0:00

Support for the show comes from

0:02

Alex Partners. The market is evolving

0:04

at a breakneck speed, and we're

0:06

only starting to understand how disruptive

0:09

forces such as AI, cyber threats,

0:11

and tariffs will change the game.

0:13

The winners will be those who

0:15

prioritize execution and know when to

0:17

adapt. And for unparalleled insights, they

0:19

can turn to the Alex Partners

0:22

Disruption Index. Stay tuned to hear

0:24

more about it later in the

0:26

show. In the face of disruption,

0:28

Businesses trust Alex Partners to get

0:30

things done when it really

0:32

matters. Read more on the

0:35

latest trends in C-sweet insights

0:37

at disruption dot Alex partners.com.

0:40

Disruption dot alix partners.com. Support

0:42

comes from service now. We're for

0:44

people doing the fulfilling work they actually

0:46

want to do. That's why this ad

0:48

was written and read by a real

0:51

person and not AI. You know what

0:53

people don't want to do? Boring busy

0:55

work. Now with AI agents built into

0:57

the ServiceNow platform, you can automate millions

0:59

of repetitive tasks in every corner of

1:01

your business, IT, HR, and more, so

1:03

your people can focus on the work

1:05

that they want to do. That's putting

1:08

AI agents to work for people. It's

1:10

your turn. Visit servicenow.com. Support

1:12

for Decoder comes from AR.

1:14

Have you ever wondered what's

1:16

powering your smartphone and other

1:19

devices we interact with daily,

1:21

or what lies at the

1:23

heart of life-savingaving drug discoveries

1:25

and roboticotic surgeries? The answer

1:28

is AR. Arm Technology is

1:30

moving the world forward, enabling

1:32

AI to create a more

1:34

meaningful, more connected life for

1:37

everyone, everywhere. Arm believes the

1:39

future isn't about technology. It's

1:41

about people, and the possibilities

1:44

technology can offer us

1:47

all. The future is built on

1:49

Arm. You can discover

1:51

more at arm.com/Discover. Today

1:53

I'm talking with Kako

1:55

Shavastav, CEO of music

1:57

creation platform Splice. I

2:00

don't think I... need

2:02

to introduce Splice. I

2:04

just need to play

2:06

this clip. If you

2:08

exist on planet Earth,

2:10

you know that guitar

2:12

loop. It's the main

2:14

loop from Sabrina Carpenters

2:16

hit Espresso, which is

2:19

an inescapable pop music

2:21

phenomenon. That

2:31

loop is part of a sample pack you

2:33

can get on Splice. In fact, most of

2:35

Espresso is built from samples you can get

2:38

on Splice, which is one of the biggest

2:40

marketplaces for loops and samples around. Anyone can

2:42

just go sign up, pay the money, download

2:44

royalty-free loops, and try to make pop hits

2:47

all day long. Making songs like this is

2:49

just part of music making now, and it

2:51

has been ever since Riana's monster hit Umbrella

2:53

was built around a garage band loop called

2:56

Vintage Func03 in 2007. Seriously,

3:02

this just comes with garage band.

3:04

Now if you're at a Coder

3:06

listener, you know that some of

3:08

my favorite conversations are with the

3:10

people who build technology products for

3:12

creatives, and that I'm obsessed with

3:14

how technology changes the music industry

3:16

in particular, because it feels like

3:18

whatever happens to music happens to

3:20

music happens to everything else five

3:22

years later. So talking a cockle

3:24

really hit all those notes for

3:26

me, because Splice is wrapped up

3:28

in all of it. And some

3:30

of its new products, including some

3:32

new AI tools, might change how

3:34

music is made all over all

3:37

over again. Cockle's been the CEO

3:39

of Splice for three years now.

3:41

Before that, she was with Adobe.

3:43

So she has a lot of

3:45

experience working at a company that

3:47

makes tools for a creative user

3:49

base that's threatened by things like

3:51

AI. But if you've listened to

3:53

any of our Adobe episodes or

3:55

episodes about other kinds of AI-powered

3:57

creative tools, you know that the

3:59

flip side is that people actually

4:01

use the tools at high rates,

4:03

because they're fun to play with

4:05

and they do. hard lines about

4:07

what Splice will and won't do

4:09

with AI, but also to see

4:11

how the broader music industry can

4:13

try and make sense of all

4:15

this technology and what it might

4:17

do to music in the future.

4:19

I also wanted to talk about

4:22

how Splice is navigating the incredibly

4:24

complex minefield of copyright law and

4:26

attribution on the internet, something that's

4:28

only getting more complicated with AI

4:30

and the ever-increasing number of copyright

4:32

lawsuits filed against the big AI

4:34

companies by creatives of all kinds.

4:36

There's a lot in this one.

4:38

And Kaku was willing to fall

4:40

pretty deep on some of these

4:42

rabbit holes with me. I think

4:44

you'll like it. Let me know

4:46

what you think. Okay, Splice CEO

4:48

Kaku Shirestov. Here we go. I

4:50

have wanted to do this conversation

4:52

forever, so I'm glad I'm here.

4:54

Yeah, we ran into each other

4:56

at the code conference last year

4:58

and we just, we're off to

5:00

the races talking about music and

5:02

technology and AI, and I'm glad

5:04

you're finally here. Because so much

5:07

has changed since then, but all

5:09

of the issues are kind of

5:11

still there and still working towards

5:13

resolution. Yeah, yeah, a lot's changed,

5:15

lots going to keep changing. And

5:17

you're right. Some of the core

5:19

issues are still the core issues.

5:22

And along the way, at least

5:24

one like massive hit single has

5:26

been made using loops from Splice.

5:29

So there's that. Oh, come on,

5:31

not just one, not just one.

5:34

I think Aspreso, there's Aspreso and

5:36

there's like a lot of other

5:38

ones. I love Aspreso, it's, you

5:41

know, it's awesome, but a very

5:43

large proportion of top music everywhere

5:46

using Splice. What do you do

5:48

for folks? Splice is a music

5:50

creation platform that is used by

5:53

music creators musicians music creators. That's

5:55

our focus, you know, who are

5:57

our creators. And what we provide

6:00

to them, we have this tagline

6:02

start with sound, you know, so

6:05

we do start with sound. And

6:07

we provide them with probably the

6:09

world's most diverse, most high quality

6:12

sonic palate. We send people all

6:14

over the world. Right before this,

6:16

I was looking at a report

6:18

from our team that just came

6:20

back from Brazil. And we're recording

6:22

sounds, we're meeting artists on the

6:24

ground. So we're capturing the sounds

6:26

of the sounds of the sounds

6:29

of the world. and we make

6:31

that available through our platform. We

6:33

also provide AI-based creative tools that

6:35

help you start with the sound,

6:37

but make it your own. We

6:39

have compositional AI. We just launched

6:41

something brand new at South by

6:43

Southwest called Splice Mike, which allows

6:45

you to hum an idea, start

6:47

with a musical idea right in your

6:49

phone, and we'll help you compose around

6:52

that by putting the right samples next

6:54

to it to help you get you

6:56

to your final track. So there's

6:58

a lot here, which is here are the

7:00

foundational pieces of making a song, right? We're

7:02

going to do loops and samples. We're going

7:04

to have this library of audio. And then

7:07

there's this turn, which I see a lot

7:09

of companies that make creative software starting to

7:11

make, which is we're going to do it

7:13

for you. It used to work at Adobe,

7:16

Adobe I think is the. the paradigmatic example

7:18

of this right now, you can just push

7:20

generative fill in Photoshop and it just does

7:22

a bunch of stuff for you. You can

7:24

prompt Photoshop now in various ways and it

7:27

does stuff for you. Are you all the

7:29

way there with Splice and where you're going

7:31

where you can say, write me a country

7:33

song and Splice will just do it for

7:35

you? Actually, we're totally not. That's totally not

7:37

what we're trying to do. And I'm so

7:39

glad to ask this question because I really

7:42

want to put this idea out there. Our

7:44

creatives, our musicians, our artists, the

7:46

people who we think about all

7:48

day long, the last thing they

7:50

want is someone to make the

7:52

song for them. In fact, one

7:55

of the things that we learned

7:57

when we launched Create out of the

7:59

gate was People are like, oh,

8:01

this feels like cheating. This feels

8:03

too easy. I need more controls.

8:05

And so right from day one,

8:07

we've been adding more sophistication, more

8:09

technology, and more customization, more personalization

8:11

for users out of the gate.

8:14

Because for our users, it's really

8:16

about the creative process. And how

8:18

is that interesting? How are they

8:20

able to get the tools to

8:22

kind of capture what's happening inside?

8:24

and turn it into a song,

8:26

turn it into a vibration, turn

8:28

it into something that they can

8:30

share with other people. So it

8:33

is absolutely not push-button creation. That's

8:35

not fun. This is a long

8:37

argument in music. It goes past

8:39

decades. It stretches before AI hit

8:41

the scene. You must at the

8:43

company have some sense of how

8:45

people perceive building music out of

8:47

sample packs. And even before espresso

8:49

umbrella by Riana was garage band

8:52

loo, which I think is... That's

8:54

just a moment in music that

8:56

should belong in the history books,

8:58

you know? Totally. How have you

9:00

dealt with that? Okay, music is

9:02

now just assembling a bunch of

9:04

pre-made samples and that's good or

9:06

bad. People have a lot of

9:08

feelings about that. Is that a

9:10

framework you're using as you enter

9:13

sort of the AI generation era?

