AI Hustle: News on Open AI, ChatGPT, Midjourney, NVIDIA, Anthropic, Open Source LLMs: AI Music Revolution with Recordless COO Roshni Lulla
Jaeden Schafer & Jamie McCauley 10/10/23 - Episode Page - 26m - PDF Transcript
Welcome to the OpenAI podcast, the podcast that opens up the world of AI in a quick and
concise manner.
Tune in daily to hear the latest news and breakthroughs in the rapidly evolving world
of artificial intelligence.
If you've been following the podcast for a while, you'll know that over the last six
months I've been working on a stealth AI startup.
Of the hundreds of projects I've covered, this is the one that I believe has the greatest
potential.
So today I'm excited to announce AIBOX.
AIBOX is a no-code AI app building platform paired with the App Store for AI that lets
you monetize your AI tools.
The platform lets you build apps by linking together AI models like chatGPT, mid-journey
and 11Labs, eventually will integrate with software like Gmail, Trello and Salesforce
so you can use AI to automate every function in your organization.
To get notified when we launch and be one of the first to build on the platform, you
can join the wait list at AIBOX.AI, the link is in the show notes.
We are currently raising a seed round of funding.
If you're an investor that is focused on disruptive tech, I'd love to tell you more
about the platform.
You can reach out to me at jaden at AIBOX.AI, I'll leave that email in the show notes.
Welcome to the AI Chat podcast.
I'm your host, Jayne Schaefer.
Today on the podcast, we have the pleasure of being joined by Roshni Lula, who is a
PhD student in brain and cognitive sciences at USC, where she investigates emotion and
decision making through functional neuroimaging and computational mythologies.
She is also the co-founder and CEO of Recordless, an innovative music streaming platform and
holds a degree in neuroscience from NYU with a diverse background in clinical, computational
and laboratory based research.
Welcome to the show today.
Thank you so much.
I'm excited to be here.
We're super excited to have you on the show.
I wanted to kick it off with a question I would be curious to ask you, which is, how
does your research in emotion and decision making in the brain kind of intersect with
your work in AI and in the music industry?
Yeah.
So I mean, to most people, it kind of feels like I'm in two separate worlds, which is
what it feels like to me as well sometime.
But honestly, there's a lot more overlap than you would think.
One of the big things just generally about being in a PhD and being in the startup world
is that there's a lot of questions and you don't necessarily know what's going to happen
tomorrow or next month or next year.
And you're kind of always innovating and trying new things.
So that's one big parallel that I've found between these two sides.
But I think one big thing that I bring over from my work in emotion and decision making
is just the insights and the data that we get from whether it's neuroscience, whether
it's surveys, whether it's just observing people and observing their behavior.
That I think is really lacking in the music industry.
And in somewhat in AI, but mostly in the music industry, the amount of data that people get
on an individual song and how their music performs aside from just being on billboard
charts or getting a certain number of stream.
So I think that's one thing that I've really been trying to bring over to the music industry
and it's a recordless.
Okay.
Very cool.
So that kind of brings me to my next question, which is, would you be able to explain to
everyone a little bit about recordless, what it is, how you guys got started and your vision
for it?
Yeah.
So it's me and my co-founder Bradley and his sister Whitney who are all working on this
product.
It's been about two years for the product as it stands today.
We have had this idea of having music be a little bit more dynamic and responsive to
whether that's data or any other type of input.
So recordless is an AI remix engineer, which means you can go in and change a song, rearrange
it, remix it without needing the technical skills or technical equipment that a DJ or
a producer needs without having a full studio and a mixed set and years of experience.
So that's kind of the basis of it.
We have a lot of plans in the future to add some kind of gamification to the system where
users almost get rewarded for making mixes and interacting with the platform and really
driving a lot of insights back to artists, like specifically independent artists who
maybe have a harder time releasing music or a harder time kind of blowing up and getting
viral without a record label behind you supporting you.
So really like that insights part is one big part of it, but also just making music
mixing a little bit more easier and accessible for more people.
Very cool.
Super interesting.
So talk to me a little bit about the platform, because as I would understand it, if it's
integrating AI essentially, so you are kind of like telling AI to change a song in a certain
way, you don't need to have like maybe like the stripped down vocals or like the different
actual files in the song that AI is able to kind of work on it.
Is that how it functions?
