WEBVTT

1
00:00:00.131 --> 00:00:21.564
Hi, every week I do a little something on Instagram called a week in review where on Saturday mornings I review the previous week of updates in technology and AI and the creative industries and all these things that are kind of blending together and raise serious ethical questions and economic questions and job displacement questions and all those sorts of things.

2
00:00:22.185 --> 00:00:22.385
Um...

3
00:00:22.745 --> 00:00:25.326
And they've been doing really well and getting a lot of really good feedback.

4
00:00:25.366 --> 00:00:28.167
So I've decided to do them as a little audio episode.

5
00:00:28.627 --> 00:00:38.130
So you can either get them on a Saturday morning on my Substack, or you can tune in here on a Monday morning and I'll release it and give you a short summary of,

6
00:00:38.811 --> 00:00:41.973
what I think are the most important highlights from the previous week.

7
00:00:42.434 --> 00:00:45.877
There's so much going on right now, and it's really hard to keep track.

8
00:00:45.977 --> 00:00:56.227
And so what I'd like it to be is just a very short summary of what went on and to kind of prep you for the week if you don't feel that you have nearly enough time to see the forest for the trees.

9
00:00:56.427 --> 00:01:00.530
This is the week ending August 28th, 2026.

10
00:01:01.150 --> 00:01:13.998
So last week, the Ankler reported that for nearly three years, Kathleen Kennedy and AFI Dean Susan Ruskin have been quietly bringing Hollywood and Silicon Valley together to talk about how AI should actually be used in filmmaking.

11
00:01:14.439 --> 00:01:16.100
If you don't know who Kathleen Kennedy is...

12
00:01:16.440 --> 00:01:21.864
She was the co-founder and lead producer at Amblin Entertainment, Steven Spielberg's production company.

13
00:01:21.884 --> 00:01:23.585
I mean, she produced E.T.

14
00:01:23.625 --> 00:01:24.866
She produced Jurassic Park.

15
00:01:24.886 --> 00:01:27.027
She produced Schindler's List and The Sixth Sense.

16
00:01:27.988 --> 00:01:31.710
And then since 2012, she's been the head of Lucasfilm.

17
00:01:31.750 --> 00:01:35.372
So there couldn't be somebody who's more trusted within the Hollywood environment.

18
00:01:35.392 --> 00:01:39.615
But still, these talks have been going on for three years, and they've been having them behind closed doors.

19
00:01:39.995 --> 00:01:51.243
They've not been tweeting about it, not issuing statements about it, actually getting filmmakers, actors, studio executives, technologists, and other people who have skin in the game into rooms together and asking some of the most difficult questions.

20
00:01:52.762 --> 00:01:56.665
And what's interesting is some of the biggest leaders in Hollywood have been in this room.

21
00:01:57.065 --> 00:02:00.688
The only person who's put her name out there out front is Kathleen Kennedy.

22
00:02:01.028 --> 00:02:03.970
No one else is even admitting to be part of these talks.

23
00:02:04.750 --> 00:02:10.534
But where they landed on was somewhere that I strongly agree with, that we have to stop calling everything AI.

24
00:02:11.435 --> 00:02:13.397
Now, this sounds semantic, but it really isn't.

25
00:02:13.897 --> 00:02:27.608
According to the Ankler, these conversations began in November 2023 at artist Refik Anadol's studio in Los Angeles, just after the strikes, and have now produced a proposed framework that they're calling Human Generative Workflows.

26
00:02:27.808 --> 00:02:37.473
And they're designed to distinguish generative technology that operates under a granular human creative control from machine generated work that displaces human creative contribution.

27
00:02:38.034 --> 00:02:42.336
And I love this because the basic premise is something I've been banging on about for ages.

28
00:02:42.756 --> 00:02:44.718
We have to stop calling everything AI.

29
00:02:45.618 --> 00:02:48.021
The framework essentially splits the field into three.

30
00:02:48.161 --> 00:02:50.945
First, utility techniques and embedded AI.

31
00:02:51.305 --> 00:03:03.340
The machine learning already sitting inside creative tools doing things like noise reduction, cleanup, roto, and other tasks we've been happily using for years without organizing a protest outside Adobe every two weeks.

