Integrating AI Into Creative Workflows

1

00:00:00,000 --> 00:00:03,760

It's funny as human beings we go

2

00:00:02,080 --> 00:00:06,000

through this seemingly every time

3

00:00:03,760 --> 00:00:08,080

there's a technology disruption.

4

00:00:06,000 --> 00:00:10,000

What happens when creative teams embrace

5

00:00:08,080 --> 00:00:11,679

AI as a critical technology?

6

00:00:10,000 --> 00:00:13,200

We're past that, you know, call it the

7

00:00:11,679 --> 00:00:14,719

hype cycle, whatever you want to call

8

00:00:13,200 --> 00:00:15,839

it. We're into sort of practical

9

00:00:14,719 --> 00:00:17,520

applications

10

00:00:15,839 --> 00:00:19,119

of how this is going to manifest itself

11

00:00:17,520 --> 00:00:20,560

in every industry and media and

12

00:00:19,119 --> 00:00:21,039

entertainment is no different.

13

00:00:20,560 --> 00:00:23,279

Friend,

14

00:00:21,039 --> 00:00:25,199

in this episode of AI Cloud Essentials,

15

00:00:23,279 --> 00:00:28,160

we talk with Matt, head of media and

16

00:00:25,199 --> 00:00:30,880

entertainment at CoreWeave about how studios

17

00:00:28,160 --> 00:00:32,559

are moving from isolated AI experiments

18

00:00:30,880 --> 00:00:33,280

to integrating it in their production

19

00:00:32,559 --> 00:00:34,559

workflows.

20

00:00:33,280 --> 00:00:36,559

And I think what we're realizing is

21

00:00:34,559 --> 00:00:38,640

there is no one perfect solution. It's

22

00:00:36,559 --> 00:00:39,600

all going to be what I consider assist

23

00:00:38,640 --> 00:00:43,719

of AI.

24

00:00:39,600 --> 00:00:43,719

Don't miss this lively conversation.

25

00:00:50,800 --> 00:00:55,680

Hello everyone. Welcome to season 2 of

26

00:00:53,360 --> 00:00:58,160

the AI cloud essentials, a podcast

27

00:00:55,680 --> 00:01:00,559

series brought to you by CoreWeave. We're

28

00:00:58,160 --> 00:01:02,559

kicking off the season with a very

29

00:01:00,559 --> 00:01:04,720

interesting set of focus. It's the value

30

00:01:02,559 --> 00:01:06,799

realization phase and this is the first

31

00:01:04,720 --> 00:01:07,360

episode and I'm super excited to have

32

00:01:06,799 --> 00:01:09,040

Mac.

33

00:01:07,360 --> 00:01:09,280

I'm excited to be the first. Let's do

34

00:01:09,040 --> 00:01:12,159

this.

35

00:01:09,280 --> 00:01:15,600

Yes. Awesome. And this episode is going

36

00:01:12,159 --> 00:01:18,000

to focus on the AI creative revolution.

37

00:01:15,600 --> 00:01:20,000

So Mac, I'm going to kick off things by

38

00:01:18,000 --> 00:01:22,640

saying that, you know, we have watched

39

00:01:20,000 --> 00:01:25,439

the journey of every human being playing

40

00:01:22,640 --> 00:01:27,360

on with AI in the last couple of years,

41

00:01:25,439 --> 00:01:30,000

but the media and entertainment industry

42

00:01:27,360 --> 00:01:32,400

has been specifically kind of, you know,

43

00:01:30,000 --> 00:01:35,439

a little bit explosive if I have to say

44

00:01:32,400 --> 00:01:38,079

there's a human paranoia and there's

45

00:01:35,439 --> 00:01:40,240

a lot of value, but at the same time

46

00:01:38,079 --> 00:01:42,159

there's a mixed feeling about it. Could

47

00:01:40,240 --> 00:01:43,920

you shed some light on that in with your

48

00:01:42,159 --> 00:01:46,240

own personal point of view and your own

49

00:01:43,920 --> 00:01:48,079

journey? Yeah, I'm I'm I'm old enough

50

00:01:46,240 --> 00:01:50,640

now in my life to have a unique

51

00:01:48,079 --> 00:01:52,560

perspective. it's funny as human

52

00:01:50,640 --> 00:01:54,640

beings, we go through this seemingly

53

00:01:52,560 --> 00:01:55,439

every time there's a technology

54

00:01:54,640 --> 00:01:58,159

disruption.