9:15

So I'm going to take exception

9:17

with what you just said. Sure.

9:19

I don't think music making today

9:21

is putting just a bunch of

9:23

samples together. I think that... By

9:25

the way, I will concede that

9:27

that is a very reductive criticism,

9:29

but it is a criticism. I

9:32

think of samples as the building

9:34

blocks for how modern music is

9:36

made, and it used to be

9:38

a hip-hop thing, it used to

9:40

be, you know, very specific genres,

9:42

and now it's in every genre.

9:44

One of our largest growing genres

9:46

is country music, which I never

9:48

thought would happen, but it is.

9:50

You're using samples to make country

9:53

music. I think the... Artistry of

9:55

using samples to make music is

9:57

that you start with a sample,

9:59

you start with the sound, but

10:01

then you make it your own.

10:03

One, how you assemble it, but

10:05

how you change it. the sound,

10:07

how you vary it, what you

10:09

do inside the digital audio workstation,

10:12

which is like the primary creative

10:14

canvas. I think that's really important.

10:16

Sorry, I got really heeded about

10:18

this, this is important to me.

10:20

No, I asked that question somewhat

10:22

to provoke that response, right? Because

10:24

like I said, it is a

10:26

reductive criticism, but it's a criticism

10:28

that has existed. And I guess

10:30

I'm curious, you have that response

10:33

to the criticism of sample usage.

10:35

Fundamentally, it's about what is the

10:37

creative process. And I personally spend

10:39

a ton of time with craters.

10:41

And what they are telling me

10:43

over and over and over again

10:45

is I want better tools. And

10:47

when I was at Adobe, this

10:49

is also something that we heard

10:52

from people. I want better tools.

10:54

And so the work for us,

10:56

the work for any company that's

10:58

wanting to really meet the needs

11:00

of this growing and large market.

11:02

is how do you build better

11:04

tools in this era of AI?

11:06

It's not going to be, oh,

11:08

let me type a bunch of

11:11

prompts and I get a song

11:13

out at the end. What happens

11:15

next? How do I edit that?

11:17

How do I change this particular

11:19

part of the song and get

11:21

it to sound a certain different

11:23

way? How do I take this

11:25

sample and make it into something

11:27

else? How do I get my

11:29

musical idea? And I mean, and

11:32

you saw this with the Splice

11:34

Mike launch as well. A lot

11:36

of it is... How do we

11:38

get more of you into the

11:40

music creation process as quickly as

11:42

possible? So that's fundamental, whether we're

11:44

talking about using a synthesizer to

11:46

make music or samples to make

11:48

music or AI to make music,

11:51

how do you make sure the

11:53

creative process is respected throughout those

11:55

different transitions and music innovation? That's

11:57

a lot of sort of incoming

11:59

about what your product should look

12:01

like, right? You're getting feedback from

12:03

artists, from musicians, from other creators.

12:05

Yeah. There's another side of the

12:07

puzzle, particularly music, which is... copyright

12:09

holders, labels. Now there are these

12:12

huge private equity companies that own

12:14

huge catalogs that want to assert

12:16

their rights in various ways. There's

12:18

the distributors themselves, now like Spotify

12:20

and YouTube. Do they have a

12:22

point of view that's informing how

12:24

you're using AI or how you're

12:26

thinking about sample licensing? Because that

12:28

seems like the most complicated part

12:31

of your business. Yes and no.

12:33

We are aligned across the industry,

12:35

whether it's with universal music or...

12:37

any of the other key, you

12:39

know, high quality players in the

12:41

industry, we are very aligned that

12:43

the rights of the creator have

12:45

to be respected. And again, our

12:47

position is super simple. We're going

12:50

to focus on the craters and

12:52

what craters want, and we're going

12:54

to try to meet their needs.

12:56

So the rights of the craters

12:58

have to be respected. On the

13:00

splice side of things... We take

13:02

this pretty seriously and we take

13:04

it seriously throughout the entirety of

13:06

our process, starting with ingestion, right?

13:08

How does a sample producer, sample

13:11

pack creator, come into the Slice

13:13

platform? We have an entire organization

13:15

that sort of does the intake,

13:17

does the QC, checks the providence.

13:19

If you're telling us you're the

13:21

creator, do we know that you're

13:23

really the creator, etc. So that...

13:25

process of ingesting is something we

13:27

take seriously. Is it tagged effectively?

13:30

Is it tagged appropriately all the

13:32

way through getting onto the platform?

13:34

And all the quality stuff, like

13:36

is the sound clear? Is the

13:38

recording nice? All of that. To

13:40

the other end of it, which

13:42

is, what is the experience of

13:44

someone who is downloading a sample

13:46

from Splice and able to use

13:48

it? We want to make sure

13:51

that every single download that you

13:53

do on Splice. has the potential,

13:55

but you can download the PDF

13:57

that says you've got royalty free,

13:59

you've got full rights to this

14:01

material to use it for any

14:03

kind of creation. So that's something

14:05

that's a basic part of what

14:07

we do, and it's been that

14:10

way. for a long time. On

14:12

top of that is the AI

14:14

story, and that's the big story

14:16

that everyone's talking about. I think

14:18

there, it's really simple as well.

14:20

It should be, which is if

14:22

you're gonna use content to train,

14:24

you should train on content that

14:26

you have rights to. It's not

14:28

okay to disrespect the rights of

14:31

creators. And I think, again, most

14:33

players in this space are pretty

14:35

aligned on that. Yeah, it occurs

14:37

to me just as you describe

14:39

that that you are a creator

14:41

platform for creators. There are people

14:43

who sit around making sample packs

14:45

and then they might make money,

14:47

put uploading sample packs to splice,

14:50

and then on the other end

14:52

you've got artists who are downloading

14:54

sample packs, paying you money to

14:56

go use them in other songs.

14:58

That's a unique situation. Are there...

15:00

It really is. Can I just

15:02

go back to your sample thing

15:04

because you really poked me on

15:06

that and I want to come

15:09

back to that for one second?

15:11

That's what's really magical about using

15:13

samples to make music. It's not

15:15

just a random sound that you

15:17

got on splice, there's an artist

15:19

at the other end of that.

15:21

And we work with, you know,

15:23

so just the Sao Paulo team,

15:25

we work with some of the

15:27

people who are really at the

15:30

forefront of funk and what that

15:32

means and what that sound is

15:34

and how it's evolving. So when

15:36

you're using a sample pack from

15:38

splice, you're collaborating with those people

15:40

and it's a collaboration, it's a

15:42

storytelling between those two different artists

15:44

coming together. I think that's really

15:46

fun. Yeah. It's a neat part

15:49

of the place. Are there creators

15:51

who make their entire living just

15:53

making sample packs for you? There

15:55

are some, yeah. Is that a

15:57

viable approach to being a professional

15:59

musician? I think it is for

16:01

some people. For some people, you

16:03

know, they make hundreds of thousands

16:05

of dollars. For some people, you

16:07

know, they're building their own musical

16:10

career and this is part of

16:12

what they're doing. So we see

16:14

a big range of people. I

16:16

will say that the revenue that

16:18

we've shared. with the artists on

16:20

our platform over time. It's at

16:22

an all-time high. So it's nice

16:24

to feel good, you know, to

16:26

be able to feel good about

16:29

that too. I've been spending a

16:31

lot of time just thinking about

16:33

the economics of creator platforms, whether

16:35

they're sustainable over time, on sort

16:37

of the big consumer platforms, you

16:39

see that creators have to augment

16:41

their income. They all have to

16:43

do brand deals, they all have

16:45

to do brand expansions or sponsored

16:48

content or whatever. You can't really

16:50

do that on a splice. Sabrina

16:52

Carpenter makes espresso. I'm guessing the

16:54

person who made that sample pack

16:56

did not get paid more money

16:58

because that song was a hit

17:00

just based on how your licenses

17:02

work. That's the that's the pro

17:04

and the con of being royalty

17:06

free. We are royalty free and

17:09

that what that means for the

17:11

creators is they don't have to

17:13

get stressed about it. You can

17:15

use the sample. It's clear. You

17:17

know, you don't have to worry

17:19

about clearing the rights. The downside

17:21

is you don't get to share.

17:23

in the sort of upside when

17:25

something big like that happens. We're

17:28

really here to make sure that

17:30

as many people can create as

17:32

possible and you know that that's

17:34

part of how our model works.

17:36

How does your revenue work? What

17:38

does Splice take the money? We

17:40

are a subscription platform so people

17:42

buy a subscription to Splice and

17:44

that gives them access to this

17:46

unlimited library of sounds along with

17:49

the creative tools that were... investing

17:51

in heavily for the future, you

17:53

get a certain number of credits

17:55

per month and use those credits

17:57

to download sounds that you can

17:59

then use as you want. And

18:01

is growth just getting more and

18:03

more artists to use splice on

18:05

both sides as creators and as

18:08

people who are subscribers? Yeah, that

18:10

is growth in both. It's been

18:12

a interesting journey over the last

18:14

three years while I've been here,

18:16

but growth is really good. We

18:18

need to take a quick break.

