WEBVTT

1
00:00:00.000 --> 00:00:03.480


2
00:00:03.480 --> 00:00:05.070
SAM RANSBOTHAM:

3
00:00:05.070 --> 00:00:05.160
How can AI help bring
new ideas and products

4
00:00:05.160 --> 00:00:05.250
to market in industries where
risk aversion is rampant?

5
00:00:05.250 --> 00:00:05.350
Find out today when we talk
with Kartik Hosanagar, professor

6
00:00:05.350 --> 00:00:05.430
at the Wharton School
and founder of Jumpcut,

7
00:00:05.430 --> 00:00:05.490
a startup helping previously
undiscovered talent

8
00:00:05.490 --> 00:00:06.407
produce movies and tv.

9
00:00:06.407 --> 00:00:11.780
Welcome to Me, Myself,
and AI, a podcast

10
00:00:11.780 --> 00:00:21.760
on artificial
intelligence in business.

11
00:00:21.760 --> 00:00:26.110


12
00:00:26.110 --> 00:00:29.480
Each episode, we introduce you
to someone innovating with AI.

13
00:00:29.480 --> 00:00:32.800
I'm Sam Ransbotham, professor
of information systems

14
00:00:32.800 --> 00:00:34.140
at Boston College.

15
00:00:34.140 --> 00:00:37.510
I'm also the guest editor for
the AI and Business Strategy

16
00:00:37.510 --> 00:00:40.880
Big Ideas program at MIT
Sloan Management Review.

17
00:00:40.880 --> 00:00:43.580
SHERVIN KHODABANDEH: And
I'm Shervin Khodabandeh,

18
00:00:43.580 --> 00:00:47.740
senior partner with BCG, and
I colead BCG's AI practice

19
00:00:47.740 --> 00:00:48.920
in North America.

20
00:00:48.920 --> 00:00:52.450
Together, MIT SMR
and BCG have been

21
00:00:52.450 --> 00:00:55.870
researching AI for five
years, interviewing hundreds

22
00:00:55.870 --> 00:00:58.220
of practitioners and
surveying thousands

23
00:00:58.220 --> 00:01:02.320
of companies on what it takes
to build and to deploy and scale

24
00:01:02.320 --> 00:01:05.850
AI capabilities and
really transform

25
00:01:05.850 --> 00:01:08.040
the way organizations operate.

26
00:01:08.040 --> 00:01:11.940
SAM RANSBOTHAM: Today, we're
talking with Kartik Hosanagar.

27
00:01:11.940 --> 00:01:14.230
He's a professor at The
Wharton School and founder

28
00:01:14.230 --> 00:01:15.610
and CEO of Jumpcut.

29
00:01:15.610 --> 00:01:16.960
Kartik, thanks for joining us.

30
00:01:16.960 --> 00:01:17.720
Welcome.

31
00:01:17.720 --> 00:01:19.770
KARTIK HOSANAGAR: Thanks for
having me, Sam and Shervin.

32
00:01:19.770 --> 00:01:21.520
SAM RANSBOTHAM: So this is a
bit of a different interview

33
00:01:21.520 --> 00:01:22.940
for me, because
I've known Kartik

34
00:01:22.940 --> 00:01:24.730
for years in academic circles.

35
00:01:24.730 --> 00:01:27.790
But Kartik, your latest
venture is Jumpcut.

36
00:01:27.790 --> 00:01:29.840
I think the idea
basically is to help

37
00:01:29.840 --> 00:01:32.700
surface new and fresh
stories for Hollywood.

38
00:01:32.700 --> 00:01:34.470
Can you tell us a
little bit about that?

39
00:01:34.470 --> 00:01:36.630
KARTIK HOSANAGAR: My new
startup is called Jumpcut,

40
00:01:36.630 --> 00:01:37.500
as you mentioned.

41
00:01:37.500 --> 00:01:39.740
What we are doing is
essentially trying

42
00:01:39.740 --> 00:01:43.140
to create a new
data-driven studio that's

43
00:01:43.140 --> 00:01:45.810
reimagining the way
films and TV shows

44
00:01:45.810 --> 00:01:48.160
are developed, with
the specific goal

45
00:01:48.160 --> 00:01:51.160
of elevating fresh new voices.

46
00:01:51.160 --> 00:01:55.380
The context here that got
me interested in this is,

47
00:01:55.380 --> 00:01:58.910
Hollywood has historically
been an old boys' club --

48
00:01:58.910 --> 00:02:03.820
a few execs making decisions
on what movies get made,

49
00:02:03.820 --> 00:02:06.750
who's in those movies,
[and] at what budgets.

50
00:02:06.750 --> 00:02:09.850
All of these [are] just
pretty much based on gut

51
00:02:09.850 --> 00:02:12.550
and relationships
-- who knows who.

52
00:02:12.550 --> 00:02:15.410
There are costs [to] this
kind of decision-making.

53
00:02:15.410 --> 00:02:18.870
There's the economic cost;
Hollywood has historically had

54
00:02:18.870 --> 00:02:20.650
a very poor batting average.

55
00:02:20.650 --> 00:02:24.560
There's the social cost;
by just about any measure,

56
00:02:24.560 --> 00:02:27.190
Hollywood has not been
a particularly inclusive

57
00:02:27.190 --> 00:02:28.090
industry.

58
00:02:28.090 --> 00:02:30.680
And then there's the
cost to audiences.

59
00:02:30.680 --> 00:02:33.790
So, what we're
trying to do is break

60
00:02:33.790 --> 00:02:37.780
that mold and use data to
make more objective, better

61
00:02:37.780 --> 00:02:38.820
decisions.

62
00:02:38.820 --> 00:02:42.900
But ultimately,
by doing that, we

63
00:02:42.900 --> 00:02:45.800
can assess storytellers
and stories on their merit

64
00:02:45.800 --> 00:02:50.310
as opposed to who is connected
to whom or just [the]

65
00:02:50.310 --> 00:02:52.680
gut feelings of a few people.

66
00:02:52.680 --> 00:02:55.930
So that's how we're
trying to democratize

67
00:02:55.930 --> 00:02:57.343
Hollywood using data.

68
00:02:57.343 --> 00:02:59.010
SAM RANSBOTHAM: I
mean, maybe your world

69
00:02:59.010 --> 00:03:02.030
is slightly different than
mine, but no movie execs

70
00:03:02.030 --> 00:03:04.060
are in my classes.