55

00:01:55,439 --> 00:02:01,280

Okay. I'm a mechanical engineer by

56

00:01:58,159 --> 00:02:04,159

trade way back 30 years ago when I

57

00:02:01,280 --> 00:02:05,680

was in college and my first internship

58

00:02:04,159 --> 00:02:08,959

was

59

00:02:05,680 --> 00:02:13,040

at this sort of architecture and

60

00:02:08,959 --> 00:02:16,239

engineering firm and I come in from the

61

00:02:13,040 --> 00:02:18,879

university with CAD experience, AutoCAD

62

00:02:16,239 --> 00:02:21,280

Progineer and you walk in and there are

63

00:02:18,879 --> 00:02:24,640

these engineers with drafting boards and

64

00:02:21,280 --> 00:02:26,640

they're manually drawing all the things

65

00:02:24,640 --> 00:02:29,200

and they are freaking out about what

66

00:02:26,640 --> 00:02:31,599

this means and how they how they sort of

67

00:02:29,200 --> 00:02:34,480

deal with this disruption and we've

68

00:02:31,599 --> 00:02:36,800

been through this repeatedly and it's

69

00:02:34,480 --> 00:02:38,319

always this oh well I've built my brand

70

00:02:36,800 --> 00:02:41,040

and I think that's a lot of what we talk

71

00:02:38,319 --> 00:02:42,800

about media entertainment engineering

72

00:02:41,040 --> 00:02:44,400

any space when we talk about that

73

00:02:42,800 --> 00:02:46,800

technology disruption it's what have I

74

00:02:44,400 --> 00:02:50,480

built my brand around

75

00:02:46,800 --> 00:02:54,959

and I think what gets dismissed a lot

76

00:02:50,480 --> 00:02:57,840

is this concept of your personal

77

00:02:54,959 --> 00:02:59,360

IP or creative expertise in our field,

78

00:02:57,840 --> 00:03:03,040

right? And I think that's really where

79

00:02:59,360 --> 00:03:05,519

we are is I think 2024 into 2025, we

80

00:03:03,040 --> 00:03:07,519

were squarely in the oh, we're going to

81

00:03:05,519 --> 00:03:08,720

prompt something up and it's going to

82

00:03:07,519 --> 00:03:09,120

generate a movie.

83

00:03:08,720 --> 00:03:11,280

Yeah.