18:20

We'll be right back. This is

18:22

advertiser content from nerd wallet. The

18:24

brain can store vast amounts of

18:26

information up to 2.5 million gigabytes.

18:29

But unlike a hard drive, our

18:31

brain is always reorganizing information based

18:33

on what's most important to us.

18:35

So instead of using brain power

18:37

researching, say, credit cards or auto

18:39

insurance, the nerds at nerd wallet

18:41

can give your brain a break

18:43

to find the right financial product

18:45

for you. It's this kind of

18:48

processing that researchers like Carlos Gershinson

18:50

study as they explore how our

18:52

brains are both like and unlike

18:54

computers. This depends on how our

18:56

brain evolved. So there are some

18:58

things that might seem complex, but

19:00

since our ancestors relied on solving

19:02

certain problems for survival, then we

19:04

do it naturally. And there are

19:07

other things that might seem not

19:09

so complex, but since our brains

19:11

were not evolved to solve those

19:13

tasks, it is difficult for us.

19:15

Thankfully, that's when our brains can

19:17

reach for other tools. Extended mind

19:19

theory basically says that since our

19:21

brains are limited, we extend cognitive

19:23

abilities to our environment. We are

19:25

becoming smarter simply by relying on

19:28

this external technology. Neuro wallet could

19:30

be seen in context of the

19:32

extended mind, because instead of memorizing

19:34

where I spend what, then you're

19:36

just extending. When it comes to

19:38

your finances, you're not expected to

19:40

remember everything. Give your brain a

19:42

break and let nerd wallet find

19:44

the right financial products for you.

19:47

Get matched in minutes at nerd

19:49

wallet.com. Missions to Mars, driverless cars,

19:51

AI chapots. Feels like we're already

19:53

living in the future. Well, Robin

19:55

Hood is built for the future

19:57

of trading. Robin Hood's intuitive design

19:59

makes trading seamless. Spot opportunities and...

20:01

Take control of your trades with

20:03

tools like screeners, simulated returns, and

20:05

strategy builder. On Robin Hood, traders

20:08

have access to hundreds of popular

20:10

stocks and ETFs, 24 hours a

20:12

day, 5 days a week, so

20:14

you can keep up with today's

20:16

fast-paced markets. You can now even

20:18

trade stocks in crypto all in

20:20

one place with a new desktop

20:22

platform, Robin Hood legend. The future

20:24

of trading is fast, powerful, and

20:27

precise. Experience it now on Robin

20:29

Hood. Investing is risky. Trading during

20:31

extended hours involves additional risk. Cryptocurrency

20:33

services are offered through an account

20:35

with Robin Hood Crypto LLC NMLS

20:37

ID 1702840. Robin Hood Crypto is

20:39

licensed to engage in virtual currency

20:41

business activity by the New York

20:43

State Department of Financial Services. Today

20:49

at T-Mobile I'm joined by a

20:51

special co-anchor. What up everybody? It's

20:54

your boy, Big Snope deal, double

20:56

sheet. Snoop! Where can people go

20:58

to find great deals? Hand to

21:00

t-mobile.com to get four iPhone six

21:02

teams with Apple intelligence on us.

21:05

Plus four lines plus four lines

21:07

for Apple intelligence on us. Plus

21:09

four lines for 25. That's four

21:11

lines for Apple intelligence, with Apple

21:13

Intelligence, with Apple intelligence,

21:16

requires IOS-18 point one

21:18

or later. We're

21:21

back with Splice O'Cuckle Shavastav, getting into

21:23

the big decoder questions to understand how

21:25

Cuckle's background at Adobe and other big

21:27

software companies has influenced how Splice is

21:30

approaching the music industry. I think that

21:32

brings us kind of to the decoder

21:34

questions. You're a newish CEO, you're three

21:36

years in, you were at Adobe before,

21:39

I think you had two different stints

21:41

at Adobe. I did, I did, right

21:43

out of business school, and when it

21:45

was a perpetual business business, and more

21:48

recently, well on its subscription journey. Adobe

21:50

is the creative software company. They have

21:52

a very, I would say, back and

21:54

forth relationship with creatives. You know, we

21:56

had Shunton and Ryan on the show.

21:59

We got feedback on that. episode of

22:01

Dakota like nothing else we've ever experienced.

22:03

People have a lot of feelings about

22:05

Adobe, what that software represents, what that

22:08

subscription is worth, what AI means to

22:10

Adobe as a company and its user

22:12

base. You obviously have some of that

22:14

experience. As you've come into Splice, how

22:17

have you thought about applying those lessons

22:19

to what is well on its way as being

22:21

one of those companies for the musical

22:23

community? Adobe's been a really

22:25

important part of my career journey. I

22:27

learned a ton of great things. at Adobe

22:29

both good and bad. I was also one

22:32

of the early people at Flickr, the

22:34

photo sharing site, which I don't know

22:36

if you ever used, but a lot

22:38

of people loved it. We... There's some

22:40

Flickr users right now who are

22:42

writing us emails. I'm just letting

22:44

you know. They still like... It's...

22:46

It's... I was also the head

22:48

of product and marketing at GitHub.

22:50

So I've had a chance to

22:53

see crater tools in multiple different

22:55

places. And all of that has

22:57

really informed what I'm bringing here

22:59

to Splice. The journey for me

23:02

has been a little bit around

23:04

pattern recognition. One thing that I've

23:06

seen at Flickr, at GitHub, some

23:08

parts of Adobe that I see

23:11

here at Splice, is that you

23:13

have a business that's centered around

23:15

content, and you have a lot

23:17

of rich metadata around that

23:20

content. and you have lots

23:22

and lots of impressions around

23:24

that content so that users

23:26

are giving you information about

23:28

it. And so at Splice, we

23:30

have about a million songs

23:33

that are samples that are

23:35

sounds that are downloaded today.

23:37

That's a lot. We have

23:39

28 million stacks that have been

23:41

created using our AI tools. So

23:43

we have a lot of impressions

23:46

of what sounds are people listening

23:48

to. How are they creating things

23:50

together? What sounds go well together?

23:52

And that's been a really interesting thing.

23:55

Once you have that data, once

23:57

you have that metadata, you can

23:59

use that. build rich experiences on

24:01

top, which is what we're doing

24:03

now with the creative tools, the

24:05

AI-based creative tools. That feels very

24:07

familiar to bring to Splice, to

24:09

bring to the music industry where

24:11

I've seen it at get hub,

24:13

I've seen it at Flickr, I've

24:15

seen it at Flickr, I've seen

24:17

it at these other places. That

24:19

turn to we're going to make

24:21

the tools that actually help you

24:23

create the music. Yeah. You can

24:25

look at it in a slightly

24:28

more abstract way, right? you go

24:30

off to the races, Splice doesn't

24:32

see what you're doing in those

24:34

apps. But those are the dominant

24:36

music creation apps. To this day,

24:38

they're still the dominant music creation

24:40

apps. You're suggesting with something like

24:42

Splice Mike or Splice Create on

24:44

the phone that you have on

24:46

mobile, that you're going to take

24:48

some of that creation, right? Particularly

24:50

on a phone, I think there's

24:52

a lot of opportunity to reinvent

24:54

how we make music. It's still

24:56

fairly cumbersome and phone screenscreens are

24:58

small. the features you launch at

25:00

South by Southwest Southwest are interesting

25:02

because they use AI to make

25:04

that a little bit faster, more

25:06

seamless, more sketchy, right? You can

25:08

sketch an idea very quickly on

25:10

a phone now. Is that the

25:13

extension? We're going to go take

25:15

some of Pro Tools Market Share.

25:17

We're going to go take some

25:19

of Logics Market Share. So I

25:21

think that word take suggests a

25:23

zero-sum game. This is not a

25:25

zero-sum game. It's about expanding and

25:27

exploring the creative process. Many of

25:29

our users use Splice Mic or

25:31

user mobile app as an adjunct

25:33

part of their process. that they

25:35

will ultimately finish inside a digital

25:37

audio workstation. And I love to

25:39

see that. So one of our

25:41

super, super top-end producers has worked

25:43

with all, many of the big

25:45

names that you would recognize. He'll

25:47

tell me, you know, I'll get

25:49

into my Uber, I'll start playing

25:51

with the Splice app, I'll generate

25:53

a bunch of stacks, but by

25:55

the time I get to the

25:57

studio, I've got a bunch of

26:00

ideas that I can show the

26:02

artists right away, and then... That's

26:04

a really core part of his

26:06

creative process. You know, I was

26:08

just at my kids' school where

26:10

they have a digital music production

26:12

class. And for them, listening to

26:14

sounds on Splice is a really

26:16

core part of learning. What does

26:18

this genre sound like? What does

26:20

it mean to create a Bollywood

26:22

hit? What does it mean to

26:24

create something that's a K-pop sound?

