things that

214
00:12:54,479 --> 00:12:57,873
went viral.
[laugh] The other one is, we just launched

215
00:12:57,873 --> 00:13:01,283
a way for.
For having richer content on the site.

216
00:13:01,283 --> 00:13:07,641
So normally you would have like a small
almost a network feed which is about 140

217
00:13:07,641 --> 00:13:11,811
to a bit more characters.
And we give the opportunity to some

218
00:13:11,811 --> 00:13:17,007
individuals to say, "Well, you are an
influential in this domain, why don't you

219
00:13:17,212 --> 00:13:20,084
use a longer post?" Almost like a blog
type?

220
00:13:20,289 --> 00:13:23,639
Uh-huh.
And again, it's standard, it's, it's, it's

221
00:13:23,639 --> 00:13:28,014
keyed on, on your identity.
And we ask about 140, 150, influentials

222
00:13:28,014 --> 00:13:29,012
to.
To message.

223
00:13:29,012 --> 00:13:33,914
And these are, from one side, the, the
Romneys and the Obamas today.

224
00:13:33,914 --> 00:13:37,518
To the Richard Branson, who's the most
popular one.

225
00:13:37,734 --> 00:13:42,420
To, the head of the world bank, to, a
bunch of individuals that are.

226
00:13:42,420 --> 00:13:48,195
Whose opinion you might want to, to read
from time to time, and you hope from them

227
00:13:48,195 --> 00:13:52,764
to get the very good content part.
So that's another one that's not well.

228
00:13:52,764 --> 00:13:56,530
You see, there's some individuals that we
follow this person, that person.

229
00:13:56,687 --> 00:13:59,355
Um-hm.
To get access, to their, to their person.

230
00:13:59,512 --> 00:14:01,971
Um-hm.
Or to get dynamically access to their

231
00:14:01,971 --> 00:14:04,481
person because you call always go on the
public.

232
00:14:04,481 --> 00:14:08,457
But to dynamically hear that to be
notified and that has a different way to

233
00:14:08,457 --> 00:14:11,805
progress through the network.
Cause I follow that say you or him.

234
00:14:11,962 --> 00:14:16,408
Then you say oh he's following him and her
and maybe I should check what he's saying.

235
00:14:16,565 --> 00:14:18,919
Um-hm.
And, and that has a different way to

236
00:14:18,919 --> 00:14:21,430
diffuse as compared to the previous one.
Okay.

237
00:14:21,649 --> 00:14:22,892
Yeah.
Sounds neat.

238
00:14:22,892 --> 00:14:29,035
So what's diffussing is the polymer.
Yes As opposed to this activity which you

239
00:14:29,035 --> 00:14:31,741
are still curating.
You're Correct.

240
00:14:31,960 --> 00:14:36,105
You're identifying.
One of the high level if you look at it,

241
00:14:36,105 --> 00:14:38,732
the posting on an infant here.
It follows.

242
00:14:38,732 --> 00:14:41,615
It's something new, it's the follower
posting.

243
00:14:41,615 --> 00:14:45,716
The other one is the endorsement.
Is there is such a thing as an

244
00:14:45,716 --> 00:14:47,189
endorsement?
What is it?

245
00:14:47,189 --> 00:14:51,290
They knew it was before.
What are these skills thing, that I knew

246
00:14:51,290 --> 00:14:55,967
what I had been, that was before?
They are both very, especially today, it's

247
00:14:55,967 --> 00:15:00,388
been, what, about three weeks, four weeks
since we started endorsements?

248
00:15:00,581 --> 00:15:04,233
Yeah, three weeks.
And by in front here is about a week.

249
00:15:04,233 --> 00:15:08,960
So, today you are really seeing this.
This thing explode through the network.

250
00:15:09,133 --> 00:15:09,824
Mm-hm.
Mm-hm.

251
00:15:09,824 --> 00:15:14,550
And, I guess identifying influences here.
You mentioned, you know, maybe, maybe you

252
00:15:14,550 --> 00:15:18,066
can use these endorsements?
And, and some fun to you actually.