84

00:03:09,120 --> 00:03:13,200

Right. And I think now we're we're past

85

00:03:11,280 --> 00:03:14,720

that, you know, call it the hype cycle,

86

00:03:13,200 --> 00:03:16,640

whatever you want to call it. We're into

87

00:03:14,720 --> 00:03:18,319

sort of practical applications

88

00:03:18,319 --> 00:03:21,920

in every industry. And media and

89

00:03:19,840 --> 00:03:23,360

entertainment is no different. I

90

00:03:21,920 --> 00:03:27,200

think you have a little bit of a unique

91

00:03:23,360 --> 00:03:29,440

scenario with me and around unions in

92

00:03:27,200 --> 00:03:30,879

particular. And so you talk about

93

00:03:29,440 --> 00:03:33,360

well you look at sports and you talk

94

00:03:30,879 --> 00:03:34,879

about name image and likeness. You

95

00:03:33,360 --> 00:03:38,159

look at the media world and you talk

96

00:03:34,879 --> 00:03:41,200

about how they deal with sort of

97

00:03:38,159 --> 00:03:44,560

their personal IP their name,

98

00:03:41,200 --> 00:03:45,920

image and likeness and how do you I

99

00:03:44,560 --> 00:03:47,680

think it really comes down to money,

100

00:03:45,920 --> 00:03:51,040

right? How do you monetize that? How

101

00:03:47,680 --> 00:03:53,040

do you distribute that revenue that you

102

00:03:51,040 --> 00:03:55,519

are you are essentially assigning

103

00:03:53,040 --> 00:03:57,840

through that process? But if you if

104

00:03:55,519 --> 00:03:59,920

you really look at where we've gone

105

00:03:57,840 --> 00:04:03,040

probably from the end of 2025 to the

106

00:03:59,920 --> 00:04:05,280

beginning of 2026 is really around I

107

00:04:03,040 --> 00:04:08,319

think a lot of the uncertainty and doubt

108

00:04:05,280 --> 00:04:10,480

right the FUD around what it is and how

109

00:04:08,319 --> 00:04:13,200

it's going to

110

00:04:10,480 --> 00:04:14,319

replace or not replace things. I think I

111

00:04:13,200 --> 00:04:16,160

think we're starting to wrap our

112

00:04:14,319 --> 00:04:19,199

collective heads around how it's going

113

00:04:16,160 --> 00:04:21,519

to go and we've started to execute on

114

00:04:19,199 --> 00:04:22,320

some of the incremental improvements

115

00:04:21,519 --> 00:04:24,080

that it'll

116

00:04:22,320 --> 00:04:25,840

Yeah. Yeah. I think the

117

00:04:24,080 --> 00:04:28,080

realization that it's going to augment

118

00:04:25,840 --> 00:04:30,320

and empower the individual is really

119

00:04:28,080 --> 00:04:32,400

kind of settling in a little bit better

120

00:04:30,320 --> 00:04:34,720

and I was listening to you talking about

121

00:04:32,400 --> 00:04:36,639

you know your journey and being a

122

00:04:34,720 --> 00:04:38,240

mechanical engineer. I started my career

123

00:04:36,639 --> 00:04:41,280

as a electronics and communication

124

00:04:38,240 --> 00:04:43,440

engineer and my very first job was on a

125

00:04:41,280 --> 00:04:46,320

you know an autonomous car plant

126

00:04:43,440 --> 00:04:48,880

manufacturing plant and my first

127

00:04:46,320 --> 00:04:50,800

foray into my you know my professional

128

00:04:48,880 --> 00:04:53,360

experience was working with the robots

129

00:04:50,800 --> 00:04:54,960

and that time we did not have the fear

130

00:04:53,360 --> 00:04:56,639

we actually felt that it was really

131

00:04:54,960 --> 00:04:58,560

doing all the painting and the welding

132

00:04:56,639 --> 00:05:00,800

automatically so we always thought of

133

00:04:58,560 --> 00:05:02,800

them as a partner I think that feeling

134

00:05:00,800 --> 00:05:05,280

has to come here as well so now coming

135

00:05:02,800 --> 00:05:08,479

to this section where I would love to

136

00:05:05,280 --> 00:05:10,960

get your in perspective or examples, you

137

00:05:08,479 --> 00:05:12,400

know, case studies and customer wins or

138

00:05:10,960 --> 00:05:14,479

how you're working with the customers

139

00:05:12,400 --> 00:05:17,360

and if you could share some insights as

140

00:05:14,479 --> 00:05:20,080

to how people are taking more generic AI

141

00:05:17,360 --> 00:05:22,479

systems and building them into much more

142

00:05:20,080 --> 00:05:24,240

tuned for this industry. And as you

143

00:05:22,479 --> 00:05:25,759

said, it's not just writing a prompt and

144

00:05:24,240 --> 00:05:28,880

creating a movie.

145

00:05:25,759 --> 00:05:30,639

Yep. No, I think you're spot on. I

146

00:05:28,880 --> 00:05:33,360

think there was a lot of experimentation

147

00:05:30,639 --> 00:05:35,360

around that that latter step.

148

00:05:33,360 --> 00:05:36,639

you know, early on I think AI was

149

00:05:35,360 --> 00:05:38,560

essentially a black box. And I talk

150

00:05:36,639 --> 00:05:40,720

about that a lot. I talk about this

151

00:05:38,560 --> 00:05:42,880

this analogy between AI being this black

152

00:05:40,720 --> 00:05:46,639

box of you put things in, things come

153

00:05:42,880 --> 00:05:49,120

out. You don't know how to refine. Maybe

154

00:05:46,639 --> 00:05:51,280

it's adding things to your prompt or

155

00:05:49,120 --> 00:05:53,520

adding this that and the other element.

156

00:05:51,280 --> 00:05:55,440

And it's really hard to sort of

157

00:05:53,520 --> 00:05:56,479

direct that, so to speak,

158

00:05:55,440 --> 00:05:57,280

right? And I use that word

159

00:05:56,479 --> 00:05:59,840

intentionally.

160

00:05:57,280 --> 00:06:02,320

Yeah. now what you're looking at is

161

00:05:59,840 --> 00:06:04,080

is that sort of directed approach of and

162

00:06:02,320 --> 00:06:07,120

I think I really like your

163

00:06:04,080 --> 00:06:09,039

robotics example because it is not

164

00:06:07,120 --> 00:06:11,520

dissimilar to a lot of the mundane

165

00:06:09,039 --> 00:06:12,880

activities that go along within the

166

00:06:11,520 --> 00:06:15,520

creative community.