32
00:03:04.081 --> 00:03:20.347
Second, human-generative workflows, where artists control pipelines that put generative models inside VFX or production tooling that already exists, which the group argues remain human-authored, copyrightable, and the product of deliberate creative decisions by humans.

33
00:03:20.967 --> 00:03:26.829
And third, machine-generative work prompt goes in, finished thing comes out, with minimum human creative control.

34
00:03:27.709 --> 00:03:29.390
Three categories instead of one.

35
00:03:29.530 --> 00:03:30.831
And that's their whole idea.

36
00:03:30.991 --> 00:03:32.011
And thank God.

37
00:03:32.632 --> 00:03:46.259
Because earlier this month, because earlier this month, that private conversation expanded into an AFI summit involving 120 people from entertainment, technology, labor, studios, streamers, government, and academia.

38
00:03:46.460 --> 00:03:53.003
The working groups covered labor, compensation, production economics, copyright, permission, and standards and disclosure.

39
00:03:53.563 --> 00:03:55.785
So in other words, all the actual difficult stuff.

40
00:03:55.805 --> 00:03:56.365
Okay.

41
00:03:56.465 --> 00:04:01.667
And this is encouraging because the public conversation around AI in Hollywood has become almost comically binary.

42
00:04:02.107 --> 00:04:05.168
AI is the future or AI is the death of human creativity.

43
00:04:05.669 --> 00:04:10.691
And then everyone screams at each other until somebody posts a picture with six fingers and everybody calls them a fascist.

44
00:04:10.831 --> 00:04:13.652
So the binary conversation isn't getting us anywhere.

45
00:04:13.712 --> 00:04:19.834
And I applaud everyone for at least admitting that these conversations are happening because we're seeing them happen at every single studio.

46
00:04:20.174 --> 00:04:21.695
Just no one's willing to admit it right now.

47
00:04:21.775 --> 00:04:26.578
And meanwhile, the technology itself has become vastly more complicated than the argument we're having around it.

48
00:04:27.118 --> 00:04:30.220
Automated rotoscoping isn't generating an actor's performance.

49
00:04:30.480 --> 00:04:33.722
AI-assisted denoising isn't replacing a cinematographer.

50
00:04:33.842 --> 00:04:42.107
Machine learning-assisted depth extraction isn't the same thing as typing, make me a Nike commercial and make it look like it's directed by Denis Villeneuve.

51
00:04:43.503 --> 00:04:46.065
and putting that into a prompt box and going to lunch.

52
00:04:46.885 --> 00:04:55.331
A tool that helps an artist perform a task faster is not automatically ethically equivalent to a system designed to remove that artist from the process altogether.

53
00:04:56.132 --> 00:04:59.974
These things may share some underlying technologies that doesn't make them the same thing.

54
00:05:00.815 --> 00:05:02.736
There is, however, an important caveat.

55
00:05:03.056 --> 00:05:09.901
I don't want AI has been used in VFX for years to become the industry's new rhetorical escape hatch either, because that's bullshit too.

56
00:05:10.763 --> 00:05:15.348
So yes, machine learning has existed inside creative production tools for years.

57
00:05:15.528 --> 00:05:25.317
But no, that doesn't mean synthetic actors, generative video, automated roto, and an autocomplete feature in Photoshop suddenly all belong in the same ethical category.

58
00:05:25.518 --> 00:05:26.358
Quite the opposite.

59
00:05:26.819 --> 00:05:29.301
The fact that they're different is exactly the fucking point.

60
00:05:30.182 --> 00:05:40.547
If the same two letters describe everything from noise reduction to generating a human performance without the human, then those two letters, AI, become almost useless as a description.

61
00:05:41.188 --> 00:05:48.671
And if we can't describe the technology accurately, how the hell are we supposed to write sensible rules about labor, copyright, consent, and compensation?

62
00:05:48.812 --> 00:05:51.293
Which, on another note, brings us neatly around to music.

63
00:05:51.933 --> 00:05:54.557
It turns out the music industry isn't anti-AI.

64
00:05:54.657 --> 00:05:56.260
It's anti-not-getting-paid.

65
00:05:56.861 --> 00:05:58.523
This might be my favorite story of the week.

66
00:05:58.984 --> 00:06:07.356
Universal Music Group, Sony Music Group, and Warner Music Group all invested in Stability AI's $76 million Series B, electronic...

67
00:06:07.857 --> 00:06:09.138
Electronic Arts joined them.