253
00:15:18,066 --> 00:15:20,832
[laugh].
Well. How does one do that?

254
00:15:20,832 --> 00:15:22,676
[laugh].
As it always happens.

255
00:15:22,849 --> 00:15:25,673
Uh-huh.
Sometimes we're wildly more successful

256
00:15:25,673 --> 00:15:27,575
than we thought we were.
Mm-hm.

257
00:15:27,748 --> 00:15:30,802
Going to be.
So we were rather careful to create very

258
00:15:30,802 --> 00:15:32,589
well at the beginning.
Mm-hm.

259
00:15:32,762 --> 00:15:35,837
So we went through, a high level of
editorial review.

260
00:15:35,837 --> 00:15:39,592
So there's just going to be a few of them,
and there were a couple of things that

261
00:15:39,592 --> 00:15:42,327
needed to happen.
One is that they were willing to actually,

262
00:15:42,327 --> 00:15:46,221
do this porting, so that it seems dynamic,
and not that I'm not posting one every

263
00:15:46,221 --> 00:15:46,731
year.
Mm.

264
00:15:46,870 --> 00:15:51,117
And that you have this mechanism.
And then what we did is, well let's get

265
00:15:51,117 --> 00:15:54,954
ready for that second wave.
If it works well, let's figure it out.

266
00:15:54,954 --> 00:15:59,690
And there we used multiple signals to
figure out who is a potential influencer.

267
00:15:59,848 --> 00:16:03,067
Within the network.
Part of the signals that we were using is

268
00:16:03,067 --> 00:16:05,917
the skills map.
Part of the signal that we're using is

269
00:16:05,917 --> 00:16:09,981
that people are sharing content at
LinkedIn, and what happens to that content

270
00:16:09,981 --> 00:16:12,355
at issue.
And people share content within the

271
00:16:12,355 --> 00:16:16,050
LinkedIn share button that's under where.
What happens to that content?

272
00:16:16,050 --> 00:16:19,533
It sounds just being static, or if some
where else being picked up.

273
00:16:19,533 --> 00:16:23,597
So they were already acting like in
trances without them really knowing that

274
00:16:23,597 --> 00:16:27,977
they work because they're have the crowd
that would actually follow what they, what

275
00:16:27,977 --> 00:16:29,191
they click on.
Mm hm.

276
00:16:29,349 --> 00:16:30,880
What they acted upon.
Okay.

277
00:16:31,684 --> 00:16:34,797
Ooh.
So, so Yeah we'll, we'll be getting to

278
00:16:34,797 --> 00:16:38,193
the, I'm sorry, I smacked you.
[laugh].

279
00:16:40,245 --> 00:16:45,056
So I remember you talking in Dublin some.
A few months ago now.

280
00:16:45,056 --> 00:16:51,000
About predicting where peop-, where people
are going to move next for their next job.

281
00:16:51,000 --> 00:16:54,467
And somehow also incorporating geography.
So...

282
00:16:54,679 --> 00:16:56,448
Mm hm.
[laugh] Uh...

283
00:16:56,660 --> 00:16:58,783
Should we sample this?
[laugh].

284
00:16:59,980 --> 00:17:07,227
The interesting thing that we have done
these maps for the US. Now I'm invited to

285
00:17:07,227 --> 00:17:09,222
go to Dublin, and say oh, I don't have the
maps for Europe And fortunately enough,

286
00:17:09,222 --> 00:17:17,310
and even after the meeting, some of the
people who weren't in the audience.

287
00:17:17,524 --> 00:17:23,305
Labor guru of something in Europe.
What we were doing is we were like well,

288
00:17:23,305 --> 00:17:28,801
from your resume, I know where you're
working and where you are going to work.

289
00:17:28,801 --> 00:17:33,868
And from that I can infer maps of are you
speaking to a certain region?