26:26

And I think that's a different

26:28

way. to use this experience. So

26:30

for us, it's not, we're going

26:32

to take it away from this

26:34

place or this place, but how

26:36

do we expand how much we're

26:38

part of the creative journey in

26:40

different ways? But the idea that

26:42

you're going to start and finish

26:45

a song in Pro Tools, you're

26:47

not looking at that. Do we

26:49

think that we're going to directly

26:51

compete with a Pro Tools? No.

26:53

I don't think so. I think

26:55

Pro Tools has its place. Just

26:57

like Photoshop has its place. There

26:59

are people who tell us every

27:01

single day, you will take Ableton

27:03

out of my cold dead hands.

27:05

It's just not going to happen.

27:07

And there are a lot of

27:09

other parts of the creative process

27:11

that are painful. So for example,

27:13

when I sit down with one

27:15

of our creators, inevitably there will

27:17

be a situation where they will

27:19

go and they're like, oh, we

27:21

need to find a certain kind

27:23

of kick drum. They'll find a

27:25

folder and they'll do a sub

27:27

folder and they'll do a sub

27:30

folder and then they'll finally find

27:32

the sub sub sub folder that

27:34

has 20 Hick drum sounds that

27:36

they have like saved and they

27:38

And you just go through and

27:40

you listen to these sounds that

27:42

is painful. That is a painful

27:44

process and it shouldn't be that

27:46

hard And so we've just done

27:48

this new experience that we launched

27:50

in October last year, where we

27:52

integrated with Studio One, which is

27:54

one of the top digital audio

27:56

workstations. And there's a splice integrated

27:58

searchwood sound experience. So we listen

28:00

to what you're creating inside Studio

28:02

One, and we'll suggest the samples

28:04

that go with it right there,

28:06

integrated as part of. creative workflow.

28:08

Do I think I'm going to

28:10

have a play studio one? Absolutely

28:12

not. Can I make the studio

28:15

one experience a lot better? Because

28:17

Splice is there and Splice is

28:19

smart. With AI, 100% all day

28:21

long. How do these conversations work

28:23

with all those digital audio station

28:25

providers? Right. They're all very different.

28:27

The companies that make them are

28:29

all very quirky. Some of them

28:31

are Apple, right? It's all the

28:33

music tech industry is very quirky.

28:35

It's all very quirky. There's a

28:37

lot of, I would say, centric

28:39

Europeans floating around this industry in

28:41

particular. It's great. It's one of

28:43

my favorite parts of the tech

28:45

industry to cover. And then you

28:47

have a company like Apple, right,

28:49

which is, they're just going to

28:51

do whatever they want to do.

28:53

That's just how they work. Splice

28:55

has to integrate with all of

28:57

it. Some of them are expanding

28:59

into your zone, right? They are

29:02

adding sample packs and libraries and

29:04

subscription features. A lot of them

29:06

are adding AI tools. How does

29:08

that competition and cooperation work? Generally

29:10

speaking, it's a very strong cooperation.

29:12

I've actually been really impressed at

29:14

how collaborative the industry really is.

29:16

So the conversations with Studio One

29:18

was very, very positive, and we're

29:20

working with other doc partners as

29:22

well to bring that integration, and

29:24

it's been very, very positive. I

29:26

think there's generally a recognition that

29:28

we're good at what we do,

29:30

the kind of work that we

29:32

can do in terms of bringing

29:34

these sample packs to the world.

29:36

the global coverage, the high quality

29:38

or consistent high quality control process,

29:40

etc. It's not something that they

29:42

want to replicate. They want to

29:44

make great, you know, experiences inside

29:47

Ableton, the next feature. This is

29:49

not what they want to do.

29:51

I think the AI stuff is

29:53

new to a lot of people

29:55

in the industry. I come from

29:57

a core tech background. A lot

29:59

of the team that I've brought

30:01

in to Splice over the last

30:03

few years comes from a core

30:05

tech background. We have a lot

30:07

of expertise around that, which is

30:09

unique in some ways for the

30:11

music text space. So I think

30:13

there's a lot of respect around

30:15

that. I think there's an attractiveness

30:17

to a subscription business model that

30:19

has been difficult for this industry

30:21

to kind of adopt. And so

30:23

I think there's a lot of

30:25

curiosity about that. Like, can we

30:27

use a content business model to

30:29

get more recurring revenue? But I

30:32

think many people have found that

30:34

it's not as easy as it

30:36

looks, and they've struggled with it.

30:38

One of the things you say

30:40

about bringing people who have a

30:42

core tech background, that helps you

30:44

innovate in things like AI, I'm

30:46

sure, right, where you just need

30:48

to be on the cutting edge

30:50

of the technology. Tech in music

30:52

in particular have always just crashed

30:54

into each other. The thing I

30:56

say in the show over and

30:58

over again is if you pay

31:00

attention to the music industry and

31:02

what tech is doing to the

31:04

music industry, you have a view

31:06

into what tech will do to

31:08

everything else five years now. How

31:10

are you thinking about that dynamic?

31:12

Is it just for AI or

31:14

is there something else you're trying

31:17

to accomplish with the addition of

31:19

that talent? Innovation is really important

31:21

and when I look at the

31:23

music creation process, especially as an

31:25

outsider, I feel like these music

31:27

creators have been underserved with great

31:29

innovative experiences and I think it's

31:31

important to focus on the creative

31:33

workflow and provide people better tools

31:35

over time. When I think about

31:37

the collision between tech and music,

31:39

it's weird because there's actually more

31:41

similarity than dissimilarity inside Slice. We

31:43

have some really great software developers

31:45

who love music and music creators

31:47

in their own right. A whole

31:49

bunch of musicians and artists who

31:51

think in that same weird mathy

31:53

way that great software developers think.

31:55

So there's a lot of... similarity.

31:57

Surprisingly, there's a lot of similarity.

31:59

I also think that there's this

32:01

mindset out there that musicians are

32:04

scared of. technology scared of innovation.

32:06

I actually think that musicians love

32:08

hacking. They love trying new things.

32:10

You know, the, again, there was

32:12

all this threat around synthesizers and

32:14

all of that stuff, and then

32:16

Stevie Wonder took it to like

32:18

a totally different magical new plays.

32:20

I think artists love innovation, and

32:22

it allows them better tools to

32:24

get to the other place. Well,

32:26

they don't love this push button.

32:28

creation and I think if you

32:30

stay away from that if you

32:32

stay close to the creative process

32:34

you will find the right ways

32:36

to bring technology innovation here. I

32:38

think there's something else that you're

32:40

pushing on here that I think

32:42

is important and maybe it's one

32:44

of your decoder questions around how

32:46

do you bring the cultural mindset

32:49

from the tech industry and meld

32:51

it with the music industry and

32:53

is that a difference? Is that

32:55

a challenge? Yes, that's definitely where

32:57

I'm going. I might as well

32:59

ask you the decoder questions now.

33:01

You've been the CEO for three

33:03

years. How is Splice structured today?

33:05

How have you changed it? This

33:07

is the tech part. We are

33:09

fundamentally a product company first. So

33:11

my largest organization at Splice is

33:13

the product development organization, and that's

33:15

product managers, engineers, designers, and CX.

33:17

And what's neat about that is

33:19

I do keep CX very close

33:21

to product. That's customer experience. Because

33:23

I think that tight loop is

33:25

super important. Wait, customer, I just

33:27

want to be able to make

33:29

sure CX is customer experience? Customer

33:31

support. Customer support. Okay. So support

33:34

design, engineers, PMs. They're all in

33:36

one org and that's product dev

33:38

and it's our largest org. Our

33:40

second largest org is our content

33:42

team. And those are the people

33:44

that are going to Brazil, they're

33:46

going to South Korea, they're going

33:48

to India, they're recording these sounds.

33:50

It's our quality control department, it's

33:52

our data and ingesting and metadata

33:54

tagging groups. So that's the content

33:56

org. And maybe the third thing

33:58

that I'll point out that's really

34:00

important to me. how I structure

34:02

the org is we have a

34:04

very strong central data organization that

34:06

reports directly to me. So a

34:08

lot of people put that inside

34:10

product debt, but for me data

34:12

is important for content, data is

34:14

important for marketing, data is obviously

34:16

important for finance, and how we

34:19

run the business, it's important for

34:21

product. So I have that as

34:23

a central organization, and again it

34:25

reports directly to me. How big

34:27

is twice now? How many people

34:29

is it? A little bit less

34:31

than 200 people. And how is

34:33

it split between those three groups?

34:35

Product Dev is about, I don't

34:37

know, it's our largest org, maybe

34:39

80, 60, somewhere between there, and

34:41

then content is the next biggest,

34:43

and it's somewhere between 40 and

34:45

60. One of the really interesting

34:47

things here, again, it's a creator

34:49

platform for creators, which is just

34:51

an interesting dynamic. Other creator platforms

34:53

at scale, they say they have

34:55

investments in content teams, but they

34:57

really just... hope the scale carries

34:59

them forward. Instagram does not have

35:01

some huge content team that is

35:03

traveling the world to get content.