71
00:03:04.060 --> 00:03:05.790
So how did you end
up with this idea,

72
00:03:05.790 --> 00:03:08.210
and how did you end
up with Jumpcut?

73
00:03:08.210 --> 00:03:09.790
KARTIK HOSANAGAR:
It's interesting --

74
00:03:09.790 --> 00:03:12.480
no movie execs in my
classes either, Sam.

75
00:03:12.480 --> 00:03:13.990
SAM RANSBOTHAM:
Do you know that I

76
00:03:13.990 --> 00:03:16.080
went to high school with
Julia Roberts, though?

77
00:03:16.080 --> 00:03:18.000
I mean, there's a
connection there.

78
00:03:18.000 --> 00:03:18.750
But anyway, sorry.

79
00:03:18.750 --> 00:03:19.310
Keep going.

80
00:03:19.310 --> 00:03:21.570
KARTIK HOSANAGAR: Well, I
hope you are still connected

81
00:03:21.570 --> 00:03:22.190
to your high school --

82
00:03:22.190 --> 00:03:23.607
SAM RANSBOTHAM:
No, unfortunately;

83
00:03:23.607 --> 00:03:26.190
this is pre-social media, and
I made lots of tragic mistakes

84
00:03:26.190 --> 00:03:28.310
in that era, but anyway.

85
00:03:28.310 --> 00:03:30.458
Sad tale for another time.

86
00:03:30.458 --> 00:03:31.750
KARTIK HOSANAGAR: Right, right.

87
00:03:31.750 --> 00:03:34.350
I look forward to catching
up on that story sometime.

88
00:03:34.350 --> 00:03:38.130
Coming back to your question:
What got me interested in this?

89
00:03:38.130 --> 00:03:42.860
I've always had an interest
in content and storytelling

90
00:03:42.860 --> 00:03:43.810
and films.

91
00:03:43.810 --> 00:03:45.900
I'm actually an
amateur filmmaker.

92
00:03:45.900 --> 00:03:51.630
So back when I was a
newly tenured professor,

93
00:03:51.630 --> 00:03:55.400
and also, I didn't have kids, it
was an interesting combination

94
00:03:55.400 --> 00:03:57.980
where I had my
weekends to myself,

95
00:03:57.980 --> 00:04:01.880
so I would go make short
films and just put them out

96
00:04:01.880 --> 00:04:02.630
on YouTube.

97
00:04:02.630 --> 00:04:05.300
It was really just
a fun hobby of mine.

98
00:04:05.300 --> 00:04:07.860
In fact, during my first
sabbatical from Wharton,

99
00:04:07.860 --> 00:04:09.050
I wrote a screenplay.

100
00:04:09.050 --> 00:04:12.630
I flew into Mumbai, met with
a bunch of film producers

101
00:04:12.630 --> 00:04:14.630
there, and pitched
my screenplay.

102
00:04:14.630 --> 00:04:17.019
A number of them liked it.

103
00:04:17.019 --> 00:04:21.019
But the common response
from many of them

104
00:04:21.019 --> 00:04:23.190
was, "We really
like this, but how

105
00:04:23.190 --> 00:04:26.773
do we take a bet on a
completely new writer/director?"

106
00:04:26.773 --> 00:04:28.440
One of them said,
"I'll buy your script,

107
00:04:28.440 --> 00:04:29.962
but I can't have you direct it.

108
00:04:29.962 --> 00:04:32.170
You may be good, but I just
can't have you direct it,

109
00:04:32.170 --> 00:04:36.030
because I can't go to
financiers and to actors

110
00:04:36.030 --> 00:04:39.330
and ask them to take
a bet on a new voice."

111
00:04:39.330 --> 00:04:43.280
So my response back then was,
"Fair enough; makes sense.

112
00:04:43.280 --> 00:04:44.840
I realize why you can't do it."

113
00:04:44.840 --> 00:04:47.870
But at the same time, I didn't
want to give my script away.

114
00:04:47.870 --> 00:04:49.520
So I came back.

115
00:04:49.520 --> 00:04:52.780
Then my sabbatical was
over, so back to Wharton --

116
00:04:52.780 --> 00:04:55.210
back to teaching
and all of that.

117
00:04:55.210 --> 00:04:59.260
But over the years, I've
met with so many writers

118
00:04:59.260 --> 00:05:02.910
and directors who have
shared similar perspectives.

119
00:05:02.910 --> 00:05:04.770
There's a friend
of mine who told me

120
00:05:04.770 --> 00:05:08.130
it took him 15 years to
break into the industry.

121
00:05:08.130 --> 00:05:10.280
And he's a successful
writer/director.

122
00:05:10.280 --> 00:05:12.500
Fifteen years for his
first opportunity.

123
00:05:12.500 --> 00:05:15.300
And I hear this so many times.

124
00:05:15.300 --> 00:05:20.070
In recent years, we see really
successful movies or shows

125
00:05:20.070 --> 00:05:22.860
like Get Out or
Stranger Things come

126
00:05:22.860 --> 00:05:26.190
from very unexpected
places, so I've always

127
00:05:26.190 --> 00:05:28.370
been fascinated by this space.

128
00:05:28.370 --> 00:05:32.170
Initially it was a hobby, but
one fine day I kind of felt

129
00:05:32.170 --> 00:05:36.230
like the problem I'm
seeing is something

130
00:05:36.230 --> 00:05:40.893
my skills and data might have
some relevance to use to solve.

131
00:05:40.893 --> 00:05:43.060
SHERVIN KHODABANDEH: It's
actually quite interesting

132
00:05:43.060 --> 00:05:46.230
because there are several
angles to the story here.

133
00:05:46.230 --> 00:05:49.790
One is the equity angle
and the social angle,

134
00:05:49.790 --> 00:05:54.660
and giving those who
deserve and have the merit

135
00:05:54.660 --> 00:05:57.450
the voice that they
might otherwise not get.