167

00:06:12,880 --> 00:06:17,600

rotoscoping is a great example. So

168

00:06:15,520 --> 00:06:21,280

there are literally people that have

169

00:06:17,600 --> 00:06:22,800

historically taken a frame with a hero

170

00:06:21,280 --> 00:06:26,240

element, whether it's an actor or

171

00:06:22,800 --> 00:06:29,360

something else filmed Yeah. on a set and

172

00:06:26,240 --> 00:06:31,440

outlined that character so that

173

00:06:29,360 --> 00:06:33,120

character can be removed and the

174

00:06:31,440 --> 00:06:34,560

background then be replaced. That's why

175

00:06:33,120 --> 00:06:37,759

you have things like green screens and

176

00:06:34,560 --> 00:06:40,240

blue screens and so that mundane thing.

177

00:06:37,759 --> 00:06:42,800

Yeah. Okay. Maybe there's an AI to

178

00:06:40,240 --> 00:06:45,039

replace that. But is that your

179

00:06:42,800 --> 00:06:47,039

your value, right? And so we really

180

00:06:45,039 --> 00:06:49,280

start kind of extracting away what

181

00:06:47,039 --> 00:06:49,759

the perceived value is.

182

00:06:49,759 --> 00:06:54,160

from that artist. And so something

183

00:06:52,479 --> 00:06:57,199

really interesting happened very

184

00:06:54,160 --> 00:06:59,759

recently with this, you know, AI

185

00:06:57,199 --> 00:07:02,080

complete replacement bit. OpenAI had

186

00:06:59,759 --> 00:07:04,000

had some technology called Sora. It was

187

00:07:02,080 --> 00:07:06,160

a video generation tool and everybody

188

00:07:04,000 --> 00:07:08,560

thought, "Oh my gosh, look at this

189

00:07:06,160 --> 00:07:11,360

high-res photoreal

190

00:07:08,560 --> 00:07:13,440

output coming out of this model." And

191

00:07:11,360 --> 00:07:16,000

then you see Disney doubling down on it

192

00:07:13,440 --> 00:07:18,400

with a billion dollar investment.

193

00:07:16,000 --> 00:07:19,280

And just recently, they scrapped the

194

00:07:18,400 --> 00:07:21,599

whole project.

195

00:07:19,280 --> 00:07:24,400

Yes. the agreement is gone. And I think

196

00:07:21,599 --> 00:07:26,639

what that signals is really

197

00:07:24,400 --> 00:07:29,280

this understanding or this recognition

198

00:07:26,639 --> 00:07:31,360

that we're never going to be in a prompt

199

00:07:29,280 --> 00:07:32,800

to movie scenario.