290
00:17:33,868 --> 00:17:37,722
Are you willing to get a job opening to
another place?

291
00:17:37,722 --> 00:17:42,219
If you look at it says well the migration
pattern in real life.

292
00:17:42,219 --> 00:17:48,000
And we use part of that signal to figure
out which job I should recommend to you.

293
00:17:48,000 --> 00:17:49,000
It's.
Mm-hm.

294
00:17:49,154 --> 00:17:52,715
It's one of the things that I wish we
would do much more on Linkedin.

295
00:17:52,869 --> 00:17:55,501
Much more looking into this labor market.
Uh-huh.

296
00:17:55,810 --> 00:17:58,803
One thing that the reason we talk about
was the skill gap.

297
00:17:58,803 --> 00:18:02,466
How come some people are not employed when
we have so many job opening?

298
00:18:02,621 --> 00:18:05,046
Oh.
Is there a gap between the, the skills of

299
00:18:05,046 --> 00:18:07,729
what you need for the job compared to what
you have.

300
00:18:07,884 --> 00:18:09,587
Mm-hm.
Another part is well.

301
00:18:09,742 --> 00:18:12,425
I thinking of taking some Coursera
classes, right?

302
00:18:12,425 --> 00:18:13,869
[laugh] No.
That actually.

303
00:18:13,869 --> 00:18:16,707
No seriously.
That's one of the things we are trying to

304
00:18:16,707 --> 00:18:18,720
figure out.
Well if there is such a gap.

305
00:18:18,720 --> 00:18:22,874
How do you,
Provide, that is organizations with

306
00:18:22,874 --> 00:18:25,781
individuals.
See I', m looking for a job x, and there,

307
00:18:25,781 --> 00:18:30,340
well there's not that many job x on the
market, but there's but this is job y on

308
00:18:30,340 --> 00:18:34,900
the market which is close to the job x,
but these are the skills that you missed.

309
00:18:35,073 --> 00:18:38,824
How do I get these guys?
Well I can work on this, but I cannot get

310
00:18:38,824 --> 00:18:40,728
the job.
So I cannot work on this.

311
00:18:40,728 --> 00:18:43,787
So how do I get it?
Well, classes, courses, books, online

312
00:18:43,787 --> 00:18:45,287
this, online that.
Mm-hm.

313
00:18:45,460 --> 00:18:49,267
And, and provide a way to actually get
that, or how to get internships.

314
00:18:49,408 --> 00:18:51,711
Or maybe you can discover stud companies.
[laugh].

315
00:18:51,852 --> 00:18:54,297
[laugh].
So, maybe you can't just directly go from

316
00:18:54,297 --> 00:18:56,037
[CROSSTALK].
Say now, wait a minute!

317
00:18:56,037 --> 00:18:57,588
I'm going to mark that down.
[laugh].

318
00:18:57,729 --> 00:18:58,616
[laugh].
That,s.

319
00:18:58,622 --> 00:19:02,196
Yeah, I think it's really interesting in
the context of veterans for example.

320
00:19:02,196 --> 00:19:05,722
Like we have, you know, we were open up
these things where, you know, veterans

321
00:19:05,722 --> 00:19:09,201
come they, they leave the military.
They almost have the skills but what

322
00:19:09,201 --> 00:19:12,021
skills do those map to in, you know, in
non-military contacts?

323
00:19:12,021 --> 00:19:15,689
And, you know, how, like what can a radio
operator in the military, what, what are

324
00:19:15,689 --> 00:19:20,014
they really qualified to do and how can we
help them get a job which can more utilize

325
00:19:20,014 --> 00:19:22,130
the type of skills they used in the
military?

326
00:19:22,323 --> 00:19:27,551
That was actually really interesting kind
of thing to look at, and deep dive into.

327
00:19:27,745 --> 00:19:28,520
Mm-hm.
Mm-hm.