136
00:05:57.450 --> 00:06:02.100
And then there is the more
business, data-driven angle,

137
00:06:02.100 --> 00:06:08.870
which takes my mind to Moneyball
in baseball, or underwriting --

138
00:06:08.870 --> 00:06:12.910
giving credit to folks who
don't have a credit history

139
00:06:12.910 --> 00:06:17.010
because they emigrate to this
country and, but for the fact

140
00:06:17.010 --> 00:06:20.790
that they've just arrived here,
they would be perfect in terms

141
00:06:20.790 --> 00:06:23.380
of the profile and the
background and the education --

142
00:06:23.380 --> 00:06:29.930
and being able to actually
assess talent based on a bunch

143
00:06:29.930 --> 00:06:33.720
of attributes and features
and sort of de-risk the very

144
00:06:33.720 --> 00:06:37.420
executives that say, "How
do I take a bet on you,

145
00:06:37.420 --> 00:06:40.342
because I don't know
anything about you?"

146
00:06:40.342 --> 00:06:41.800
KARTIK HOSANAGAR:
Yeah, absolutely.

147
00:06:41.800 --> 00:06:44.600
And you used a very
interesting word here.

148
00:06:44.600 --> 00:06:46.170
You talked about de-risking.

149
00:06:46.170 --> 00:06:48.917
You talked about
credit applications,

150
00:06:48.917 --> 00:06:50.500
and there are people
who are applying,

151
00:06:50.500 --> 00:06:52.520
and how do we de-risk them?

152
00:06:52.520 --> 00:06:55.900
A lot of this is
really about risk.

153
00:06:55.900 --> 00:07:00.620
If you look at Hollywood, one
of the big things I've realized

154
00:07:00.620 --> 00:07:04.400
is, Hollywood would
like to take chances.

155
00:07:04.400 --> 00:07:07.120
They understand the
problems of inequity.

156
00:07:07.120 --> 00:07:10.270
They understand the industry
has not been inclusive.

157
00:07:10.270 --> 00:07:14.370
But everyone's worried
about the risk issue.

158
00:07:14.370 --> 00:07:19.310
What was interesting for me
was, a lot of execs in Hollywood

159
00:07:19.310 --> 00:07:21.200
told me, "Well, the
reason everyone's

160
00:07:21.200 --> 00:07:25.320
so risk averse is, if you do
a movie with Brad Pitt in it

161
00:07:25.320 --> 00:07:27.000
and it doesn't work
out, you won't lose

162
00:07:27.000 --> 00:07:29.250
your job for making that bet.

163
00:07:29.250 --> 00:07:34.295
But if you take a bet on
something new [and] unexpected,

164
00:07:34.295 --> 00:07:35.670
and that doesn't
work out, you'll

165
00:07:35.670 --> 00:07:38.820
have some explaining to do:
'Why did you make this bet?

166
00:07:38.820 --> 00:07:42.650
This is not the kind of bet
we've made in the past.'" And I

167
00:07:42.650 --> 00:07:44.940
remember at some point
I even read a statistic,

168
00:07:44.940 --> 00:07:47.830
which was on movies,
and it said like 74%

169
00:07:47.830 --> 00:07:53.400
of the highest-grossing movies
are all sequels or adaptations

170
00:07:53.400 --> 00:07:56.200
of existing IP
[intellectual property].

171
00:07:56.200 --> 00:07:59.380
So there's no one willing to
take a chance on something new.

172
00:07:59.380 --> 00:08:01.070
Everything is a sequel.

173
00:08:01.070 --> 00:08:02.680
As an outsider, I
was wondering, why

174
00:08:02.680 --> 00:08:05.030
is it that it has
to be a bestselling

175
00:08:05.030 --> 00:08:08.990
book or a comic book for
something to get made?

176
00:08:08.990 --> 00:08:12.000
Why can't it be an original
screenplay or an original idea?

177
00:08:12.000 --> 00:08:14.420
And it came back to, "Oh,
yes, because [the] IP

178
00:08:14.420 --> 00:08:15.700
has been de-risked."

179
00:08:15.700 --> 00:08:19.720
But the insight that I
had was that IP isn't

180
00:08:19.720 --> 00:08:21.560
the only way of de-risking.

181
00:08:21.560 --> 00:08:24.870
There are other ways
to de-risk as well.

182
00:08:24.870 --> 00:08:29.110
So that's essentially
how we approached it.

183
00:08:29.110 --> 00:08:31.620
And there's really three
pieces, and I'll mention each

184
00:08:31.620 --> 00:08:33.539
at a high level, and
then you'll tell me

185
00:08:33.539 --> 00:08:36.690
if there's any of them that's
worth going deeper into.

186
00:08:36.690 --> 00:08:39.870
One is, first of all,
how do we discover

187
00:08:39.870 --> 00:08:41.870
stories and storytellers?

188
00:08:41.870 --> 00:08:46.700
The classic Hollywood approach
is that studios get submissions

189
00:08:46.700 --> 00:08:50.850
of scripts from
agents, and agents

190
00:08:50.850 --> 00:08:52.840
are representing
talent and people who

191
00:08:52.840 --> 00:08:54.510
are outside of the
system; they don't

192
00:08:54.510 --> 00:08:56.390
have a godfather
who'd connect them

193
00:08:56.390 --> 00:08:58.710
or don't already
have the connections.

194
00:08:58.710 --> 00:09:00.760
So the first thing
we had to solve

195
00:09:00.760 --> 00:09:03.740
for is how do we find
stories and storytellers that

196
00:09:03.740 --> 00:09:05.590
are outside of the system?

197
00:09:05.590 --> 00:09:08.330
So for that, one
of the things we do

198
00:09:08.330 --> 00:09:13.780
is, our algorithms are assessing
content as well as creators

199
00:09:13.780 --> 00:09:18.670
on various platforms
like YouTube or Reddit,

200
00:09:18.670 --> 00:09:22.840
or even storytelling platforms
like Wattpad and others.

201
00:09:22.840 --> 00:09:25.360
And what we're doing
in those platforms --

202
00:09:25.360 --> 00:09:27.520
I'll use YouTube as an example.

203
00:09:27.520 --> 00:09:31.350
If a screenwriter
or if a director

204
00:09:31.350 --> 00:09:35.130
has created short films and has
posted those short films that

205
00:09:35.130 --> 00:09:37.510
are already resonating
with audiences,

206
00:09:37.510 --> 00:09:39.320
then we try and discover them.

207
00:09:39.320 --> 00:09:41.940
So the idea is to find
people where they are.

208
00:09:41.940 --> 00:09:44.230
That's whatever country,
whatever platform you're

209
00:09:44.230 --> 00:09:46.940
in, and go discover them there.