200

00:07:31,360 --> 00:07:35,199

There are certainly some applications

201

00:07:32,800 --> 00:07:38,160

around that. you know, ad agencies where

202

00:07:35,199 --> 00:07:39,759

they need to do some quick shots around,

203

00:07:38,160 --> 00:07:41,440

you know, one element or another and

204

00:07:39,759 --> 00:07:44,880

they can prompt up, hey, I want to go

205

00:07:41,440 --> 00:07:47,919

from here to here, but even that

206

00:07:44,880 --> 00:07:50,400

really doesn't fit the bill in terms of

207

00:07:47,919 --> 00:07:52,479

turning that knob ever so slightly to

208

00:07:50,400 --> 00:07:54,319

get to this sort of resonating theme in

209

00:07:52,479 --> 00:07:56,560

the industry of getting to that last

210

00:07:54,319 --> 00:07:58,319

perfect pixel. And so I think with any

211

00:07:56,560 --> 00:08:00,879

of the technology disruptions, we're

212

00:07:58,319 --> 00:08:05,039

always at this point of how can we do it

213

00:08:00,879 --> 00:08:08,000

faster with more iterations to get to

214

00:08:05,039 --> 00:08:09,840

that last perfect view and do it as

215

00:08:08,000 --> 00:08:11,520

efficiently as possible. I was reading

216

00:08:09,840 --> 00:08:13,840

one of your very interesting article

217

00:08:11,520 --> 00:08:16,000

where it talked about that you know

218

00:08:13,840 --> 00:08:18,639

this is not about just empowering an

219

00:08:16,000 --> 00:08:21,120

individual you know creator but also

220

00:08:18,639 --> 00:08:23,280

a team how they can better collaborate

221

00:08:21,120 --> 00:08:25,520

how they can actually kind of you know

222

00:08:23,280 --> 00:08:27,520

learn from each other's experiences and

223

00:08:25,520 --> 00:08:29,840

kind of accelerate the innovation that

224

00:08:27,520 --> 00:08:31,520

they are after right so that's very

225

00:08:29,840 --> 00:08:32,880

interesting could you talk a little bit

226

00:08:31,520 --> 00:08:35,120

about you know what are the new

227

00:08:32,880 --> 00:08:37,360

additional models and fine-tuning that

228

00:08:35,120 --> 00:08:39,120

code beef has kind of done to kind of

229

00:08:37,360 --> 00:08:41,839

help your customers in this journey I

230

00:08:39,120 --> 00:08:44,480

think you know where we are currently

231

00:08:41,839 --> 00:08:46,080

as of 2026 just to kind of date stamp

232

00:08:44,480 --> 00:08:47,920

where we are in terms of the thought

233

00:08:46,080 --> 00:08:49,279

leadership is really that there I

234

00:08:47,920 --> 00:08:51,120

think there's a bit of a

235

00:08:49,279 --> 00:08:54,240

and it's again pervasive in every

236

00:08:51,120 --> 00:08:58,000

industry it's utter chaos in terms of

237

00:08:54,240 --> 00:08:59,200

how this is going to establish itself in

238

00:08:58,000 --> 00:09:02,560

a work product

239

00:08:59,200 --> 00:09:04,320

and so I think you see 5,000 50,000

240

00:09:02,560 --> 00:09:06,399

different options and alternatives and

241

00:09:04,320 --> 00:09:07,360

startups and VC money going into these

242

00:09:06,399 --> 00:09:09,519

startups

243

00:09:07,360 --> 00:09:11,760

to hopefully figure out who has that

244

00:09:09,519 --> 00:09:13,120

that perfect solution.

245

00:09:18,399 --> 00:09:20,399

And so the rotoscoper

246

00:09:20,399 --> 00:09:24,959

Okay. Do we have something for that

247

00:09:21,920 --> 00:09:27,519

where that rotoscoper can then use their

248

00:09:24,959 --> 00:09:29,200

sort of human in the loop creativity and

249

00:09:27,519 --> 00:09:29,920

creative eye to do the things they need

250

00:09:29,200 --> 00:09:32,080

to do? Yeah.

251

00:09:29,920 --> 00:09:34,800

Right. great example there's a

252

00:09:32,080 --> 00:09:38,480

there's a wonderful there's a

253

00:09:34,800 --> 00:09:40,560

wonderful sort of AI model provider in

254

00:09:38,480 --> 00:09:44,800

this space called flawless

255

00:09:40,560 --> 00:09:46,720

and flawless allows studios to

256

00:09:44,800 --> 00:09:48,560

when they are doing language translation

257

00:09:46,720 --> 00:09:50,320

on a movie imagine a movie is shot in

258

00:09:48,560 --> 00:09:53,360

English but they want to distribute it

259

00:09:50,320 --> 00:09:55,279

in Japan

260

00:09:53,360 --> 00:09:58,240

they can go through and instead of just

261

00:09:55,279 --> 00:10:02,560

doing the audio voice over they can

262

00:09:58,240 --> 00:10:05,839

actually change the movement of the

263

00:10:02,560 --> 00:10:07,760

face of the actors to get that more

264

00:10:05,839 --> 00:10:10,560

immersive experience. So now something

265

00:10:07,760 --> 00:10:12,160

shot in English now looks like the actor

266

00:10:10,560 --> 00:10:14,959

is actually saying the things they are

267

00:10:12,160 --> 00:10:16,480

saying in Japanese or in whatever

268

00:10:14,959 --> 00:10:18,480

collocation piece. So that's that

269

00:10:16,480 --> 00:10:20,320

assisted AI that's allowing for a

270

00:10:18,480 --> 00:10:22,079

differentiated product that wasn't

271

00:10:20,320 --> 00:10:24,320

available before or would have taken

272

00:10:22,079 --> 00:10:26,560

months and months and driven up the

273

00:10:24,320 --> 00:10:27,920

budget beyond any kind of feasibility.