Yeah, so the way our platform works currently is we have a desktop app, which we will be
releasing soon actually for download, which is super exciting.
But we have songs that are already in there.
You can also load in your own music, so music you own and like have the rights to and it
takes that song and kind of chops it up into the different sections.
So the different verses, the chorus, intro, outro, and you can rearrange the song to maybe
play the outro before the intro and then go to verse two, then verse one and play it
backwards or something.
So for example, we had a Drake song that Bradley and I love and we took his verse and
played it in reverse.
So like the end of the verse, then the middle, then the beginning.
And the way the AI works and the way the platform works is that when you play your new
rearrangement or your new arrangement and mix of the song, it sounds like it was supposed
to be listened to that way.
So it's really cool if someone is like unfamiliar with the song and we played them a new arrangement.
They think that's how the song was supposed to sound initially, even if we created the mix
like in our platform.
So that's kind of a really exciting part.
And then we also have another type of remix thing, which is a little bit more like what
you were saying, where it splits it into a song stem and either you can tone down certain
stems, like maybe make the drums a little bit quieter, add in more complimentary drum
and create like different vibes or moods of the song.
So that's like a secondary part.
But the core part of the song and like the actual creativity of the artist is maintained
in both aspects of the mixing.
Very cool.
This is like so fascinating for me because little too many people know.
But I at one point when I was probably like 14 wanted to be a DJ, I thought that was like
the ultimate pinnacle of my career.
And I just sadly probably wasn't very good at it, or maybe I didn't have enough practice.
I don't know, even to the point where this is like, this is so funny.
I feel like I'm like doxing myself in some way.
But there's this thing called a ham radio license, super nerdy thing.
It's like these like long frequency radios.
When I was a kid, I got ham radio certified and they let you pick your like, your like
call sign and my call sign was the a seven DJ J.
Because I was like, yeah, it's like DJ, it's pretty funny.
Very, very nerdy take on it.
But I think this is a really applicable take or a space right now because we're seeing
a lot of really interesting moves with AI and music in general.
There's a lot of big record labels and producers that are actually starting to, you know,
it was kind of interesting because originally, I would say a number of months ago or even
at the beginning of the year, people were starting to make like, not like deep fakes,
right, but like they're making like verge, like taking Drake's voice and like making
a new AI song, you know, remix with the weekend, that thing went viral.
And so there's this whole space.
And originally, I kind of got like the vibe from a lot of people in the music industry
that they were like, no, shut it down, stop it.
You know, they were like threatening to sue people that were doing this pretty much.
Right. Like Drake didn't want knock off Drake songs.
I feel like the mood has shifted significantly in the last probably one to two months.
I think we've seen a bunch of like announcements out of, I think, like universal records.
It's some other places where they're saying like, no, we're going to start like figuring
out what the process is with AI music.
And essentially, you know, licenses will be given back to the original authors.
So they're like, you know, they're the work starting getting stolen and whatnot.
Of course, I think probably one of the biggest pioneers in the space was Grimes
when she kind of came out and said, anyone can use my voice and make AI song.
And then you just give me like part of the royalties for it or whatever.
So I think we're getting to a point where the big the big people
that would be fighting you, for example, you know, universal music and Warner
and all those guys are like getting to the point where they realize it's inevitable.
People are going to make bootleg versions, whether they like it or not.
And so at this point, it's kind of like, OK, well, how do we regulate and make sure that,
you know, our original artists that we have signed are going to get some sort of benefit.
I love that you guys are focusing on the independent artist angle, too.
I think that's really cool.
So you said you said you guys are planning on in the next month
launching something for download.
Talk to me a little bit about some of the things you guys are excited for,
looking forward to and kind of next steps for you and your company.
Yeah, so our ultimate goal is to have a streaming platform.
So a streaming platform where similar to Spotify or Apple music, where you subscribe to it.
But it's really a more interactive and dynamic process of listening to music.
So you have your library of music, but you can play it back in a way
that ultimately matches your mood and like your condition, which again,
like kind of tied a little bit more back into my PhD work and like neuroscience
and behavior, because music that matches your mood is like obviously preferred.
And it's like hard, you know, like if you're on a run, for example,
and like you're really into it, then a really slow like Taylor Swift song comes on.
It kind of kills the vibe a little bit.