328
00:19:28,943 --> 00:19:35,011
So I guess there's still this question of,
the textbook social network analysis and

329
00:19:35,011 --> 00:19:40,726
then social network analysis that scales
to the amount of data that you have and

330
00:19:40,726 --> 00:19:46,018
you've said a little bit about it.
Is there anything more that the students

331
00:19:46,018 --> 00:19:49,122
might want to brace themselves for?
You know?

332
00:19:49,122 --> 00:19:50,463
[laugh].
[laugh].

333
00:19:50,674 --> 00:19:53,991
Oh, yea.
I think almost everything you read in a

334
00:19:53,991 --> 00:19:58,609
textbook will break down in scale.
We've noticed that quite a bit, I think.

335
00:19:58,768 --> 00:20:02,996
I think a good, classical example of that
on LinkedIn is we have this people you

336
00:20:02,996 --> 00:20:07,383
might know feature, which, you know, tries
to go and find, the relationships on the,

337
00:20:07,383 --> 00:20:11,083
the social networking platform.
And you might think, hey, I'm just gonna

338
00:20:11,083 --> 00:20:14,783
go and, like, look at second degree, like,
close ties and figure that out.

339
00:20:14,783 --> 00:20:18,483
It turns out that really hard to do when
you have 185 million members.

340
00:20:18,641 --> 00:20:21,178
And.
Actually doing these things at scale and

341
00:20:21,178 --> 00:20:24,772
actually figuring out these well, these,
actually was very difficult.

342
00:20:24,772 --> 00:20:27,680
And it was very humbling.
I mean, words like ten million.

343
00:20:27,680 --> 00:20:31,177
It quickly breaks down when you get to 50
million or a 100 million things.

344
00:20:31,324 --> 00:20:35,166
And you can think of different kind of
approaches to do some of this stuff.

345
00:20:35,166 --> 00:20:39,205
Some of this can be parallelized, but you
know, a lot of this stuff actually cannot.

346
00:20:39,205 --> 00:20:42,900
So you have to think maybe about
approximations or other things you can do.

347
00:20:43,081 --> 00:20:45,986
Or really put domain knowledge into the
problem,

348
00:20:45,986 --> 00:20:49,799
To really understand how you can get these
algorithms to scale.

349
00:20:49,981 --> 00:20:52,946
You know, quadratic is not too really good
for us.

350
00:20:52,946 --> 00:20:56,517
[laugh] Just to say that reading something
in a book.

351
00:20:56,517 --> 00:20:59,785
I am going to a course to manage this.
[laugh] Yeah.

352
00:20:59,785 --> 00:21:03,780
Yeah, I am not saying don't.
Yeah, the theory is really important to

353
00:21:03,780 --> 00:21:06,927
understand.
To basically understand the applications

354
00:21:06,927 --> 00:21:09,892
of it.
And then you know to really distill it and

355
00:21:09,892 --> 00:21:14,673
deeper dive into these problems.
The way that I was viewing it is Being a

356
00:21:14,673 --> 00:21:17,922
statistician, the background is a bit
math, so I started as a mathematician,

357
00:21:17,922 --> 00:21:19,064
then a statistician.
Mm.

358
00:21:19,196 --> 00:21:20,952
And I'm now dealing with money.
Mm-hm.

359
00:21:21,084 --> 00:21:23,500
And there's a fundamental difference
between a, b and c.

360
00:21:23,710 --> 00:21:26,515
Mm-hm.
The math is, is geared into, the

361
00:21:26,515 --> 00:21:29,459
mathematical model.
Axioms, theorem, Lima.

362
00:21:29,459 --> 00:21:32,474
Everything is true or false in some
extent.

363
00:21:32,685 --> 00:21:35,139
At static fusion, nothing is true or
false.

364
00:21:35,139 --> 00:21:39,486
Everything is uncertain.
But you still use math as the, the, the,

365
00:21:39,486 --> 00:21:45,025
the tools that gets you towards, you know,
what is optimum, what isn't optimum.