274

00:10:26,560 --> 00:10:31,519

So there are a lot of elements like

275

00:10:27,920 --> 00:10:34,160

that. outpainting. a good

276

00:10:31,519 --> 00:10:36,160

studio friend of ours called Magnopus

277

00:10:34,160 --> 00:10:40,880

and their downstream visual effects

278

00:10:36,160 --> 00:10:43,120

vendors put on a massive project at

279

00:10:40,880 --> 00:10:43,680

Sphere called Wizard of Oz Sphere.

280

00:10:43,680 --> 00:10:46,560

If you've seen it, it's amazing.

281

00:10:44,800 --> 00:10:46,880

I was just going to that example.

282

00:10:46,880 --> 00:10:51,120

And so that was originally shot in

283

00:10:49,360 --> 00:10:51,519

4x3 ratio.

284

00:10:51,519 --> 00:10:56,000

low resolution, right? Because it's

285

00:10:54,160 --> 00:10:57,360

in the wayback machine. Well, all of a

286

00:10:56,000 --> 00:10:59,040

sudden you have to contemplate, well,

287

00:10:57,360 --> 00:11:00,240

what's happening out of camera? So you

288

00:10:59,040 --> 00:11:01,440

you think of these things not just

289

00:11:00,240 --> 00:11:03,200

outping. Yeah.

290

00:11:01,440 --> 00:11:05,120

But you but think about the main hero

291

00:11:03,200 --> 00:11:07,680

characters that have suddenly gone out

292

00:11:05,120 --> 00:11:10,720

of the 43 ratio. So you use an

293

00:11:07,680 --> 00:11:12,720

assist of AI model to contemplate that

294

00:11:10,720 --> 00:11:14,959

work and then go back into the

295

00:11:12,720 --> 00:11:18,000

traditional workflow to actually render

296

00:11:14,959 --> 00:11:20,880

and composite and get that thing on

297

00:11:18,000 --> 00:11:22,640

the massive high resolution 16K screen.

298

00:11:20,880 --> 00:11:25,200

Yeah. And so it's things like that where

299

00:11:22,640 --> 00:11:27,519

they it wouldn't previously be possible

300

00:11:25,200 --> 00:11:29,519

to do the things that we're doing and AI

301

00:11:27,519 --> 00:11:30,320

is giving us that helpful nudge in that

302

00:11:29,519 --> 00:11:32,320

direction.

303

00:11:30,320 --> 00:11:34,560

Absolutely. I've had the chance to watch

304

00:11:32,320 --> 00:11:36,079

a little bit of episode of Oz. It's just

305

00:11:34,560 --> 00:11:38,399

mindboggling. Yeah.

306

00:11:36,079 --> 00:11:41,120

So if you think and if you have to kind

307

00:11:38,399 --> 00:11:43,279

of take a you know I know people are

308

00:11:41,120 --> 00:11:45,200

hesitant in predicting but you know what

309

00:11:43,279 --> 00:11:46,880

are you guys focused on based on your

310

00:11:45,200 --> 00:11:48,640

customer requirements where the industry

311

00:11:46,880 --> 00:11:50,720

is heading where do you see the industry

312

00:11:48,640 --> 00:11:53,360

heading and advancements happening in

313

00:11:50,720 --> 00:11:54,959

for the creative industry in

314

00:11:53,360 --> 00:11:57,839

the next 12 to 24 months.

315

00:11:54,959 --> 00:12:00,000

Yep. So we have started kind of

316

00:11:57,839 --> 00:12:03,360

back to my sort of thought leadership

317

00:12:00,000 --> 00:12:05,360

point earlier. we've started with

318

00:12:03,360 --> 00:12:07,519

sort of the ideation phase of what this

319

00:12:05,360 --> 00:12:09,440

looks like building a framework

320

00:12:07,519 --> 00:12:11,600

around this and the framework

321

00:12:09,440 --> 00:12:14,000

consists of you know things that

322

00:12:11,600 --> 00:12:16,000

hyperscalers and CoreWeave certainly

323

00:12:14,000 --> 00:12:17,519

differentiates itself in the AI realm

324

00:12:16,000 --> 00:12:19,519

with the infrastructure layer and that's

325

00:12:17,519 --> 00:12:21,440

the foundation of everything right do we

326

00:12:19,519 --> 00:12:25,120

have the infrastructure and you see like

327

00:12:21,440 --> 00:12:27,680

CoreWeave and others this sort of AI

328

00:12:25,120 --> 00:12:29,440

cloud in particular you know the sort of

329

00:12:27,680 --> 00:12:32,480

the infrastructure requirements of

330

00:12:29,440 --> 00:12:36,320

AI are quite different and quite bespoke

331

00:12:32,480 --> 00:12:37,040

compared to your historical cloud

332

00:12:36,320 --> 00:12:38,800

provider.