210
00:09:46.940 --> 00:09:49.230
And essentially, the
algorithms and analytics

211
00:09:49.230 --> 00:09:51.680
are trying to
analyze the content

212
00:09:51.680 --> 00:09:55.270
to look for high
production value, which

213
00:09:55.270 --> 00:09:59.160
you can infer from the frames
and images in the videos,

214
00:09:59.160 --> 00:10:00.050
for example.

215
00:10:00.050 --> 00:10:02.570
You're looking for
strong storytelling

216
00:10:02.570 --> 00:10:04.400
and strong audience
response, which

217
00:10:04.400 --> 00:10:06.920
you can infer from
the kinds of comments

218
00:10:06.920 --> 00:10:10.140
that are elicited in response
to these videos or even tech

219
00:10:10.140 --> 00:10:11.180
stories.

220
00:10:11.180 --> 00:10:16.070
You can short-list a set
of stories or storytellers

221
00:10:16.070 --> 00:10:18.340
for our creative
team to look at.

222
00:10:18.340 --> 00:10:21.820
So it's not AI
making the decision,

223
00:10:21.820 --> 00:10:23.630
but it's AI making
humans efficient,

224
00:10:23.630 --> 00:10:25.210
because if I had a
creative team that

225
00:10:25.210 --> 00:10:27.620
had to scour through
YouTube to find

226
00:10:27.620 --> 00:10:29.280
the needle in the
haystack, I would

227
00:10:29.280 --> 00:10:31.230
need a massive team
and several years

228
00:10:31.230 --> 00:10:34.700
to comb through 250,000
short films on YouTube.

229
00:10:34.700 --> 00:10:38.043
SHERVIN KHODABANDEH:
Or just my two kids.

230
00:10:38.043 --> 00:10:39.960
KARTIK HOSANAGAR: [Laughs.]
Well, that's true.

231
00:10:39.960 --> 00:10:40.680
That works, too.

232
00:10:40.680 --> 00:10:43.610
Except I think they will
just send me Minecraft

233
00:10:43.610 --> 00:10:45.143
videos or something like that.

234
00:10:45.143 --> 00:10:47.310
SHERVIN KHODABANDEH: Given
how many hours of YouTube

235
00:10:47.310 --> 00:10:48.010
they watch.

236
00:10:48.010 --> 00:10:48.980
But I digress.

237
00:10:48.980 --> 00:10:50.317
Sorry, go ahead.

238
00:10:50.317 --> 00:10:51.900
SAM RANSBOTHAM: Let's
interrupt there.

239
00:10:51.900 --> 00:10:54.067
I mean, that's the opposite
of the Moneyball problem

240
00:10:54.067 --> 00:10:55.040
that Shervin mentioned.

241
00:10:55.040 --> 00:10:56.930
Because with Moneyball,
you had the people,

242
00:10:56.930 --> 00:10:59.000
and it's about figuring
out which one was good.

243
00:10:59.000 --> 00:11:02.480
You're more about expanding
that search space.

244
00:11:02.480 --> 00:11:05.820
KARTIK HOSANAGAR: Before we
can say, "Among the stories,

245
00:11:05.820 --> 00:11:08.830
which is the one to bet on?"
-- before we can do that,

246
00:11:08.830 --> 00:11:12.000
we want to know, "Is the
pool of people and stories

247
00:11:12.000 --> 00:11:14.010
we're looking at the right pool?

248
00:11:14.010 --> 00:11:15.730
Or is that the complete pool?"

249
00:11:15.730 --> 00:11:18.050
If we don't expand
the pool, we are not

250
00:11:18.050 --> 00:11:19.870
solving the inclusion problem.

251
00:11:19.870 --> 00:11:21.500
We work with agents.

252
00:11:21.500 --> 00:11:24.800
We work with established
writers as well.

253
00:11:24.800 --> 00:11:29.270
But in addition, we're
actively seeking out new people

254
00:11:29.270 --> 00:11:33.240
and not waiting for them to
get discovered by an agent

255
00:11:33.240 --> 00:11:36.220
and the agent then forwards
them to us, but instead

256
00:11:36.220 --> 00:11:37.720
[we're] finding
them where they are.

257
00:11:37.720 --> 00:11:39.040
SAM RANSBOTHAM: So that was one.

258
00:11:39.040 --> 00:11:39.540
I'm sorry.

259
00:11:39.540 --> 00:11:41.262
You're going on with
two and three here.

260
00:11:41.262 --> 00:11:42.470
KARTIK HOSANAGAR: Yeah, yeah.

261
00:11:42.470 --> 00:11:45.000
And the second is, again,
how do we de-risk them?

262
00:11:45.000 --> 00:11:49.600
Once we find somebody we
ask them, "Hey, do you

263
00:11:49.600 --> 00:11:50.950
have a story for us?"

264
00:11:50.950 --> 00:11:53.180
And they don't have
one -- they have 15.

265
00:11:53.180 --> 00:11:56.440
And then we hear
them or we read them,

266
00:11:56.440 --> 00:11:58.430
and we get excited with a few.

267
00:11:58.430 --> 00:12:01.200
We want to figure out, how do
we bring in some objectivity

268
00:12:01.200 --> 00:12:01.700
to this?

269
00:12:01.700 --> 00:12:03.400
And that's where data comes in.

270
00:12:03.400 --> 00:12:06.550
And some of it is
backward looking,

271
00:12:06.550 --> 00:12:08.230
which is a classic
machine learning

272
00:12:08.230 --> 00:12:10.500
kind of approach, which
is, "Let's look at data

273
00:12:10.500 --> 00:12:13.700
on what's been doing well."

274
00:12:13.700 --> 00:12:16.210
And that could be what
movies are doing well

275
00:12:16.210 --> 00:12:18.820
but also what kinds of
stories are trending.

276
00:12:18.820 --> 00:12:21.000
It could even be search
queries and looking

277
00:12:21.000 --> 00:12:23.810
at where is the cultural
zeitgeist, where

278
00:12:23.810 --> 00:12:24.550
are people going.

279
00:12:24.550 --> 00:12:30.160
And [we try to] understand which
of the stories that we have

280
00:12:30.160 --> 00:12:34.430
are stories that we think
audiences will respond to.

281
00:12:34.430 --> 00:12:36.420
But that's not
everything, because I

282
00:12:36.420 --> 00:12:38.950
think as long as we
are backward looking

283
00:12:38.950 --> 00:12:40.840
and we're looking
at what's worked,

284
00:12:40.840 --> 00:12:42.860
it is a fundamentally
conservative approach,

285
00:12:42.860 --> 00:12:45.360
because we'll do more and more
of what's worked in the past.