68
00:06:09.679 --> 00:06:21.310
Stability says the round brings the funding raised under its current leadership to $232 million, and more importantly for this conversation, that its stable Audio 3.0 family was trained on fully licensed data.

69
00:06:21.510 --> 00:06:25.834
Then on Friday, Sony Music Publishing and Warner Chapel Music sued Anthropic.

70
00:06:26.134 --> 00:06:36.688
Well then, the lawsuit accuses Anthropic of using tens of thousands of copyrighted musical works without authorization, which they did, in building Claude and seeks potentially enormous statutory damages.

71
00:06:37.269 --> 00:06:41.735
It even names Anthropic CEO Dario Moday and co-founder Benjamin Mann individually.

72
00:06:42.316 --> 00:06:47.038
So Sony and Waters spent the week simultaneously investing in AI and suing AI.

73
00:06:47.398 --> 00:06:48.099
Contradiction?

74
00:06:48.119 --> 00:06:49.039
I don't think so.

75
00:06:49.580 --> 00:06:52.701
This is the world right now, and I see this as leadership.

76
00:06:52.881 --> 00:06:55.502
This is probably why I get called a hypocrite so many times.

77
00:06:56.223 --> 00:07:02.746
I think this is one of the cleanest demonstrations yet of what the creative industry's position on generative AI is actually becoming.

78
00:07:03.226 --> 00:07:06.988
You can build it, you just don't automatically get our shit for free.

79
00:07:07.588 --> 00:07:10.371
We've spent three years pretending there are two teams.

80
00:07:10.471 --> 00:07:17.037
Team AI apparently thinks copyright is an annoying 18th century inconvenience standing between humanity and AGI.

81
00:07:17.578 --> 00:07:20.841
Team Artist apparently wants to throw every GPU into the sea.

82
00:07:21.522 --> 00:07:24.565
Meanwhile, the grown-up position has been sitting there the entire time.

83
00:07:25.105 --> 00:07:26.507
Build incredible technology.

84
00:07:26.767 --> 00:07:27.828
License people's work.

85
00:07:28.068 --> 00:07:28.809
Get permission.

86
00:07:29.109 --> 00:07:29.610
Pay them.

87
00:07:30.732 --> 00:07:34.234
The most remarkable part about all of it is how simple and logical it is.

88
00:07:35.115 --> 00:07:37.516
And there's this other piece that I find particularly interesting.

89
00:07:37.636 --> 00:07:39.237
The labels are taking equity.

90
00:07:39.657 --> 00:07:41.998
They don't want just a licensing fee for the ingredients.

91
00:07:42.038 --> 00:07:43.899
They want ownership in the restaurant.

92
00:07:44.240 --> 00:07:52.104
If generative music becomes an enormous new industry, their catalogs, artists, expertise, and licensing arrangements help make that industry possible.

93
00:07:52.424 --> 00:07:54.325
They'd quite like to participate in the upside.

94
00:07:54.966 --> 00:07:56.467
As far as I'm concerned, fair enough.

95
00:07:56.487 --> 00:07:57.087
Right.

96
00:07:57.187 --> 00:08:00.830
For years, the AI debate has been framed as artists versus technology.

97
00:08:01.190 --> 00:08:05.934
I increasingly think the actual divide is becoming licensed AI versus unlicensed AI.

98
00:08:06.434 --> 00:08:10.257
And after that, permission, provenance, compensation, and participation.

99
00:08:12.639 --> 00:08:14.000
It's actually so simple.

100
00:08:14.580 --> 00:08:20.405
If you're going to build technology off artists' backs, you very least need to compensate the artists.

101
00:08:20.625 --> 00:08:23.287
But better than that, give them a piece of the pie.

102
00:08:24.538 --> 00:08:28.622
Okay, so on to the sexiest subject of this week, metadata.

103
00:08:28.762 --> 00:08:31.965
Brace yourselves or skip ahead, but we're about to get nerdy.

104
00:08:32.486 --> 00:08:40.333
So Luminate, which is the entertainment data company whose information powers billboard charts, and it also has loads of information on box office.

105
00:08:40.793 --> 00:08:47.600
Everybody uses it, CAA, UTA, anyone who wants information about entertainment data right now, they buy it from Luminate.