35:06

They just wait for people to

35:08

come to them. It's the same

35:10

with YouTube or TikTok or whoever

35:12

else. They might manage some of

35:14

their top influencers, but really the

35:16

volume of content comes to them.

35:18

Is that a tipping point that

35:20

you think Splice can reach or

35:22

do you want to maintain control

35:24

over the library? It's really important

35:26

for us to make sure our

35:28

library is the highest quality. that

35:30

it can be. So it's not

35:32

going to be a free for

35:34

all where anyone is uploading anything

35:36

they want because we need to

35:38

maintain that high quality. Especially in

35:40

the age of AI, right? There's

35:42

all kinds of stuff that's being

35:44

uploaded to all of these big

35:46

platforms. So it's never going to

35:48

be that way for us. And

35:51

so that's just a core investment,

35:53

right? It's a core piece of,

35:55

I think, your cost model. How

35:57

was Spice Organized before? Again, you're

35:59

three years into it. How have

36:01

you changed that structure? Is this

36:03

still largely the same or have

36:05

you re-oriented the company? So I

36:07

think the biggest change... has been

36:09

around, I would say it's around

36:11

three big ideas, Nille, which are

36:13

core to how I run a

36:15

business. The first is data. I've

36:17

brought in a lot more data

36:19

people. It's very, very critical. The

36:21

reason that's important for me is

36:23

because I need to understand what

36:25

are customers actually care about. So

36:27

how are they voting with their

36:29

clicks as opposed to whatever opinions

36:31

everybody else has? So that's a

36:33

big investment. The second is design.

36:36

That is really where data and

36:38

the math and the science turns

36:40

into something else, which is a

36:42

real experience that people can feel.

36:44

It's where the art becomes magic.

36:46

And the reason that's important is

36:48

because we're serving creative people. And

36:50

that's what creative people do as

36:52

well as they take all these

36:54

inputs. And then it turned into

36:56

something that didn't create. So building

36:58

a strong design team that is

37:00

either made up of music creators

37:02

themselves or people who spend a

37:04

lot of time with music creators

37:06

is really important. And the third

37:08

thing that I really brought in

37:10

that's important is that we build

37:12

our products with the customers. So

37:14

everything that we launched, there are

37:16

tools that we built in to

37:18

allow people to give us feedback.

37:21

In fact, When we launched Create,

37:23

the biggest button in the Create

37:25

experience was the feedback button. It

37:27

was weird, but it was important

37:29

for us. And every single time

37:31

someone typed in something to give

37:33

us feedback, it comes into a

37:35

slack channel that's with all the

37:37

designers and the engineers and the

37:39

product managers. And so we're actively

37:41

talking about the feedback from the

37:43

customers as it's coming in and

37:45

responding to it for the next

37:47

version. I love that. absolutely love

37:49

that we build product that way.

37:51

I think everyone should build product

37:53

that way. One of the reasons

37:55

I always ask about structure on

37:57

the show is that it's a

37:59

proxy for culture. You kind of

38:01

get what you get. You make

38:03

some big choices about how things

38:06

are organized and that leads to

38:08

a culture. You're an interesting spot.

38:10

you took over for co-founders. One

38:12

co-founder left, he was a CTO

38:14

in 2019. The other co-founder, Steve

38:16

Martocchi, he's the executive chairman now,

38:18

but he's off doing another startup.

38:20

How have you thought about changing

38:22

the culture, inheriting the culture, and

38:24

the balance between the two? The

38:26

reason I love your question around

38:28

structure is because I do see

38:30

that it's proxy for values. And

38:32

that's why I answered it the

38:34

way I did around data design

38:36

building with customers. Those are fundamental

38:38

values that I want to bring

38:40

and inculcate into the company. There's

38:42

something else that we also did

38:44

that was around building culture. I

38:46

spent a lot of time listening

38:48

to the team, trying to learn

38:50

what made this culture unique. And

38:53

then I reflected back to the

38:55

organization. Hey, these are the values

38:57

that I'm hearing from you all.

38:59

Do you think this captures it?

39:01

And we came up with something

39:03

that we call our disco values.

39:05

Direct, inclusive, spliced together, crater-centric, and

39:07

optimistic. And even though these are

39:09

new values that we came up

39:11

with after I joined, They have

39:13

felt so authentic to the culture

39:15

that we have. That's existed for

39:17

a long time, but it's given

39:19

voice to it. So disco is

39:21

something we talk about a lot.

39:23

Every single new employee that comes

39:25

on talks about which disco value

39:27

they resonate with most. We use

39:29

it in performance reviews. We do

39:31

use it for shoutouts. It's a

39:33

core part of who we are.

39:35

Second, the good question, which is

39:38

also in many ways a proxy

39:40

for culture and values, is about

39:42

decisions. How do you make decisions?

39:44

What's your framework? This is something

39:46

that I'm working on. I've always

39:48

been a very math and science

39:50

kind of person. I've always been

39:52

someone who's very analytical. I use

39:54

a lot of data. You know,

39:56

I have a framework for decision-making.

39:58

I study, you know, all the

40:00

different tools. for decision-making, but as

40:02

the decisions that, decision

40:04

sets that come to

40:07

me become more complex,

40:09

and as we operate

40:11

in an increasingly more

40:13

complex world, fires, politics,

40:15

etc., I have found myself

40:17

relying more and more on

40:20

intuition, and I think

40:22

balancing those two, so I

40:24

would say that my decision-making process

40:26

is I will drown myself. in

40:29

data. I will really get deep.

40:31

People know in my team that I spend

40:33

a lot of time on our dashboard. I

40:35

will spend a lot of time watching

40:38

research videos and understanding how people

40:40

are using our tools. I will

40:42

spend a lot of time personally

40:44

talking to different customers. I'm talking

40:47

to customers all the time. And

40:49

once I've kind of drowned myself

40:51

in all this information, I'll just

40:54

try to listen deeply. And usually

40:56

the answer is very clear. We

40:58

have to take another

41:01

quick break. We'll

41:03

be right back.

41:06

Thumbtak presents the

41:08

ins and outs of

41:11

caring for your

41:14

home. Out. Procrastination.

41:17

Putting it off.

41:19

Kicking the can down

41:21

the road. In. Plans

41:23

and guides that make

41:26

it easy. Out. Carpet

41:28

in the bathroom. Like, why?

41:30

In. Knowing what to do, when

41:33

to do it, and who to

41:35

hire. Start caring for

41:38

your home with confidence.

41:40

Download Thumbtak today. Hear

41:43

that. Spring is here, and

41:45

the Home Depot has great

41:47

prices on grills to make

41:49

this season. Yours! So if

41:51

you're working on improving your

41:53

hosting skills, you're going to

41:55

want the next grill for

41:57

burner gas grill for 229

42:15

At Sierra, discover top workout

42:17

gear at incredible prices, which

42:19

might lead to another discovery.

42:21

Your headphones haven't been connected

42:23

this whole time. Awkward! Discover top

42:26

brands at unexpectedly low prices.

42:28

Sierra, let's get moving. We're

42:54

going to put this into practice

42:56

because the making creative software for

42:58

creative people in the age of

43:00

AI is about as tense as

43:03

it gets in the balance between

43:05

here's what the numbers are telling

43:07

us and here's how the people

43:09

feel. And what I mean specifically

43:11

is the numbers are telling everyone

43:13

that people are using the AI

43:15

tools. Just down the line every

43:18

software maker I've talked to has

43:20

introduced AI tools as any meaningful

43:22

value. It says the users are

43:24

using them. They're clicking the buttons,

43:26

they're doing generative fill all day

43:28

long. I'm sure you see that

43:30

in your numbers too. And then

43:33

what you hear from the creatives

43:35

on social media or online or

43:37

in letters to Congress is, get

43:39

this out of my face, they

43:41

stole everything for me. And there's,

43:43

that is about as big of

43:45

a divide in tech, in culture,

43:48

in creativity, is I have ever

43:50

experienced. And I think that is

43:52

challenging a lot of how everyone

43:54

is going to make decisions. I'm

43:56

going to read you a quote.

43:58

from one of your ostensible competitors.

44:00

And it tracks with everything you're

44:03

saying, but I suspect you are

44:05

going to disagree with this quote.

44:07

And I just want to. sit

44:09

with that for a minute. Yeah,

44:11

let's do it. So you have

44:13

said, right, creators just want to

44:15

create, they want all this stuff

44:18

to get out of their way.

44:20

So here's the CEO of Suno,

44:22

Mikey Shulman. Suno is just push

44:24

a button, it makes you a

44:26

song, right? You say country song,

44:28

it just spits out of country

44:30

song at you. And here's what

44:33

he recently said. It takes a

44:35

lot of time, a lot of

44:37

practice, you have to get really

44:39

good at an instrument or really

44:41

good at a piece of production

44:43

software. I think the majority of

44:45

people don't enjoy the majority of

44:48

time they spend making music. It

44:50

is not really enjoyable to make

44:52

music now. Now I've made a

44:54

lot of music. I have no

44:56

idea what Mikey Shulman is talking

44:58

about. I think it's quite fun

45:01

to make music. But that does

45:03

track with what you're saying, right

45:05

with what you're saying. But he

45:07

spun the knob all the way

45:09

to just prompt me for a

45:11

song. And a lot of people

45:13

reacted to this quote very strongly.