286
00:12:45.360 --> 00:12:47.140
SAM RANSBOTHAM: It's
the sequel problem.

287
00:12:47.140 --> 00:12:49.005
KARTIK HOSANAGAR:
The sequel problem.

288
00:12:49.005 --> 00:12:49.880
We'll be stuck there.

289
00:12:49.880 --> 00:12:51.255
So the other
question that we are

290
00:12:51.255 --> 00:12:53.910
trying to solve for is, "How do
we go take a bet on something

291
00:12:53.910 --> 00:12:57.130
that's never been
done before and there

292
00:12:57.130 --> 00:12:59.190
is no historical success there?"

293
00:12:59.190 --> 00:13:03.270
That's where we bring in
ideas as from AB testing

294
00:13:03.270 --> 00:13:05.300
and experimentation.

295
00:13:05.300 --> 00:13:08.630
We interact with
lots of audiences

296
00:13:08.630 --> 00:13:14.240
online, where we pitch
stories, and we're

297
00:13:14.240 --> 00:13:16.990
having multiple stories
compete with each other

298
00:13:16.990 --> 00:13:20.220
and see which ones people
are gravitating toward.

299
00:13:20.220 --> 00:13:22.370
We pair the classic
machine learning

300
00:13:22.370 --> 00:13:26.410
based on past data with
digital experimentation.

301
00:13:26.410 --> 00:13:28.420
And we're running
these experiments

302
00:13:28.420 --> 00:13:32.170
and then figuring out not
just which stories we like,

303
00:13:32.170 --> 00:13:36.070
but also sometimes questions,
and this is all very

304
00:13:36.070 --> 00:13:37.780
hypothesis driven.

305
00:13:37.780 --> 00:13:41.060
And it actually also creates
very interesting breakthroughs.

306
00:13:41.060 --> 00:13:42.810
Like, one of the shows
we're working on --

307
00:13:42.810 --> 00:13:46.330
it's with a very senior
writer, in this instance --

308
00:13:46.330 --> 00:13:51.330
he had an interesting
sci-fi story,

309
00:13:51.330 --> 00:13:53.930
and it's a very
high-budget show.

310
00:13:53.930 --> 00:13:57.690
He came to us saying,
"If I have to sell this,

311
00:13:57.690 --> 00:14:00.640
it has to be the case that
this is a four-quadrant show.

312
00:14:00.640 --> 00:14:02.490
Because of the
budget, it can't be

313
00:14:02.490 --> 00:14:04.290
that this has a niche audience.

314
00:14:04.290 --> 00:14:08.050
Can your data prove that this
is a four-quadrant show?"

315
00:14:08.050 --> 00:14:10.500
And we said, "OK,
let's test it out."

316
00:14:10.500 --> 00:14:13.220
And we tested it, and it
tested off the charts.

317
00:14:13.220 --> 00:14:15.740
I went back to him
and I said, "It's

318
00:14:15.740 --> 00:14:17.370
almost a four-quadrant show.

319
00:14:17.370 --> 00:14:19.210
It's testing off the charts --

320
00:14:19.210 --> 00:14:21.040
Gen Z, Gen X, millennials.

321
00:14:21.040 --> 00:14:24.230
Doesn't matter what the age
group; they're responding."

322
00:14:24.230 --> 00:14:27.400
Sci-fi, drama, all these
different audiences:

323
00:14:27.400 --> 00:14:29.230
People who are into
sci-fi are responding,

324
00:14:29.230 --> 00:14:31.860
but also people who are not
into sci-fi or into drama,

325
00:14:31.860 --> 00:14:33.070
they're responding.

326
00:14:33.070 --> 00:14:35.070
People in the U.S.
are responding.

327
00:14:35.070 --> 00:14:36.950
People outside and in
international markets

328
00:14:36.950 --> 00:14:37.575
are responding.

329
00:14:37.575 --> 00:14:41.610
So all that's great, but
women aren't responding.

330
00:14:41.610 --> 00:14:44.420
And I said, "Well, this
is what's going on."

331
00:14:44.420 --> 00:14:45.970
And his first
reaction was, "Oh, we

332
00:14:45.970 --> 00:14:48.000
can't go with that kind
of data to the buyers,

333
00:14:48.000 --> 00:14:50.420
because that hurts the show."

334
00:14:50.420 --> 00:14:52.430
And that was our
initial reaction,

335
00:14:52.430 --> 00:14:55.810
which was to be defensive about
it: Can we hide that data?

336
00:14:55.810 --> 00:14:57.650
Can we not show it, and so on?

337
00:14:57.650 --> 00:14:59.660
And then, as we
are talking, it's

338
00:14:59.660 --> 00:15:01.310
like, "Why are women
not responding?

339
00:15:01.310 --> 00:15:02.240
What's going wrong?"

340
00:15:02.240 --> 00:15:07.470
We have this show, there's
three main characters in it.

341
00:15:07.470 --> 00:15:08.640
There's a woman in there.

342
00:15:08.640 --> 00:15:10.200
There's two male characters.

343
00:15:10.200 --> 00:15:13.180
To simplify it, I'll just
say there's the good guy,

344
00:15:13.180 --> 00:15:16.110
there's the bad guy, and then
there's the woman in the show.

345
00:15:16.110 --> 00:15:19.630
As we talked about it,
we realized, "Look,

346
00:15:19.630 --> 00:15:23.760
the woman in the show doesn't
seem to have enough agency.

347
00:15:23.760 --> 00:15:25.610
She has no motivation.

348
00:15:25.610 --> 00:15:28.540
The female character is
serving the motivations

349
00:15:28.540 --> 00:15:29.960
of the male characters."

350
00:15:29.960 --> 00:15:32.480
And he realized that
he's approaching it

351
00:15:32.480 --> 00:15:34.750
with this mindset and
he's not thought hard

352
00:15:34.750 --> 00:15:38.560
about what is driving her.

353
00:15:38.560 --> 00:15:44.080
And then he reimagined
the female character.

354
00:15:44.080 --> 00:15:46.700
I'm getting into the weeds,
but now the female character

355
00:15:46.700 --> 00:15:50.390
is a genetically CRISPR
kind of female, modified,

356
00:15:50.390 --> 00:15:53.480
and she's got superpowers
or special powers and so on.