106
00:08:48.616 --> 00:08:54.380
So they announced a global framework this week for identifying and measuring AI-generated music.

107
00:08:54.820 --> 00:08:59.083
Its Connect platform can now flag AI-generated songs and artists worldwide.

108
00:08:59.623 --> 00:09:03.466
Luminate says it's already identified thousands of AI-generated items.

109
00:09:03.926 --> 00:09:08.790
Those classifications aren't visible to customers yet, but the company plans to roll them out later this year.

110
00:09:09.290 --> 00:09:17.696
It's also working with streaming platforms and AI music companies on self-declared attribution data and building a process for challenging incorrect classifications.

111
00:09:18.396 --> 00:09:20.038
I know, try to contain your excitement.

112
00:09:20.378 --> 00:09:24.443
But I actually think this is a huge deal, precisely because it sounds so boring.

113
00:09:25.244 --> 00:09:28.868
We've spent the last few years having philosophical arguments about transparency.

114
00:09:29.129 --> 00:09:30.711
AI-generated work should be labeled.

115
00:09:30.911 --> 00:09:31.932
Artists should consent.

116
00:09:32.212 --> 00:09:33.474
Training data should be licensed.

117
00:09:33.754 --> 00:09:35.897
Human-created music should be treated differently.

118
00:09:36.097 --> 00:09:36.638
Great!

119
00:09:37.411 --> 00:09:37.872
How?

120
00:09:38.072 --> 00:09:42.556
Because eventually somebody has to turn the manifesto into a database and into a business.

121
00:09:42.896 --> 00:09:53.166
You need identification, metadata, classification, a mechanism for challenging that classification, standards for eligibility, measurement, rights of information and compensation.

122
00:09:53.527 --> 00:09:58.571
Then somebody has to maintain the whole fucking thing when there's 100 million songs in it and everyone disagrees.

123
00:09:59.252 --> 00:10:04.594
The AI rights debate has reached the spreadsheet stage, and that's usually when something becomes an industry.

124
00:10:05.274 --> 00:10:06.595
And this doesn't stop at music.

125
00:10:06.795 --> 00:10:08.495
Advertising is going to need versions of this.

126
00:10:08.836 --> 00:10:10.256
Film is going to need versions of this.

127
00:10:10.576 --> 00:10:13.477
Publishing, photography, and design are going to need versions of this.

128
00:10:14.138 --> 00:10:18.479
Eventually, eventually, saying something is AI-generated won't be enough.

129
00:10:18.839 --> 00:10:20.040
Was the whole thing generated?

130
00:10:20.200 --> 00:10:20.740
One element?

131
00:10:20.800 --> 00:10:23.101
Just the last element, which overlays everything?

132
00:10:24.280 --> 00:10:25.200
What tools were used?

133
00:10:25.280 --> 00:10:26.681
What source assets went into it?

134
00:10:26.761 --> 00:10:28.162
Were those assets licensed?

135
00:10:28.442 --> 00:10:29.962
Was somebody's likeness involved?

136
00:10:30.262 --> 00:10:31.083
Did they consent?

137
00:10:31.383 --> 00:10:32.303
Who owns the result?

138
00:10:32.743 --> 00:10:33.724
Is there a human author?

139
00:10:34.024 --> 00:10:35.064
Who gets paid?

140
00:10:35.765 --> 00:10:38.346
You can't solve any of that with another ethics statement.

141
00:10:38.406 --> 00:10:39.606
You need infrastructure.

142
00:10:39.766 --> 00:10:43.728
Boring, tedious, absolutely essential infrastructure.

143
00:10:44.068 --> 00:10:45.188
So welcome to the future.

144
00:10:45.668 --> 00:10:46.889
It has metadata standards.

145
00:10:48.952 --> 00:10:57.878
So finally, NVIDIA, the absolutely massive company that is deciding that it wants to do something with all of its cash on hand, has decided to keep buying things.

146
00:10:58.618 --> 00:11:02.781
And NVIDIA reportedly wants to buy Hugging Face for somewhere around $13 billion.

147
00:11:02.961 --> 00:11:08.685
If you don't know what Hugging Face is, it's a collaborative platform, an online hub for artificial intelligence and machine learning.

148
00:11:09.045 --> 00:11:11.487
It's often described as the GitHub for AI.

149
00:11:14.086 --> 00:11:16.968
there's conflicting reporting about exactly where the deal stands.