366
00:21:45,025 --> 00:21:49,468
And you understand the methodology.
Then you go to, billions of data, or

367
00:21:49,468 --> 00:21:52,503
millions of this.
And suddenly, even doing a median is very

368
00:21:52,503 --> 00:21:54,753
complicated.
And you go back to the average.

369
00:21:54,753 --> 00:21:58,259
And you say, well, I know that the median
is better as a point is in with the

370
00:21:58,259 --> 00:22:01,504
average, but, hey, you know what, I, I can
only do one average fast enough.

371
00:22:01,661 --> 00:22:03,492
Mm-hm.
So I have to live with it.

372
00:22:03,492 --> 00:22:07,522
But then, at that point of time, what you
do, in the back of your mind, you know

373
00:22:07,522 --> 00:22:11,132
what is, what is a fundamental assumptions
of different techniques.

374
00:22:11,132 --> 00:22:13,696
And you can use them and tweak them a
little bit.

375
00:22:13,696 --> 00:22:17,360
So you know you are not doing the perfect
thing, but at scale, you have.

376
00:22:17,360 --> 00:22:21,670
No options but you know what is the impact
of the change.

377
00:22:21,824 --> 00:22:24,041
And you can maybe cry for that bias or
not.

378
00:22:24,195 --> 00:22:26,721
Mm-hm.
Because their not on part time, but they

379
00:22:26,721 --> 00:22:33,262
are just way too big and way too much.
Well, so, those were some of the

380
00:22:33,262 --> 00:22:37,704
challenges.
Any, anything else that we may have

381
00:22:37,704 --> 00:22:45,940
missed?
There is a beauty on scale.

382
00:22:46,459 --> 00:22:50,806
It's not, we, we shouldn't be afraid of
scale, but, did it, it's really

383
00:22:50,806 --> 00:22:54,960
fascinating, to see things that when they
have ton of traffic.

384
00:22:54,960 --> 00:22:59,943
Because its, its makes it, it gives it its
dynamism, it gives it its diffusion

385
00:22:59,943 --> 00:23:04,992
signature and because it had skill, the
other duties are I think its hard and

386
00:23:04,992 --> 00:23:09,975
because its hard its interesting.
So I think its a chance that people should

387
00:23:09,975 --> 00:23:14,036
pick up and up your favor.
Get tons of things that would fail but

388
00:23:14,036 --> 00:23:16,618
it's a very, very fascinating thing to
play on.

389
00:23:16,786 --> 00:23:20,602
Yeah, and the other thing is you can learn
different things at scale.

390
00:23:20,602 --> 00:23:24,811
Things that, you know, with a small amount
of data becomes very difficult to

391
00:23:24,811 --> 00:23:29,131
understand, but when you have a large
amount of data such as LinkedIn has, you

392
00:23:29,131 --> 00:23:32,050
can see patterns that you would not
necessarily see.

393
00:23:32,222 --> 00:23:35,506
Which is actually, I think a, very
interesting angle,

394
00:23:35,506 --> 00:23:37,926
But. Mm-hm.
So really large things under the

395
00:23:37,926 --> 00:23:39,020
microscope.
Yeah.

396
00:23:39,193 --> 00:23:39,942
Yeah.
Exactly.

397
00:23:40,115 --> 00:23:42,189
Yeah.
That's a good way to put it.

398
00:23:42,361 --> 00:23:45,100
Yeah.
And the other thing, these are fascinating

399
00:23:45,100 --> 00:23:50,371
times to play with that there.
Some years ago even if you thought about

400
00:23:50,371 --> 00:23:53,935
doing it, you couldn't.
Just like friends to fill.

401
00:23:53,935 --> 00:23:59,874
It's the right time to actually, it's a
new door to open up to what we can think

402
00:23:59,874 --> 00:24:02,402
about.
Cool. well, maybe we can end with that.

403
00:24:02,402 --> 00:24:04,534
Thanks very much.
Keep your eye on Sam.

404
00:24:04,534 --> 00:24:05,825
Thank you.
Thank you.