357
00:15:53.480 --> 00:15:56.020
And suddenly, when we
tested the new version,

358
00:15:56.020 --> 00:15:57.910
women were responding to it.

359
00:15:57.910 --> 00:15:59.930
Now the idea improved.

360
00:15:59.930 --> 00:16:02.460
This isn't like soulless
data telling you

361
00:16:02.460 --> 00:16:05.720
that you need to insert a
chase sequence on page 13

362
00:16:05.720 --> 00:16:08.140
to increase the audience.

363
00:16:08.140 --> 00:16:13.350
This is hypotheses and asking,
"What is my story missing?"

364
00:16:13.350 --> 00:16:17.560
and really improving
creative decisions [that are]

365
00:16:17.560 --> 00:16:20.467
very much human led
but data informed.

366
00:16:20.467 --> 00:16:22.300
SHERVIN KHODABANDEH:
This is quite exciting.

367
00:16:22.300 --> 00:16:24.610
What I particularly
like about it, Kartik,

368
00:16:24.610 --> 00:16:28.900
is you're breaking
into the last standing

369
00:16:28.900 --> 00:16:33.240
bastion of judgment-driven
decision-making, right?

370
00:16:33.240 --> 00:16:36.210
I mean, if you think
about 20 years ago,

371
00:16:36.210 --> 00:16:40.000
[to] the loan officer making a
judgment on who's going to get

372
00:16:40.000 --> 00:16:43.290
the loan, and then
fast-forward to 10 years ago,

373
00:16:43.290 --> 00:16:47.390
where it's the retailer [who]
decides what price he or she

374
00:16:47.390 --> 00:16:51.890
should charge, or how he
should stock up the shelves --

375
00:16:51.890 --> 00:16:55.120
all of those industries have
been completely revolutionized

376
00:16:55.120 --> 00:16:58.100
with data and analytics,
and they're making actually

377
00:16:58.100 --> 00:17:01.190
data-driven decisions, and
they're doing test and learn,

378
00:17:01.190 --> 00:17:02.630
and they course-correct.

379
00:17:02.630 --> 00:17:04.880
And we're seeing a lot
of that in entertainment,

380
00:17:04.880 --> 00:17:08.950
but still a lot of the studios
are very much judgment driven.

381
00:17:08.950 --> 00:17:10.530
And I think this is
very interesting,

382
00:17:10.530 --> 00:17:13.069
because it's the
beginning of the beginning

383
00:17:13.069 --> 00:17:13.990
for this industry.

384
00:17:13.990 --> 00:17:16.922
So hats off to you
for doing that.

385
00:17:16.922 --> 00:17:19.130
KARTIK HOSANAGAR: Well,
yeah, thanks for saying that.

386
00:17:19.130 --> 00:17:22.490
I mean, I'm super excited
to apply data to a setting

387
00:17:22.490 --> 00:17:25.869
where people are the
most skeptical of

388
00:17:25.869 --> 00:17:27.970
"Should data be used here?"

389
00:17:27.970 --> 00:17:33.820
While we are finding
value in this,

390
00:17:33.820 --> 00:17:38.150
I will also say that Hollywood
has enough and more skeptics

391
00:17:38.150 --> 00:17:40.470
with regard to whether
data should play a role.

392
00:17:40.470 --> 00:17:42.660
And sometimes,
there is [even this]

393
00:17:42.660 --> 00:17:45.960
misconception of what data is.

394
00:17:45.960 --> 00:17:48.280
And sometimes, it
is, "Oh, there's

395
00:17:48.280 --> 00:17:50.490
an AI system making
all the decisions,

396
00:17:50.490 --> 00:17:52.920
and I no longer have
a role to play."

397
00:17:52.920 --> 00:17:55.220
I think there's a lot of
room for subjectivity,

398
00:17:55.220 --> 00:17:59.300
but "How can data
be a useful tool?"

399
00:17:59.300 --> 00:18:00.530
is how we're looking at it.

400
00:18:00.530 --> 00:18:02.950
And hopefully in
five, 10 years, we

401
00:18:02.950 --> 00:18:06.260
can establish a track
record of what data can do.

402
00:18:06.260 --> 00:18:08.570
And that might change
the mindset here

403
00:18:08.570 --> 00:18:10.530
in the whole industry.

404
00:18:10.530 --> 00:18:12.060
I did see a lot
of that reaction,

405
00:18:12.060 --> 00:18:14.740
saying, "Oh, data in
creative has no role."

406
00:18:14.740 --> 00:18:16.820
But I saw another
kind of reaction,

407
00:18:16.820 --> 00:18:18.680
and that was interesting.

408
00:18:18.680 --> 00:18:21.880
I had some people in
the industry saying,

409
00:18:21.880 --> 00:18:26.000
"This is inevitable,
and Netflix is already

410
00:18:26.000 --> 00:18:27.910
starting some of this stuff.

411
00:18:27.910 --> 00:18:31.430
We are forced by Netflix
to get into this.

412
00:18:31.430 --> 00:18:33.850
So we have to start
doing it, and we

413
00:18:33.850 --> 00:18:35.450
may as well partner
with people like

414
00:18:35.450 --> 00:18:38.310
you to understand this game."

415
00:18:38.310 --> 00:18:43.350
So in fact, one of them told
me -- and this is quite simple,

416
00:18:43.350 --> 00:18:44.320
intuitive.

417
00:18:44.320 --> 00:18:48.110
This is a producer who told
me that when his show came

418
00:18:48.110 --> 00:18:53.800
on Netflix, within days -- first
of all, within a day or two --

419
00:18:53.800 --> 00:18:56.180
Netflix was able to give
him feedback, of course,

420
00:18:56.180 --> 00:19:04.010
on how they thought his entire
first season would play out.

421
00:19:04.010 --> 00:19:07.180
By the end of Month
1, they were already

422
00:19:07.180 --> 00:19:09.940
in conversations about
Season 2, because Netflix

423
00:19:09.940 --> 00:19:12.580
was able to project out
what things would look like.

424
00:19:12.580 --> 00:19:15.670
Apparently, he was
also given information

425
00:19:15.670 --> 00:19:18.800
about which characters audiences
are most interested in.

426
00:19:18.800 --> 00:19:22.450
They wanted to kill one of the
characters early in Season 2,

427
00:19:22.450 --> 00:19:24.460
and they were told,
"No, you cannot do this.