150
00:11:17.368 --> 00:11:21.450
Some reports say NVIDIA has agreed to acquire the company for roughly $12.5 billion.

151
00:11:21.971 --> 00:11:25.553
Others describe negotiations that may not have yet produced a final agreement.

152
00:11:25.973 --> 00:11:29.515
Neither company has publicly confirmed the acquisition as of the time of writing this.

153
00:11:29.996 --> 00:11:30.816
So, a grain of salt.

154
00:11:31.116 --> 00:11:34.218
But, holy shit, the strategic idea here is interesting.

155
00:11:34.358 --> 00:11:36.820
NVIDIA started this AI boom as the chip company.

156
00:11:37.180 --> 00:11:39.264
Then it became the AI infrastructure company.

157
00:11:39.724 --> 00:11:41.968
Then it started investing all over the ecosystem.

158
00:11:42.169 --> 00:11:47.978
And now it may want one of the most important places developers go to discover, download, and share work with open models.

159
00:11:48.881 --> 00:11:51.123
Okay, Jensen, leave some AI for everybody else.

160
00:11:51.663 --> 00:11:56.367
Because hugging face is enormously important precisely because it doesn't need to win the model race.

161
00:11:56.787 --> 00:11:57.728
It hosts the race.

162
00:11:58.368 --> 00:12:01.250
Models from thousands of developers and organizations live there.

163
00:12:01.610 --> 00:12:02.571
Datasets live there.

164
00:12:02.691 --> 00:12:03.472
Tools live there.

165
00:12:03.512 --> 00:12:08.636
Developers use it as one of the central gathering places for open AI.

166
00:12:08.736 --> 00:12:10.517
And when I say open AI, I don't mean the company.

167
00:12:10.577 --> 00:12:11.738
I mean literally open AI.

168
00:12:12.318 --> 00:12:13.379
artificial intelligence.

169
00:12:13.919 --> 00:12:16.201
And all those models eventually need somewhere to run.

170
00:12:16.381 --> 00:12:18.802
And guess who would be delighted to sell you the compute?

171
00:12:19.483 --> 00:12:21.384
There's a beautiful strategic logic to this.

172
00:12:21.544 --> 00:12:22.925
Google can build its own chips.

173
00:12:23.085 --> 00:12:24.386
Amazon can build its own chips.

174
00:12:24.626 --> 00:12:28.128
The giant AI labs can desperately hunt for alternatives to NVIDIA.

175
00:12:28.428 --> 00:12:32.250
But NVIDIA doesn't necessarily need to predict which individual model wins.

176
00:12:32.591 --> 00:12:34.352
If the entire ecosystem gets bitter,

177
00:12:34.792 --> 00:12:35.572
NVIDIA wins.

178
00:12:35.772 --> 00:12:40.593
Owning Hugging Face would potentially put it much closer to the developer and distribution layer of that ecosystem.

179
00:12:41.013 --> 00:12:43.314
But it also raises a slightly uncomfortable question.

180
00:12:43.774 --> 00:12:47.054
Hugging Face has always felt a little like Switzerland for open AI.

181
00:12:47.314 --> 00:12:54.316
Part of its value comes from being neutral territory where NVIDIA, its competitors, researchers, startups, and developers can all show up.

182
00:12:54.856 --> 00:12:58.757
So what happens when the world's dominant supplier of AI compute owns the playground?

183
00:12:58.837 --> 00:13:00.237
It no longer becomes Switzerland.

184
00:13:01.218 --> 00:13:02.739
So maybe that's the story this week.

185
00:13:03.239 --> 00:13:05.600
Four stories that initially look unrelated.

186
00:13:05.960 --> 00:13:09.561
Hollywood quietly trying to figure out what a different kinds of AI actually mean.

187
00:13:09.961 --> 00:13:13.043
Music companies investing in one AI company while suing another.

188
00:13:13.423 --> 00:13:16.724
A data company building classifications for AI generated songs.

189
00:13:17.124 --> 00:13:21.786
And Nvidia potentially spending 13 billion on one of the centers of the open model universe.

190
00:13:22.526 --> 00:13:24.087
I think they're all telling us the same thing.

191
00:13:25.254 --> 00:13:28.217
AI is becoming an almost useless nomer.