45:16

Yeah. How do you sit in

45:18

the middle of that to say

45:20

there's a line and I'm going

45:22

to enforce the line and we're

45:24

not just going to prompt it

45:26

all the way to a song?

45:28

Also, do you think he's right?

45:31

Do you think people don't enjoy

45:33

making music? Here's what I have

45:35

learned by serving creative people who

45:37

create. It's not essential for everybody.

45:39

For the people who create, it

45:41

is a sacred experience. It is

45:43

a core part of who they

45:46

are. They can't not do it.

45:48

And there is a struggle, but

45:50

the struggle is to authentically translate

45:52

what is inside you into something

45:54

else. And sometimes... Your tools will

45:56

help you, will enable you to

45:58

do that, and other times your

46:01

tools will get in the way.

46:03

And understanding the distinction between those

46:05

two is the whole ballgame. But

46:07

it's really about allowing the struggle

46:09

to come to life, giving birth

46:11

to something. new is hard, but

46:13

it's profoundly important. And to dismiss

46:16

it by this push-button set of

46:18

tools is, it's insulting. It's dismissive,

46:20

it's reductive, and I think the

46:22

creative process and creative people deserve

46:24

better. They deserve better technology that

46:26

enables them as opposed to reduces

46:28

this profound activity to a button.

46:31

So this is where I think

46:33

the line is inherently qualitative, right?

46:35

Here's what we're going to do

46:37

and here's what we're not going

46:39

to do. And the tension of,

46:41

it's not really enjoyable music now,

46:43

you can describe that as using

46:46

the software sucks, or I just

46:48

want to have an idea and

46:50

hear it as fast as I

46:52

can. And then you can describe

46:54

it the way you're saying, which

46:56

is, there's some parts of the

46:58

struggle, which are the creative process,

47:01

which make the art compelling. It's

47:03

profoundly important. If the data tells

47:05

you that people really want to

47:07

just click the button and make

47:09

the music, are your values strong

47:11

enough to not send you all

47:13

the way down the road? I

47:16

think it depends on which people

47:18

you're listening to. We're really clear

47:20

about the people that we're listening

47:22

to. We are listening to creative

47:24

people who love the process of

47:26

music creation, that it's essential for

47:28

them. Yes, are there challenges, but

47:31

the challenge is the creative process,

47:33

right? That is the challenge. And

47:35

for those people... the signals are

47:37

really clear. They do not want

47:39

push-button creativity. In fact, like I

47:41

was sharing before, when we give

47:43

them create, when we gave them

47:46

create for the first time, they're

47:48

like, this is too simple, I

47:50

don't want this. I want something

47:52

that gives me more creative freedom,

47:54

more creative control. And so for

47:56

us to signal, the people were

47:58

listening to are super clear, and

48:01

the signals they're giving us, there's

48:03

no confusion in what they want.

48:05

The other side of this marketplace

48:07

is consumers. We see consumers and

48:09

fans all the time now react

48:11

very strong. to AI generated imagery

48:13

in particular. Yeah. You make a

48:16

movie poster and it's got a

48:18

bunch of AI in it. The

48:20

fans are gonna be. It's not

48:22

just Photoshop. That movie poster is

48:24

coming coming down. It's maybe different.

48:26

It's not in your face. You

48:28

can't see that the characters in

48:31

the movie poster have 12 fingers

48:33

and their hair bleeds into the

48:35

skyscraper behind them. It's not as

48:37

obvious, but it's there. Do you

48:39

perceive that? kind of consumer or

48:41

fan backlash to AI and music

48:43

the same way that we we

48:46

see it in visual art? We

48:48

haven't seen it yet. Here's what

48:50

I have seen. Here's what I

48:52

have seen. I have seen really

48:54

clear signal from our customers that

48:56

they are not really interested in

48:58

computer generated samples. Clearly not our

49:01

strategy. We are in fact investing

49:03

in more human-created samples, human-curated samples.

49:05

This is why we're sending people

49:07

out to the, you know, sort

49:09

of subgenre locations, talking to authentic

49:11

artists, getting their voices. It's really

49:13

important for our strategy to continue

49:16

to do that, because people want

49:18

to connect with the stories of

49:20

the real artists on the other

49:22

side of the sample. So that's...

49:24

really, really important and really clear

49:26

for us. I think what an

49:28

end user who is listening to

49:31

Serena Carpenter today and will listen

49:33

to somebody else's music tomorrow, what

49:35

they can hear is going to

49:37

be interesting and is going to

49:39

evolve over time. I love that

49:41

Kendrick Lamar won the Pulitzer Prize

49:43

for music and the first people

49:46

who won the Pulitzer Prize for

49:48

music and the first people who

49:50

won the Pulitzer Prize for music.

49:52

You know, when that award was

49:54

first introduced, it was a completely

49:56

different sound. What art is and

49:58

what is acceptable changes over. time,

50:01

I would expect it to continue

50:03

to change over time. So I

50:05

don't want to reflect on that.

50:07

I know that artists will use

50:09

different tools and they will use

50:11

AI-based tools, absolutely. At least on

50:13

the visual imagery side, there's a

50:16

lot of endless sort of feudal

50:18

discussion about watermarks and encryption and

50:20

letting people know when images were

50:22

edited by AI or created by

50:24

AI. I would not say that's

50:26

come to anything and I would

50:28

say there are some... deep and

50:31

meaningful challenges with even making that

50:33

technology work consistently. There's not anything

50:35

quite like that on the music

50:37

side. Do you think there should

50:39

be? I think it's going to

50:41

be really hard to disambiguate around

50:43

sound and around images, around video.

50:46

You know, you've done, you've had

50:48

some really great conversations about this

50:50

topic on your podcast. I've listened

50:52

to them. I think it's a

50:54

really important debate and discussion to

50:56

have. There is going to be

50:58

a bunch of bad AI generated

51:01

content out there. It's already happening.

51:03

It's going to happen with music.

51:05

I think that is toothpaste is

51:07

out of the toothpaste too. I

51:09

think as an industry, we have

51:11

to do the right thing around

51:13

respecting the right subcraters and doing

51:16

the right thing with respect to

51:18

training data, respecting credits. This is

51:20

work that has to happen. I

51:22

don't think it's solved yet. Maybe

51:24

some of these cases that are

51:26

open will help us get to

51:28

the right answer. But I don't

51:31

think it's going to come out

51:33

of order marking. You talked about

51:35

the flood of AI content that's

51:37

coming. We can all see it.

51:39

The big consumer platforms are embracing

51:41

it. I think to some extent

51:43

Mark Suckerberg would love it if...

51:46

all the content on Facebook was

51:48

AI and he was paying zero

51:50

out to creators. I think to

51:52

some extent YouTube is really leaning

51:54

into the idea that you should

51:56

interact with your favorite creators through

51:58

AI avatars. and that they should

52:01

make even more videos, or AI

52:03

should help them make even more

52:05

videos to increase the volume of

52:07

content that appears. That's all very

52:09

complicated. I don't know exactly how

52:11

it's going to play out, but

52:13

I understand the incentives for those

52:16

platforms to make those choices, to

52:18

say, we actually, what we want

52:20

always is more content because I

52:22

will create more attention and we

52:24

can serve more ads and we're

52:26

in this finite zero-sum attention game.

52:28

You're not in that game specifically.

52:31

You don't have those incentives. You

52:33

do allow artists to make music

52:35

with AI using your tools. Do

52:37

you allow AI generated samples to

52:39

enter your library? We do not.

52:41

We do not. Draw that distinction.

52:43

Why is it okay to make

52:46

music with AI but not to

52:48

have it in the sample library?

52:50

I think it's what users are

52:52

coming to splice for today. They

52:54

are coming to find those authentic

52:56

sounds made by humans. That's not

52:58

to say that people aren't using

53:01

AI to master. sounds or things

53:03

like that. You're using AI to

53:05

master your audio and video probably

53:07

here. I think those are tools

53:09

and that's fine as long as

53:11

there's an authentic artist artistic vision

53:13

and voice behind it. So that's

53:16

super important for us to continue

53:18

to be focused there. With respect

53:20

to these social platforms that you're

53:22

talking about, I think that's a

53:24

really important insight. And inasmuch as

53:26

these social platforms are important for

53:28

our creators as a way to

53:31

share their output, to share their

53:33

musical idea, they're really important for

53:35

us. But these social platforms have

53:37

grown because they allow people to

53:39

have emotional connection with each other.

53:41

You know, I'm really angry about

53:43

this particular issue or I'm reaching

53:46

out for support for these fires

53:48

in LA or... you know, these

53:50

connections that we make, and finding

53:52

support around this very specific, you

53:54

know, cancer that I have that

53:56

I can't find other people to

53:58

connect with online. If we

54:01

erode that, if we erode those

54:03

actual emotional connections between people in

54:05

order to save a buck and

54:07

paying out creators, I think the

54:09

value of these platforms will diminish

54:11

over time. And maybe that's okay.