428
00:19:24.460 --> 00:19:26.120
That's your main character."

429
00:19:26.120 --> 00:19:29.000
So this producer was
telling me, "Yes, I mean,

430
00:19:29.000 --> 00:19:32.030
Netflix is already pushing
us to do some of this stuff."

431
00:19:32.030 --> 00:19:34.740
The difference is,
Netflix's data is coming in

432
00:19:34.740 --> 00:19:37.550
after the show is
released, and Jumpcut

433
00:19:37.550 --> 00:19:41.090
is trying to bring in data
before the show is created.

434
00:19:41.090 --> 00:19:43.820
But I think some
of them see this

435
00:19:43.820 --> 00:19:46.950
as being inevitable because
of companies like Netflix

436
00:19:46.950 --> 00:19:49.467
and are happy to
embrace our approach.

437
00:19:49.467 --> 00:19:51.050
SAM RANSBOTHAM: I
was a bit surprised,

438
00:19:51.050 --> 00:19:53.980
because obviously your book
from a couple of years ago

439
00:19:53.980 --> 00:19:56.690
was about how algorithms,
I think in your words,

440
00:19:56.690 --> 00:19:58.950
are shaping our lives.

441
00:19:58.950 --> 00:20:02.230
But this seems to be one area
that algorithms are really

442
00:20:02.230 --> 00:20:03.540
not shaping our lives.

443
00:20:03.540 --> 00:20:07.180
So how are you getting
this constituency of people

444
00:20:07.180 --> 00:20:09.390
to pay attention to
data and algorithms that

445
00:20:09.390 --> 00:20:10.970
have historically not?

446
00:20:10.970 --> 00:20:13.175
That's got to be not
just selling a story;

447
00:20:13.175 --> 00:20:14.800
you're also having
to sell an approach.

448
00:20:14.800 --> 00:20:16.900
KARTIK HOSANAGAR:
So in the book,

449
00:20:16.900 --> 00:20:19.440
which is called A Human's
Guide to Machine Intelligence ,

450
00:20:19.440 --> 00:20:24.580
I documented several ways in
which AI is shaping our lives,

451
00:20:24.580 --> 00:20:28.190
as Sam was just mentioning,
and included some examples

452
00:20:28.190 --> 00:20:29.240
in the movie world.

453
00:20:29.240 --> 00:20:33.760
So, for example, I talk about
how on Netflix, algorithms

454
00:20:33.760 --> 00:20:36.090
are driving a lot
of our consumption.

455
00:20:36.090 --> 00:20:38.480
There was a paper by data
scientists at Netflix

456
00:20:38.480 --> 00:20:42.720
that said something like 80%
of viewing hours on Netflix

457
00:20:42.720 --> 00:20:44.890
originated from algorithmic
recommendations.

458
00:20:44.890 --> 00:20:48.310
So it's pretty clear it
is shaping our lives, even

459
00:20:48.310 --> 00:20:49.970
in a setting like the movies.

460
00:20:49.970 --> 00:20:53.290
But I think while it's
shaping our lives --

461
00:20:53.290 --> 00:20:56.840
certainly in terms of
how we consume content --

462
00:20:56.840 --> 00:20:59.480
how content is created,
that side of the business,

463
00:20:59.480 --> 00:21:00.660
hasn't really evolved.

464
00:21:00.660 --> 00:21:02.860
So the supply side
still looks the same,

465
00:21:02.860 --> 00:21:04.610
even though the
demand has completely

466
00:21:04.610 --> 00:21:07.867
been reshaped by the market.

467
00:21:07.867 --> 00:21:10.200
SHERVIN KHODABANDEH: To me,
there [are] a lot of analogs

468
00:21:10.200 --> 00:21:16.930
in what you're doing here to
the evolution of data-driven

469
00:21:16.930 --> 00:21:20.160
decision-making -- not
replacing [the] human,

470
00:21:20.160 --> 00:21:23.700
but basically creating a
smarter, more effective,

471
00:21:23.700 --> 00:21:25.130
or efficient human.

472
00:21:25.130 --> 00:21:27.750
And there is exactly
the same thing

473
00:21:27.750 --> 00:21:30.940
in retail and in personalization
and in marketing.

474
00:21:30.940 --> 00:21:35.150
Like I said, it feels to me
that entertainment and media

475
00:21:35.150 --> 00:21:37.600
has been sort of the
last bastion of it.

476
00:21:37.600 --> 00:21:40.430
I think, Kartik, as
you said, a lot of it

477
00:21:40.430 --> 00:21:42.810
is risk rather
than anything else.

478
00:21:42.810 --> 00:21:44.570
And I think in
another book of yours,

479
00:21:44.570 --> 00:21:46.990
you alluded to the
chicken-and-egg problem:

480
00:21:46.990 --> 00:21:50.070
If you don't have
data on the talent,

481
00:21:50.070 --> 00:21:53.430
how do you make those
decisions without some level

482
00:21:53.430 --> 00:21:56.130
of experimentation and
some level of data?

483
00:21:56.130 --> 00:21:58.690
And then, of course, there
are disruptors like Netflix

484
00:21:58.690 --> 00:22:03.450
who are forcing everybody
to become more data driven

485
00:22:03.450 --> 00:22:05.380
as a way of surviving.

486
00:22:05.380 --> 00:22:09.720
I wonder whether studios could
survive 10 years from now

487
00:22:09.720 --> 00:22:12.160
without making a
major step that way.

488
00:22:12.160 --> 00:22:13.160
KARTIK HOSANAGAR: Right.

489
00:22:13.160 --> 00:22:16.180
I would think that
the time has come.

490
00:22:16.180 --> 00:22:18.290
In fact, maybe the
time was yesterday

491
00:22:18.290 --> 00:22:20.760
for something like this,
which is of course why

492
00:22:20.760 --> 00:22:22.770
I went and started Jumpcut.

493
00:22:22.770 --> 00:22:28.910
I think it's inevitable, because
we know that human judgment has

494
00:22:28.910 --> 00:22:31.230
its big share of problems.

495
00:22:31.230 --> 00:22:35.470
Obviously, human judgment
is also great in many ways,

496
00:22:35.470 --> 00:22:39.600
but we have our biases;
we're colored by them

497
00:22:39.600 --> 00:22:42.410
in ways we don't realize.