192
00:13:28.818 --> 00:13:31.701
And thank God, because maybe we're finally moving down a level.

193
00:13:32.161 --> 00:13:33.222
Was AI involved?

194
00:13:33.402 --> 00:13:34.203
Who cares?

195
00:13:34.824 --> 00:13:37.346
The better questions are, what kind of AI?

196
00:13:37.647 --> 00:13:38.327
What's it doing?

197
00:13:38.588 --> 00:13:39.509
Who's it controlled by?

198
00:13:39.629 --> 00:13:40.530
What was it trained on?

199
00:13:40.790 --> 00:13:41.591
Licensed by whom?

200
00:13:41.911 --> 00:13:42.632
Augmenting what?

201
00:13:42.892 --> 00:13:43.753
Replacing who?

202
00:13:44.133 --> 00:13:45.074
Who owns the output?

203
00:13:45.395 --> 00:13:46.156
Who gets paid?

204
00:13:46.716 --> 00:13:49.741
Those are much harder questions than AI, good or bad.

205
00:13:50.181 --> 00:13:53.286
They also make much worse LinkedIn thought leadership posts.

206
00:13:53.626 --> 00:14:01.418
There isn't a little green box for it's complicated and it depends on the specific application, rights framework, economic incentives, and degree of human control.

207
00:14:02.808 --> 00:14:03.308
It's a pity.

208
00:14:03.368 --> 00:14:10.770
It's one of the reasons that I'm pulling back from doing things on LinkedIn is because nuanced discussion doesn't get to happen there.

209
00:14:10.790 --> 00:14:13.690
It's a simulacra of nuanced conversation.

210
00:14:14.050 --> 00:14:15.231
It's not where it should happen.

211
00:14:15.291 --> 00:14:18.331
And that's why, to be honest, we've started doing these things.

212
00:14:18.511 --> 00:14:24.193
I've started posting on Substack somewhere where these conversations can actually have a middle ground.

213
00:14:24.213 --> 00:14:30.414
Because increasingly, I think that that's the answer is the entire world we're looking at.

214
00:14:30.774 --> 00:14:34.135
is polarized on either side of every single debate.

215
00:14:35.115 --> 00:14:38.976
And what we need is places where it's safe to have the conversation in the middle.

216
00:14:39.456 --> 00:14:40.497
And that's what we try to do here.

217
00:14:41.237 --> 00:14:43.117
Because this technology isn't going away.

218
00:14:43.597 --> 00:14:48.599
Neither are copyright, labor, artists, economics, consent, or human creativity.

219
00:14:49.339 --> 00:14:53.000
So the interesting work isn't deciding whether we're team AI or team human.

220
00:14:53.080 --> 00:14:57.221
It's figuring out how all of these things coexist without completely fucking each other over.

221
00:14:58.148 --> 00:15:01.111
And perhaps that's what I find genuinely encouraging about this week.

222
00:15:01.512 --> 00:15:03.674
The AI debate isn't becoming less contentious.

223
00:15:04.314 --> 00:15:07.337
Maybe we're finally getting sophisticated enough to have the right argument.

224
00:15:07.358 --> 00:15:08.739
So anyway, thank you.

225
00:15:08.819 --> 00:15:10.000
We'll be doing this every week.

226
00:15:10.461 --> 00:15:15.105
And tomorrow we have our podcast episode that will launch as usual.

227
00:15:15.906 --> 00:15:19.210
And this will just be another part of your week, we hope.

228
00:15:19.650 --> 00:15:20.791
So thank you for listening.

229
00:15:21.071 --> 00:15:26.033
And do let us know in the comments or in DMs if you have any things you'd like us to cover.

230
00:15:26.733 --> 00:15:27.614
We're just learning here.

231
00:15:28.634 --> 00:15:30.435
I started doing these posts on Instagram.

232
00:15:31.075 --> 00:15:32.836
They kind of took off and people started liking them.

233
00:15:33.236 --> 00:15:34.777
So they grew here.

234
00:15:35.357 --> 00:15:40.599
But I'd really like to know what parts of this are useful to you, whether there are any particular stories you'd like us to cover.

235
00:15:41.840 --> 00:15:42.840
And yeah, please reach out.

236
00:15:43.060 --> 00:15:43.601
Thanks so much.

237
00:15:43.841 --> 00:15:44.081
Bye-bye.