333

00:12:37,040 --> 00:12:40,160

And so it sort of sets this

334

00:12:38,800 --> 00:12:42,240

foundation of infrastructure and then

335

00:12:40,160 --> 00:12:44,240

you look at models and models are

336

00:12:42,240 --> 00:12:46,000

certainly the basis for helping the

337

00:12:44,240 --> 00:12:48,959

creatives understand the world that it

338

00:12:46,000 --> 00:12:50,320

is. that's an interesting point that

339

00:12:48,959 --> 00:12:52,079

I'll I'll kind of touch on a bit and

340

00:12:50,320 --> 00:12:54,160

maybe we circle back to at some point.

341

00:12:52,079 --> 00:12:54,959

But there is this idea of data

342

00:12:54,160 --> 00:12:57,360

providence.

343

00:12:54,959 --> 00:12:59,040

Yeah. that's really causing a lot of

344

00:12:57,360 --> 00:13:00,800

strife and I think going to drive a lot

345

00:12:59,040 --> 00:13:04,160

of solutions in our space which is

346

00:13:00,800 --> 00:13:06,000

really around is the data that the model

347

00:13:04,160 --> 00:13:06,320

has trained against clean.

348

00:13:06,320 --> 00:13:10,800

And can you trace it back to a certain

349

00:13:08,800 --> 00:13:13,519

area and so this is where we're playing

350

00:13:10,800 --> 00:13:16,000

as well from a software as a service

351

00:13:13,519 --> 00:13:18,240

standpoint within CoreWeave which is we

352

00:13:16,000 --> 00:13:20,800

are allowing studios to build luras. So

353

00:13:18,240 --> 00:13:22,639

that's low rank adaptations. Yeah. And

354

00:13:20,800 --> 00:13:25,440

so what that allows a studio to do is

355

00:13:22,639 --> 00:13:27,440

take their own internal IP against that

356

00:13:25,440 --> 00:13:30,160

base model that's representing the world

357

00:13:27,440 --> 00:13:33,600

as we know it and train it specifically

358

00:13:30,160 --> 00:13:36,800

on that IP so that they can then go ask

359

00:13:33,600 --> 00:13:39,760

that custom model questions around hey

360

00:13:36,800 --> 00:13:41,040

this is my IP. My IP is of this famous

361

00:13:39,760 --> 00:13:43,120

dinosaur.

362

00:13:41,040 --> 00:13:46,399

Now I want to see that famous dinosaur

363

00:13:43,120 --> 00:13:48,160

in this or I want to match move these

364

00:13:46,399 --> 00:13:51,120

things or I want to track these. So it

365

00:13:48,160 --> 00:13:54,160

it opens up this potential of this

366

00:13:51,120 --> 00:13:55,839

jumping off point of a customization of

367

00:13:54,160 --> 00:13:57,279

the model in particular and we see that

368

00:13:55,839 --> 00:13:59,360

as also as a layer.

369

00:13:57,279 --> 00:14:01,600

So once we establish those first three

370

00:13:59,360 --> 00:14:03,279

layers infrastructure models and then

371

00:14:01,600 --> 00:14:05,839

customized models

372

00:14:03,279 --> 00:14:09,199

now you start building workflows you

373

00:14:05,839 --> 00:14:11,040

start building agents that behave on

374

00:14:09,199 --> 00:14:12,560

behalf of producers and directors

375

00:14:11,040 --> 00:14:15,279

autonomously

376

00:14:12,560 --> 00:14:16,399

and then you have ultimate creativity

377

00:14:15,279 --> 00:14:18,560

and flexibility.

378

00:14:16,399 --> 00:14:20,399

Yeah. So this is all aligned with your

379

00:14:18,560 --> 00:14:20,959

pyramid model if I correct. Perfect.

380

00:14:20,399 --> 00:14:21,519

Awesome.

381

00:14:20,959 --> 00:14:23,519

Correct.