So our ultimate goal is for it to either automatically detect your mood
or you can input your mood and like say a couple of things,
like imagine like a chat dbt, like I'm about to go on a run for an hour
and I'm really trying to hit this goal and it'll create
not necessarily just a playlist, but create renditions of the songs
in your library that match that.
So that's kind of like our ultimate goal and obviously integrating AI a little bit more.
And also just what you were saying, right?
Like you experimented with being a DJ.
I feel like a lot of people that I've talked to are like,
yo, I thought about being a DJ like I for when I was a kid
and I would like open up a garage band, like play around with little songs,
like make little mixes and like they sounded so bad, like, oh, my God.
But like if we could make this almost like democratizing the art of DJing,
I think that's another like big goal that we see moving forward.
This is like the funnest AI company I've talked to in a while.
It's just like so fun and I think consumers are going to love this.
So it's actually you guys essentially are doing the vision of this.
I was listening to a podcast that's called My First Million.
And probably four months ago,
they were talking about pretty much what you guys are doing, the concept.
They're just brainstorming ideas and they're like, yeah, it's going to be so awesome.
At some point, you're going to pop up your Spotify or whatever
and you're going to like go find all your favorite artists.
You're going to be able to like make music and mixes off of like them
and their songs and their lyrics, but you change it to what you want
and like your moods and like pretty much build these like custom songs.
And something that was interesting they were talking about was the fact that they're like,
yeah, and wouldn't it be cool if like you made like some sort of variation
of like a Drake song remix thing, right?
But it's like your own new thing.
And then you put it out there on the platform or whatever
and you get some sort of royalties or some sort of like following
because you do this kind of thing.
Anyway, so it's kind of talk to me about like, what's your guys vision in that regard?
I know you said you're thinking about some gamification things in the future.
But like, what does that look like? Is that challenging?
Is it even viable or possible to do like royalties?
I like create a Drake song or is Drake not going to let that happen?
Yeah. So one of the things that then I'm really excited about
that we've been thinking about is remix contest.
So what we have right now is, again, rearranging a song,
taking different parts and placing them in different areas, right?
And when people are creating a song or a single for the first time,
they go into the studio.
A lot of the times that we've been talking to producers,
they come out with like five to ten minutes of just content.
And then they have to go back and like mix and master it to be like,
we want this verse to be next to this part and like this chorus to come here.
And like basically just rearranging the song and like this
longer mix that they have into their final single.
So one of the things because my background is in like data science
and analytics and we would do this a lot.
I worked at a company when I actually first met Bradley.
That was focused on consumer insights and advertising.
So we would literally show people ads, measure biometrics and tell
whatever company like Walmart, Netflix, whoever it was and be like, Hey, listen,
like people really lose attention in this part of your ad.
Like this, these 10 seconds, people are losing it.
So make those 10 seconds and people will be more engaged throughout the ad.
Remember your brand more like more brand purchase, like loyalty, all those different things.
And I was like, what if we bring a concept kind of similar to that into music
where it's not necessarily how is your overall song doing?
But how are people reacting to the little parts of your song?
So for remix contests, we imagine, say you're an independent artist,
you have this like five to ten minutes of content.
You release it to a smaller version of your fan base, like whoever your core fans
are, say there's 200 of them, they can come in and they submit their own arrangement
of your song. So each fan goes in, into record list, creates an arrangement
of the song saying like, I like when this verse is next to this verse
and take out this part, like have this intro in and submit that.
And the artists can go through and see which types of arrangements
and which like transitions are most commonly used, which parts of their song
are most commonly taken out or which verses are most commonly like doubled up.
So when they're going to create their final mix, they can do it based
on what they know their fans already want to hear.
And then rewarding that fan, right?
First of all, I mean, I think if an artist even asked me, like an artist
that I really like asked me like, yo, like help me create my next mix,
I would be like so stoked to be like, yo, I can tell them create their next single.
Whether I get rewarded or identified for that or not, like just that experience,
I think is really cool.
But ultimately, like it could be as simple as a shot on an Instagram, right?
Like, yo, this person helped me finalize my mix or it could go as far
as getting backstage passes and meeting your fan or like meeting an artist.
And just because you created that final arrangement, you get to meet the artist
or you get writing credits on their actual song.
All of those things are ways of rewarding the fan for their behaviors, basically.
But also kind of creating this like sense of collaboration and co-creation
that I think is really like missing in the music industry right now.
Totally, I love that.
Yeah, I think that's super cool.