54:13

Maybe we shouldn't spend so much

54:15

time on tech. Maybe we should

54:17

spend more time creating music on

54:19

our own. You know, so I

54:21

think these are really interesting evolutions

54:23

that are going to happen in

54:25

the industry. And as a mom,

54:27

you know, I care a lot

54:29

about where some of this stuff

54:31

goes. For Splice, and as a

54:33

CEO of Splice, my focus was

54:35

going to stay the same, which

54:37

is I'm focused on creators, I'm

54:39

focused on what they need, and

54:41

so many things our users create

54:43

just to hang out on their

54:45

desktop because it was just for

54:47

the joy of creating. And some

54:49

of it goes on and becomes

54:51

a, you know, top, billboard top

54:53

100 hit. Great, I'm happy that

54:55

that happens, but I'm just as

54:57

happy that someone is spending time

54:59

creating, and it just hangs out

55:01

on their desktop. Let me ask

55:03

that again in just a different

55:05

frame, because I want to push

55:07

on it. It feels important to

55:09

me. Yeah. We've talked a lot

55:11

about the active creation and what

55:13

the tools are for, and the

55:15

fact that your customers, artists, do

55:17

not want ready-made push-button songs. They

55:19

want controls. They want to add

55:21

something to what the... the computer-generated

55:23

product is giving them, right? They

55:25

want to add something to the

55:27

AI, they want to add something

55:29

to the samples, and that process

55:31

of addition creates additional value, right?

55:33

Some very important songs have been

55:35

made that way using Splice and

55:37

other tools. But you're saying that

55:39

is not a good enough argument

55:41

to get AI-generated audio into the

55:43

sample library, and I'm just wondering

55:45

why the difference? Because you could

55:47

make the same argument. Like I

55:49

use Splice to generate some samples,

55:51

I tweak them, I filter them,

55:53

I made a bunch of different

55:55

things. If it's good enough for

55:58

me to send to a major

56:00

label and play on the radio,

56:02

shouldn't it be good enough to

56:04

get into the Splice sample library?

56:06

distinction really in my mind and

56:08

I think for many of our

56:10

creators is that is it AI

56:12

generated or was AI used as

56:14

a tool to bring a human

56:16

creators idea to life? Do people

56:18

use technology to create the samples

56:20

that end up on splice? Absolutely.

56:22

People are using pro tools, people

56:24

are using synthesizers, people are using

56:26

lots of tools and technology. And

56:28

like I said, some of those

56:30

tools might be AI-based, like mastering

56:32

tools or mixing tools, things like

56:34

that. That is really different from,

56:36

I've created an algorithm to pump

56:38

out a whole bunch of samples

56:40

that are computer generated mass market.

56:42

Those are not going to end

56:44

up on Splice, I will guarantee.

56:46

But if there's an authentic user,

56:48

the Stevie Wonder of the AI

56:50

age, who is creating art, you

56:52

know, that they care deeply about

56:54

and they're using AI tools as

56:56

part of that process, absolutely. And

56:58

that distinction is very important. I

57:00

agree it's important. I just don't

57:02

know how to write it down

57:04

in a way that can be

57:06

consistently enforced across all the geographies

57:08

that you're operating in with all

57:10

of your teams going out in

57:12

the world. Or... in a way

57:14

that's understandable to artists who might

57:16

want to be part of Splice.

57:18

Is there a definition you have

57:20

of where the line is, of

57:22

how much AI is too much?

57:24

For me, again, I will take

57:26

it down to something very simple.

57:28

There's a human being who we

57:30

have a relationship with on both

57:32

sides of our platform. And so

57:34

on the side of the platform

57:36

where we are working with a

57:38

musician, artist, instrumentalist, who wants to

57:40

provide a sample. to Splice. We

57:42

actually have a relationship with them

57:44

and we talk to them about

57:46

what they're trying to do, what

57:48

the idea behind their label is,

57:50

what is their artistic... vision and

57:52

we will work with them. What

57:54

is your tool set? How are

57:56

you doing it? How many sample

57:58

packs do we need every quarter?

58:00

All of those kinds of things.

58:02

And some of those people are,

58:04

there's a Japanese potter who is

58:06

making handmade percussion instruments that he

58:08

then records that end up on

58:10

splice. That's a really cool part

58:12

of the process. And then we've

58:14

got like crazy kids. making all

58:16

kinds of super electronics, super grungy,

58:18

super sharp technical sounds, and they've

58:20

got a different tool set that

58:23

they're using as part of their

58:25

process. We're not going to tell

58:27

them, oh, you can't use this

58:29

tool because it's AI generated or

58:31

not, but do you have that

58:33

authentic vision for what you're creating?

58:35

And it's not that difficult to

58:37

tell the difference between. a person

58:39

who's creating that way and a

58:41

person who's like, I typed in

58:43

a bunch of prompts and have

58:45

got a whole, you know, plethora

58:47

of computer-generated sounds. The other extremely

58:49

challenging piece of the puzzle with

58:51

AI-generated content is when you veer

58:53

into impersonation, we've seen this in

58:55

the hip-hop industry a lot recently,

58:57

we've seen it with open AI

58:59

and Skellety Hansen's voice, there's a

59:01

lawsuit, the voice got pulled, who

59:03

knows how that's going to play

59:05

out. We see there's the Elvis

59:07

Act, I think, in Tennessee, where

59:09

impersonation is illegal and I don't

59:11

think there's a great answer for

59:13

whether Elvis impersonators themselves are now

59:15

illegal. Are you playing in that

59:17

space where you're letting people use

59:19

artist voices or sound-alikes? We're not.

59:21

I think there are lots of

59:23

people who are playing in that

59:25

space or interested in that space.

59:27

We are focused on creative people

59:29

and creative people are actually really

59:31

clear with us. They are coming

59:33

to Splice because they want to

59:35

find their authentic sound and so

59:37

we work. really hard at the

59:39

very other end of that, which

59:41

is how do we allow our

59:43

users to authentically find their own,

59:45

their own vibe? Voices is one

59:47

thing, right? They're pretty recognizable. The

59:49

Fake Drake song set the industry

59:51

ablaze. It was just very obviously

59:53

a Fake Drake song, or Drake's

59:55

voice. There's not a great legal

59:57

system for saying that's Drake's voice.

59:59

You can't use it. We'll get

1:00:01

there. It seems like we're on

1:00:03

our way to understanding how to

1:00:05

understanding how to get there. Then

1:00:07

there's... Kind of the existing mess

1:00:09

of music copyright. We talk about

1:00:11

the blurred lines case on the

1:00:13

show a lot I think more

1:00:15

than any other podcast We've talked

1:00:17

about blurred lines a song which

1:00:19

came and went and whose moment

1:00:21

is over But it continues to

1:00:23

come up on Dakota maybe once

1:00:25

a month That lawsuit was like

1:00:27

you guys stole a vibe from

1:00:29

Marvin Kaye Not notes not chords

1:00:31

anything direct, but the jury was

1:00:33

like the vibes are too close

1:00:35

Robin thickened for I'll have to

1:00:37

pay the money That's something you

1:00:39

could very easily see a user

1:00:41

of Splice wandering into. We're going

1:00:43

to prompt for a beat, we're

1:00:45

going to do a stack, we're

1:00:48

going to layer some samples, and

1:00:50

we're going to get to a

1:00:52

vibe that's too close to a

1:00:54

James Brown. Is that something you

1:00:56

worry about? Is that something you

1:00:58

try to protect users from? It

1:01:00

feels like a... conversation around reheated

1:01:02

nachos and what that means. And

1:01:04

I think artists and musicians build

1:01:06

upon each other's work. They're influenced

1:01:08

by each other. And you know,

1:01:10

this conversation's been around since the

1:01:12

beginning of sampling, right? Which is

1:01:14

how, what am I referring to

1:01:16

when I use this sample and

1:01:18

what's the story that I'm trying

1:01:20

to tell? You could argue that

1:01:22

it's derivative or you could argue

1:01:24

that it's an homage or you

1:01:26

could argue that it's building on

1:01:28

a shared piece of work that's

1:01:30

a community piece of work that

1:01:32

continues to evolve over time. I

1:01:34

think that that's what makes art.

1:01:36

and music in particular, super fascinating.

1:01:38

I love that you guys have

1:01:40

this whole debate around that particular

1:01:42

song. I think it's fascinating. I

1:01:44

think it's going to continue to

1:01:46

go. And what's right and wrong

1:01:48

should be defined by the artists.