382

00:14:21,519 --> 00:14:25,839

So this is going to be a very

383

00:14:23,519 --> 00:14:28,800

fascinating and fast evolving industry

384

00:14:25,839 --> 00:14:32,160

in the next 12 to 24 months and if I see

385

00:14:28,800 --> 00:14:34,320

there lot of you know fantastic

386

00:14:32,160 --> 00:14:36,000

you know innovations happening in the

387

00:14:34,320 --> 00:14:38,160

horizontal space but in the vertical

388

00:14:36,000 --> 00:14:40,320

space an area which is really poised and

389

00:14:38,160 --> 00:14:42,320

I see the media and industry also

390

00:14:40,320 --> 00:14:44,560

embracing it now in a much better way.

391

00:14:42,320 --> 00:14:46,800

Now if you have to give your parting

392

00:14:44,560 --> 00:14:48,880

advice to the viewers of this particular

393

00:14:46,800 --> 00:14:51,120

podcast, what would it be and what would

394

00:14:48,880 --> 00:14:52,639

you advise them to kind of do tomorrow

395

00:14:51,120 --> 00:14:55,120

or the next you know

396

00:14:52,639 --> 00:14:56,639

I think it would really be around that

397

00:14:55,120 --> 00:14:56,959

incremental improvements.

398

00:14:56,959 --> 00:14:58,800

Looking towards incremental

399

00:14:58,160 --> 00:14:59,279

improvements.

400

00:14:59,279 --> 00:15:03,920

It's when you when you take a small

401

00:15:02,000 --> 00:15:05,600

slice of your workflow, you know, look

402

00:15:03,920 --> 00:15:08,959

for the mundane, look for the things

403

00:15:05,600 --> 00:15:09,440

that are timeintensive but creativity

404

00:15:08,959 --> 00:15:12,800

limited.

405

00:15:09,440 --> 00:15:14,959

Yeah. and build high level objectives

406

00:15:12,800 --> 00:15:19,360

and return on investment objectives

407

00:15:14,959 --> 00:15:22,720

around possible replacement with AI

408

00:15:19,360 --> 00:15:24,800

or assisted AI or agentic AI because I

409

00:15:22,720 --> 00:15:27,440

think the sort of the failure thus

410

00:15:24,800 --> 00:15:29,120

far when we've been in the FUD of

411

00:15:27,440 --> 00:15:31,199

prompt to movie yeah

412

00:15:29,120 --> 00:15:33,360

is really around okay well could this

413

00:15:31,199 --> 00:15:36,240

replace my entire studio and then that

414

00:15:33,360 --> 00:15:38,000

makes everyone posture in terms of okay

415

00:15:36,240 --> 00:15:39,760

what is my personal brand and what do I

416

00:15:38,000 --> 00:15:42,079

bring to the table and I think if we

417

00:15:39,760 --> 00:15:44,320

really focus on empowering the

418

00:15:42,079 --> 00:15:45,120

individuals and the creatives in the

419

00:15:44,320 --> 00:15:47,040

space

420

00:15:45,120 --> 00:15:49,040

and unleashing them and allowing them to

421

00:15:47,040 --> 00:15:51,360

get to that last perfect pixel. I think

422

00:15:49,040 --> 00:15:53,199

that's going to bode really well for

423

00:15:51,360 --> 00:15:55,040

some of these studios moving forward to

424

00:15:53,199 --> 00:15:57,040

differentiate themselves in the market.

425

00:15:55,040 --> 00:16:00,639

Awesome. I like to say that you know

426

00:15:57,040 --> 00:16:02,079

it's not a eitheror. Every studio if

427

00:16:00,639 --> 00:16:03,839

they don't explore and get into the

428

00:16:02,079 --> 00:16:05,839

journey of infinite possibility they'll

429

00:16:03,839 --> 00:16:07,920

definitely be left behind. So they

430

00:16:05,839 --> 00:16:10,320

better get onto this journey. Thank you

431

00:16:07,920 --> 00:16:12,079

so much Matt. So wonderful to hear

432

00:16:10,320 --> 00:16:13,759

and get your perspective and insights.

433

00:16:12,079 --> 00:16:15,680

I'm sure our viewers enjoyed it. And

434

00:16:13,759 --> 00:16:17,440

thank you everyone for listening to us

435

00:16:15,680 --> 00:16:19,600

and stay tuned and we have more many

436

00:16:17,440 --> 00:16:22,839

more interesting episodes coming up.

437

00:16:19,600 --> 00:16:22,839

Thank you.

‍