498
00:22:42.410 --> 00:22:46.240
Having a tool that can free
us from those, I think --

499
00:22:46.240 --> 00:22:49.880
it's a no-brainer that we
should be embracing them.

500
00:22:49.880 --> 00:22:52.020
And yeah, the time
has come to do this

501
00:22:52.020 --> 00:22:55.340
in fields that were unexpected.

502
00:22:55.340 --> 00:22:58.990
I think sports is not, again,
a place where I would have

503
00:22:58.990 --> 00:23:01.640
guessed there would
be early adopters,

504
00:23:01.640 --> 00:23:03.920
but clearly there
have been, and they've

505
00:23:03.920 --> 00:23:06.620
shown that the Moneyball
kind of approach works.

506
00:23:06.620 --> 00:23:10.830
What you really need
is to integrate data

507
00:23:10.830 --> 00:23:13.270
into decision-making
along the way.

508
00:23:13.270 --> 00:23:17.980
You have a deep understanding of
data to know when to lean on it

509
00:23:17.980 --> 00:23:19.630
but also when to
question it, and you

510
00:23:19.630 --> 00:23:22.640
have a strong creative
point of view,

511
00:23:22.640 --> 00:23:24.258
and you bring the two together.

512
00:23:24.258 --> 00:23:26.550
So you've got to almost create
it [from the] ground up,

513
00:23:26.550 --> 00:23:29.800
which is why we said we're
not a data insights vendor.

514
00:23:29.800 --> 00:23:32.620
We're a company
that's creating really

515
00:23:32.620 --> 00:23:34.225
a new kind of
business that brings

516
00:23:34.225 --> 00:23:35.350
data and creative together.

517
00:23:35.350 --> 00:23:38.180
SHERVIN KHODABANDEH: You have
to really close the loop,

518
00:23:38.180 --> 00:23:41.360
because otherwise
you have the problem

519
00:23:41.360 --> 00:23:45.820
that I think a lot
of early adopters

520
00:23:45.820 --> 00:23:48.830
of data-driven
decision-making fell into,

521
00:23:48.830 --> 00:23:50.560
which is they said,
"Well, data is

522
00:23:50.560 --> 00:23:52.360
our future
competitive advantage.

523
00:23:52.360 --> 00:23:57.100
Let's acquire as much of it
as possible and put it in some

524
00:23:57.100 --> 00:24:00.970
Hadoop cluster somewhere, only
to know we can't do anything

525
00:24:00.970 --> 00:24:03.070
with it because we haven't
thought through --"

526
00:24:03.070 --> 00:24:04.730
SAM RANSBOTHAM: But they've
got it, they've got it.

527
00:24:04.730 --> 00:24:06.688
SHERVIN KHODABANDEH: They
have the data, right?

528
00:24:06.688 --> 00:24:09.720
And they're like the number of
companies that have done just

529
00:24:09.720 --> 00:24:13.450
that 10 years ago, 15 years
ago, even doing it now --

530
00:24:13.450 --> 00:24:15.240
they're like, "We'll
get all the data.

531
00:24:15.240 --> 00:24:18.580
Once we have it all, we'll
figure out what to do with it."

532
00:24:18.580 --> 00:24:21.340
But as you're saying,
sometimes it's

533
00:24:21.340 --> 00:24:24.450
about doing so much
more with the data

534
00:24:24.450 --> 00:24:26.870
you already have
by connecting it

535
00:24:26.870 --> 00:24:29.140
to the business process
or the creative process

536
00:24:29.140 --> 00:24:31.025
and connecting the endpoints.

537
00:24:31.025 --> 00:24:32.400
SAM RANSBOTHAM:
Kartik, we really

538
00:24:32.400 --> 00:24:35.570
enjoyed both your background
and how you're applying this

539
00:24:35.570 --> 00:24:38.680
into your perspective on
algorithms, and the difference

540
00:24:38.680 --> 00:24:41.330
that algorithms can make in
terms of exploring our search

541
00:24:41.330 --> 00:24:42.780
space I think is fascinating.

542
00:24:42.780 --> 00:24:44.370
But also those
parts two and three

543
00:24:44.370 --> 00:24:47.298
were pretty fascinating
too, about how to integrate

544
00:24:47.298 --> 00:24:48.090
that with creative.

545
00:24:48.090 --> 00:24:50.400
We appreciate you taking
the time to talk with us.

546
00:24:50.400 --> 00:24:51.650
KARTIK HOSANAGAR: My pleasure.

547
00:24:51.650 --> 00:24:52.680
Thank you for having me.

548
00:24:52.680 --> 00:24:54.120
SAM RANSBOTHAM: In
our next episode,

549
00:24:54.120 --> 00:24:55.760
we speak with Sarah
Karthigan about how

550
00:24:55.760 --> 00:24:59.220
she's helping ExxonMobil use
AI for self-healing process

551
00:24:59.220 --> 00:25:01.370
improvement across
business units.

552
00:25:01.370 --> 00:25:02.572
Please join us.

553
00:25:02.572 --> 00:25:04.030
ALLISON RYDER:
Thanks for listening

554
00:25:04.030 --> 00:25:05.550
to Me, Myself, and AI.

555
00:25:05.550 --> 00:25:07.980
We believe, like you,
that the conversation

556
00:25:07.980 --> 00:25:10.210
about AI implementation
doesn't start and stop

557
00:25:10.210 --> 00:25:11.360
with this podcast.

558
00:25:11.360 --> 00:25:13.870
That's why we've created a
group on LinkedIn, specifically

559
00:25:13.870 --> 00:25:14.990
for leaders like you.

560
00:25:14.990 --> 00:25:17.740
It's called AI for Leaders,
and if you join us,

561
00:25:17.740 --> 00:25:19.780
you can chat with show
creators and hosts,

562
00:25:19.780 --> 00:25:23.380
ask your own questions, share
insights, and gain access

563
00:25:23.380 --> 00:25:25.860
to valuable resources
about AI implementation

564
00:25:25.860 --> 00:25:27.970
from MIT SMR and BCG.

565
00:25:27.970 --> 00:25:34.227
You can access it by visiting
mitsmr.com/AIforLeaders.

566
00:25:34.227 --> 00:25:35.810
We'll put that link
in the show notes,

567
00:25:35.810 --> 00:25:38.240
and we hope to see you there.

568
00:25:38.240 --> 00:25:42.000