And I also feel like it creates like a totally like a community
that feels like a club where it's like, oh, yeah, we're like super fans of this artist.
We're like literally mixed out there chatting with them.
People will love that, right?
If they're going to get feedback and talk to their people.
And I think another thing I'm sure maybe you guys are already thinking about this,
but just off the top of my head, that would be super cool is like when you have
all of those remixes of the song away for the users on the thing
to like rank or upvote or downvote ones.
That's another really easy way for the artist to be like, oh, my gosh,
90 percent of people love this mix.
Let's go with it.
So super, super cool.
What are what are some of the biggest struggles you guys have faced
in kind of rolling out this vision and working on this right now?
Yeah, so I'm obviously a very like technical person,
like very sciency and like into brains and stuff.
And Bradley's an artist.
He's like spent years like mixing and producing music, but he's ultimately
also a software engineer.
His sister is also a doctor.
So I have a medical degree.
I spent years and hours in the OR, like literally rebuilding the human body.
So all of us are very like technical, like in our heads, people.
And I think one of the biggest difficulties was like talking to artists, right?
Who maybe don't have that technical background.
And once we start saying things like artificial intelligence
and like data insights and stuff, like sometimes people get scared
and they like pull back a little bit because they're like,
I don't understand this world as well as you guys do.
And I think that was one thing that I mean, it's gotten better
through like a lot of conversations.
We've partnered with people who are who have spent years in the music industry,
who have spent years as a producer and work with artists all the time.
So I think that was an initial like struggle was that we were very
and obviously when you work with the company, you're so involved with the product
that like your head is so into the product and like it's kind of hard
to take a step back and see the bigger picture sometimes
because you're like I've been working on this product for like two years straight,
you know, so I think that was like a bit of a difficulty in the beginning
was just like trying to understand like, yeah, this is a complicated concept.
Like it's very technical.
So how do we step away from the technicality of it and be more marketing almost
like being more like, yo, like this is how it's going to look.
You know what I mean?
Like talking to an artist and like speaking their language almost.
So we started like trying to use words that they were familiar with.
Like we were like, imagine it's like a live listening session.
So like when you're in the studio with your crew and like listening to a song
and like you have a listening party or like a listening session before a drop,
like imagine doing that with your super fan and like getting their feedback.
And like, you know, so that was a way of almost speaking their language
that helped break that barrier a little bit.
Very cool.
That's super cool.
One one thing I'd love to pick your brain on and ask you about is like,
how does like how does the this whole like studio
A.I. assist feature work and what kind of like algorithms are behind it?
How how does that?
How do you guys put that together?
Yeah.
So as I said, the first the version that we're releasing right now
is really this rearrangement and what it's been what's been done.
I mean, obviously Bradley is like the brains behind the tech and stuff.
So I can't speak too too much about like the details of the algorithm.
But basically it's using A.I.
It's training on a bunch of other songs.
And when you load a song in, it's detecting what the different parts of the songs are.
Like when does the verse one start and end?
When is the intro start and end?
Obviously intro is a little bit easier, but like verses or like the chorus
and like really identifying those splits between the song.
But also creating efficient transitions across all those sections.
So you can imagine if there's 10 sections of the song,
each section should have a transition to every other section.
So that's a lot of transitions to make.
Yeah. And to make sure all those transitions sound good enough
that when you create your final make, it sounds like a normal song.
So that's what we've been using A.I. for.
And we're training it even more to learn from the typical transitions
that people make within a song.
And then ultimately recommend transitions to fans when they're creating a new mix.
We can say like, hey, like based on previous data
and based on like previous mixes that have been made,
there's a transition that might sound cool or like this is the most common
transition that's been used and kind of learning from the data
that we collect from other fans and users of the of the platform.
But currently it's like that.
And then for our like mood version of mixing,
it's like figuring out what the stems are within a song
and like automatically detecting those stems and then recommending
like maybe you tone the vocal down, vocals down a little bit here
or like add in these complimentary drums.
So like doing that in a more automated way.
Very cool. Man, that's super interesting.
Something I'd be interested in asking you about is from your kind of like
vantage point right now, you guys have been working on this for a couple of years
and kind of looking at this space.
Where do you see like music and AI kind of evolving
if we're looking into like the next five years, right?
Like, I mean, your your platform and everything you're doing is one thing.
And I wonder if there's any external industry ships
you're going to start seeing other people make as well.