1:01:50

But the idea that you would

1:01:52

accidentally boost too much of an

1:01:54

existing song by using an AI

1:01:56

tool, which is trained on bits

1:01:58

and pieces of existing songs. That's

1:02:00

a new danger, right? I mean,

1:02:02

the cycle you're talking about with

1:02:04

music, I agree, has existed since

1:02:06

music. We're all building on one

1:02:08

another. We're all lifting bits and

1:02:10

pieces. Great artists steal, everybody. Everybody

1:02:12

kind of understands it. And along

1:02:14

the way, there has been a

1:02:16

lot of litigation, right? That's the

1:02:18

other part of the cycle. The

1:02:20

push and pull is. people being

1:02:22

very unhappy about the money. And

1:02:24

now we're at a place where

1:02:26

it's easier than ever to be

1:02:28

derivative, and the money is absolutely

1:02:30

not clear, right? The artists are

1:02:32

very upset about their work being

1:02:34

trained on. Maybe not in your

1:02:36

tools, but certainly in other tools,

1:02:38

the labels are suing, Suno and

1:02:40

ODO, it's competitor, for training on

1:02:42

their data. Do you see that

1:02:44

resolving? Because it seems like the

1:02:46

problem is going to get worse

1:02:48

faster than the legal system, will

1:02:50

even comprehend the technology. Most of

1:02:52

these problems get worse before the

1:02:54

legal system catches up. I mean,

1:02:56

we know this. We know this

1:02:58

for privacy. We know this for

1:03:00

many areas. Technology outpaced how quickly

1:03:02

legislative action catches up. I think

1:03:04

in the music industry, we're doing

1:03:06

a lot of work to try

1:03:08

to create standards within. So we're

1:03:10

a part of a coalition of,

1:03:13

again, great companies in the music

1:03:15

space that are saying we've got

1:03:17

to... support ethical AI. We have

1:03:19

to support the rights of creators.

1:03:21

We have to make sure our

1:03:23

training data is clean. So I

1:03:25

think there are a lot of

1:03:27

companies that are trying to do

1:03:29

the right thing. Is there one

1:03:31

standard that has won out amongst

1:03:33

all the others? No, but I

1:03:35

know that a lot of people

1:03:37

are working. really hard on this

1:03:39

problem and we are too. We

1:03:41

care deeply about the rights of

1:03:43

craters. So that's going to stay

1:03:45

really important for us. How do

1:03:47

you feel about those the labels

1:03:49

suing Sun Oudio? Is that something

1:03:51

that's a warning sign for you?

1:03:53

Is that something you do support?

1:03:55

Do you think that that is

1:03:57

going to get resolved? I think

1:03:59

what the labels are trying to

1:04:01

do is support the rights of

1:04:03

the craters. Do I take one

1:04:05

side or the other? No. Ultimately,

1:04:07

it's always going to be about

1:04:09

the Quaders first. And I know

1:04:11

my customers deeply care about the

1:04:13

fact that they have rights to

1:04:15

the content they create using Splice.

1:04:17

That's why we allow people to

1:04:19

download the rights PDF. It matters

1:04:21

to people. Even if they're not

1:04:23

putting their song up on Spotify

1:04:25

or trying to make, you know,

1:04:27

a billion dollars from it, they

1:04:29

want to know that they have

1:04:31

the ability to do that. That's

1:04:33

what governs our decisions around clean

1:04:35

training data, ethical AI. If I

1:04:37

wanted to send up for a

1:04:39

splice account, download a bunch of

1:04:41

tracks, and then train my own

1:04:43

AI on them, is that allowed

1:04:45

in your license? It's not allowed.

1:04:47

And you spell that out. You

1:04:49

say you can't train AI on

1:04:51

these tracks. I bring this up,

1:04:53

and you probably don't. I'm going

1:04:55

to pre-apologize to you this question

1:04:57

because I know you haven't seen

1:04:59

this document, but just go with

1:05:01

it, is basically what I'm saying.

1:05:03

I'm asking that question. Just today,

1:05:05

I'm sure you haven't seen it,

1:05:07

but Google filed a letter with

1:05:09

the government basically saying, look, you

1:05:11

need to make an exception of

1:05:13

fair use to allow us to

1:05:15

train on everything. Open AI has

1:05:17

filed a similar letter in the

1:05:19

past few days. There's a big

1:05:21

push from the AI companies to

1:05:23

say, look, we just need the

1:05:25

stuff, give it to us. Like,

1:05:27

we don't want to pay for

1:05:29

this. It has to be fair

1:05:31

use. This is going to slow

1:05:33

us down too much. At the

1:05:35

same time, you're saying here in

1:05:38

our license, we're saying you can't

1:05:40

do that. Do you think that

1:05:42

can get resolved? That seems like

1:05:44

a big problem, right, where if

1:05:46

you steal enough of it, you

1:05:48

steal enough of it, you get

1:05:50

to. write a letter to the

1:05:52

government saying, write us an exception.

1:05:54

And if you steal a little

1:05:56

bit, you might end up in

1:05:58

court. And I don't know how

1:06:00

to resolve that. Yeah, I don't

1:06:02

either. It's such an important issue.

1:06:04

And the scale of the internet,

1:06:06

the scale of content on the

1:06:08

internet is so vast that what

1:06:10

is fair use, what is public

1:06:12

consumption, what is public record, what

1:06:14

is public ownership. We are an

1:06:16

uncharted territory and we're going to

1:06:18

be watching it just like you

1:06:20

are. How would you write a

1:06:22

fairer system if you were clean

1:06:24

sheeting this? How would you write

1:06:26

a fair system that makes creators

1:06:28

feel valued, gets them paid, and

1:06:30

still allows people to build these

1:06:32

AI systems that a lot of

1:06:34

people are getting some value out

1:06:36

of? I would love to say

1:06:38

that I'm the expert who could

1:06:40

write something like that. creative people

1:06:42

be creative and get the ideas

1:06:44

from their parts and minds out

1:06:46

there. Yeah, I'm gonna I'm gonna

1:06:48

leave that problem to people way

1:06:50

smarter than me who are legal

1:06:52

minds who are working really hard

1:06:54

on this. Well, if I get

1:06:56

anyone that show who has an

1:06:58

answer, I'll let you know. You're

1:07:00

a lawyer, right? I just talk

1:07:02

for living. I don't, I haven't,

1:07:04

I haven't done anything useful in

1:07:06

a long time. So what's next

1:07:08

for Splice is that we're going

1:07:10

to keep going deeper into the

1:07:12

creative process. I've been really public

1:07:14

about this with my blog posts

1:07:16

and all of that. Users keep

1:07:18

telling us, I love Splice, I

1:07:20

want to deeper in my creative

1:07:22

process. So whether it's these partnerships

1:07:24

that we're doing with Dawes, thinking

1:07:26

through how we build more creative

1:07:28

flexibility for users on our own

1:07:30

platform, whether it's with Create or

1:07:32

Splice Mike, there's a lot for

1:07:34

us still to do. And we'll

1:07:36

keep going down that path. All

1:07:38

right, we'll have to have you

1:07:40

back soon as some of these

1:07:42

issues play out. Thank you so

1:07:44

much for coming on. Dakota. I

1:07:46

would love to. I had such

1:07:48

an enjoyable conversation. Thank you so

1:07:50

much Neil. I'd like to thank

1:07:52

Alco for taking the time to

1:07:54

join Decoder and thank you for

1:07:56

listening. I hope you enjoyed it.

1:07:58

If you like to let us

1:08:00

know what you thought about this

1:08:03

episode or really anything else, drop

1:08:05

us a line. You can email

1:08:07

us at Decoder at the birch.com.

1:08:09

We really do read all the

1:08:11

events. You all see me directly

1:08:13

on threads or Blue Sky and

1:08:15

we have a Tik and an

1:08:17

Instagram. Check those out at at

1:08:19

at at Decoder at Decoder. At

1:08:21

Decoder. Decoder. Decoder. Decoder. Decoderpoder. They're

1:08:23

a Decoder. They're a Decoder. They're

1:08:25

a Decoder. They're a lot of

1:08:27

fun. If you like Decoder, please

1:08:29

share it with your friends and

1:08:31

subscribe or ever get your podcast.

1:08:33

If you really like the show,

1:08:35

hit us with that five-star review.

1:08:37

Decoder's Production the Virgin, part of

1:08:39

the Vox Media Podcast Network, our

1:08:41

producers, our Kate Cox, Nick Stat,

1:08:43

our editor, is Ursa Wright. The

1:08:45

Decoder Music is Red Break Master

1:08:47

cylinder. We'll see you next time.

1:08:49

Support for the show, comes from

1:08:51

Alex Partners Disruption Index, explores what

1:08:53

the best performing and fastest growing

1:08:55

companies are doing differently, as they

1:08:57

anticipate and respond to disruption. How

1:08:59

are the world's top leaders identifying

1:09:01

ways to integrate AI into their

1:09:03

businesses? Learn more on the latest

1:09:05

tech insights at disruption. Alex partners.com.

1:09:07

Disruption.aalyx partners.com. In the face of

1:09:09

disruption, businesses trust Alex Partners to

1:09:11

get straight to the point and

1:09:13

deliver results when it really matters.

1:09:16

Today at T-Mobile I'm joined

1:09:18

by a special co-anchor. What

1:09:20

up everybody? It's your boy

1:09:23

big snow deal double sheet.

1:09:25

Snoop! Where can people go

1:09:27

to find great deals? Hand

1:09:29

to t-mobile.com and get four

1:09:32

iPhone six teams with Apple

1:09:34

intelligence on us. Plus four

1:09:36

lines for online intelligence on

1:09:38

us. Plus four lines for 25

1:09:41

bucks for $25 bucks. Plus four

1:09:43

lines for 25 bucks for $25

1:09:45

bucks. later.