Like, where do you kind of see the future of music at this point?
Yeah, I mean, I think one, I don't think like the automatically generated
music is going to stop at any point, like using other people's vocals
and like creating completely new Drake songs or completely new
like songs with grinds voice in it.
Like, I think that's only going to continue and get even better,
honestly, in the next couple of years.
I'm hoping we see more of this like collaboration co-creation happening
because by using AI and like these advanced tools,
like the ability to create new music and create new mixes is just going to get easier.
So I'm hoping there should be more of like a collaboration
where you feel less far from your artist, right?
Like right now it's like the artist has so many talents,
which I don't think an artist's creativity is going to go away at any point.
Like as a span, I don't think I will be ever as like creative and talented
as like Drake is, right?
But I'm hoping like you feel less like distant from them.
So I'm hoping that that is kind of where we see like music industry shifting a little bit.
And there's less of like the artist has so much like power in in like the music industry.
And there's like a little bit more of like the fans have some say as well.
So I think that's kind of where I'm like hoping to see that the music industry
like move into with the rise of AI and like new technology
that kind of just makes the music making process and like feedback process a little bit easier.
Super cool. Oh my gosh, this is you guys are working on some incredible things.
I absolutely love it.
It's super fascinating.
I'm excited to hear about your launch.
If people listening today to the show want to try it out
and when you launch or get on the wait list or whatever, what's the best way for them to do that?
Yeah, so about like two weeks ago, we just released a new version of our website,
which I'm really excited about.
And we actually have the rearrangement technology available for a demo on the website itself.
It's recordlist.app.
That's our website.
And there you can also sign up for the wait list.
But the coolest thing is that we have a song up there
where you can you see the original version of the song like intro, verse,
verse, chorus, bridge, whatever.
And you can drag and drop the sections to a new mix of the song and preview your own mix and you can hear it.
And if you're not familiar with the song, it literally sounds like a normal song,
but like it's supposed to sound like that, which is kind of our whole goal.
So that's really exciting.
That's already on the website for anyone to try.
And then if you get on the wait list,
hopefully we'll be able to send you a version of the actual desktop app to download.
And then you can play with your own music, like say you have some mixes that you're trying to make by yourself
that you own and you have on your on your computer.
You can load those in and kind of rearrange them without having to have, again,
like the technical skills that are needed for a DJ or like going to GarageBand.
GarageBand itself is super complicated to you sometimes.
So this just makes it a little bit easier and like gives everyone the ability to like be a little bit of a DJ.
Super cool.
Well, this is absolutely fascinating.
I'm super excited to the listeners.
I'll leave a link in the show notes to go try record lists, recordlist.app.
So you can go and do that.
And Roshni, if people want to get in contact with you or the team to collaborate
or find out more about what you guys are doing, what's a good way for them to reach out?
Yeah, so my email is just Roshni at recordlist.app.
You can feel free to email me.
Happy to talk about science literally anytime.
Happy to talk about a record list anytime.
So, yeah, amazing.
Well, thank you so much for coming on the show today.
I really appreciate your insights and everything you guys are working on.
Absolutely fascinating stuff to the listeners.
Thank you so much for tuning in to the AI chat podcast.
Make sure to rate us wherever you get your podcasts and have an amazing rest of your day.
If you are looking for an innovative and creative community of people using chatGPT,
you need to join our chatGPT creators community.
I'll drop a link in the description to this podcast.
We'd love to see you there where we share tips and tricks of what is working in chatGPT.
It's a lot easier than a podcast as you can see screenshots,
you can share and comment on things that are currently working.
So if this sounds interesting to you, check out the link in the comment.
We'd love to have you in the community.
Thanks for joining me on the OpenAI podcast.
It would mean the world to me if you would rate this podcast
wherever you listen to your podcasts and I'll see you tomorrow.
Machine-generated transcript that may contain inaccuracies.
Embark on a sonic journey into the future of music with Roshni Lulla, COO of Recordless, in this captivating episode. Join us as we explore the transformative role of AI in the music industry and how Recordless is leading the way. Discover the cutting-edge innovations that are reshaping the way we create, experience, and enjoy music!
Get on the AI Box Waitlist: https://AIBox.ai/
Join our ChatGPT Community: https://www.facebook.com/groups/739308654562189/
Follow me on Twitter: https://twitter.com/jaeden_ai