🎵 Can You Tell If Your New Favourite Music Artist Is A Real Person Or AI?

Em Gillespie joins Jonesy & Amanda with the latest in entertainment.

See omnystudio.com/listener for privacy information.

Episode Transcript

Speaker 1 (00:01):
And Amanda gem Nation the Daily o. Emma Gillespie is here.
AI music, Well.

Speaker 2 (00:08):
I have one of the great philosophical questions of our time.
Are you listening to a real artist or a robot?

Speaker 1 (00:15):
Music?

Speaker 2 (00:16):
AI, of course, is a huge topic of conversation at
the moment. There's a big study that's come out, a
global survey from around the world ten thousand people who
were asked if they could correctly identify whether a track
was fully AI generated or fully human generated. So I
thought we'd play little game. I'm gonna play you two songs.

(00:38):
I want you to tell me which one you think
is AI. So one of them is an AI track,
one of them is a real human person track. So
you've got option A, which is this one?

Speaker 3 (00:50):
If you don't like.

Speaker 1 (00:52):
I'm a key walk, I like it.

Speaker 3 (00:56):
Change changer man changing my song?

Speaker 1 (01:03):
I like it. Yeah, it's pretty good. Right, that's a
but is it aile real? No comment? Here's b see
I happened. I went to see this man in concert,
so I know. Sorry to record for you. This one

(01:25):
is Chris Stapleton. Well, I was going to say that
the first one was AI. I thought the first one
was AI. And that's Chris Stableton.

Speaker 2 (01:33):
Well, congratulations, you were in the top three percent of
the populace.

Speaker 1 (01:37):
Probably because we knew the christ. We do this for
a job. I've got to listen.

Speaker 3 (01:42):
But if you didn't ask us to be quizzed on that,
I wouldn't have known that that first song, you could easily.

Speaker 1 (01:47):
Have just listened to the first ever thought anything? And
so is that who's put that together? So that is
an AI song.

Speaker 2 (01:52):
It's called Walk My Walk by an artist called Breaking Rust.
It was number one on the Billboard Country Charts last week.
Really lately AI generated Oh, this artist Breaking Rast, who
doesn't exist, has over two point three million monthly listeners
on Spotify. That song that we heard that little clip
of has three point six million streams to date.

Speaker 1 (02:16):
Is it a secret that he's AI. Well, that's the question.

Speaker 2 (02:20):
I guess it's not obvious on Spotify anywhere. You have
to go into the credits of the song to find
the names of the songwriters and then if you look
those people up, you can quickly realize that they're not real.
But this is part of the problem.

Speaker 1 (02:34):
At the moment.

Speaker 2 (02:34):
There was this Reuter's Global survey of nine thousand people
from eight countries ninety seven percent couldn't correctly identify an
AI track, but seventy five percent said they want clear
labels on AI generated music so that they can see
those songs they can't identify as being AI, so they
have a quick and easy way of recognizing and acknowledging that. Okay, cool,

(02:55):
we like the way that sounded, but it wasn't made
by a person.

Speaker 1 (02:58):
Do people care?

Speaker 2 (03:00):
People do care, They care greatly. Sixty five percent think
that AI models shouldn't be able to be trained using
copyrighted music. Seventy percent believe that AI threatens the livelihood
of musicians, and over half said they feel uncomfortable not
knowing if they're listening to a machine generated piece of content.

(03:20):
So I guess there are these huge ethical concerns around
it that are ongoing, the legal concerns around copyrighted music.

Speaker 1 (03:27):
Being used to heay to train it.

Speaker 2 (03:29):
Well. Exactly Universal did a massive deal with an AI
song generator just a couple of months ago, and at
Universal Artists an'ch or what that means for them? Does
that mean that their songs are kind of being offered
up to the AI song generator gods and it's out
of their control. So there's a huge lack of trust
and transparency clearly over this issue.

Speaker 3 (03:49):
So is this like say you're a musician and you
signed a universal and they take your song to train AI. Yeah,
it's like painting by numbers. It gets broken down and
then re constant ted it somewhere.

Speaker 1 (04:00):
Yeah.

Speaker 2 (04:01):
And if you say John Farnam, who we just heard from,
if you put into a song generator, I want a
song in the style of John Farnham with a voice
that sounds like his singing about the Sydney Harbor Bridge
or whatever. You know that that song could be generated
and sound just like him. But the BBC I love
This released a new article with a bit of a
kind of how to train yourself to detect AI music

(04:22):
with a few good tips. They recommend listening for high
frequency fuzziness. So some AI songs the vocals or the
synths have this unnatural kind of fuzzy sound around them
or high frequencies. They say to always check the artist's profile,
So if you're skeptical of a song, look up the
name of the artists, look at it.

Speaker 1 (04:40):
Don't They have fake profiles too.

Speaker 2 (04:42):
Yes, but you can tell if you look close enough
at those images. Of the AI generated artists. They look
a bit cartoon they're a little bit cartoonish. You might
not think anything if you just scrolled past it quickly,
but if you look at it for longer than five
or ten seconds, you can usually tell okay, that doesn't
look like a real person. So look up the art artist,
Google them. You can use filters on some streaming platforms

(05:04):
to exclude AI content, so there are calls for those
filters to be more easy to use. And honestly, this
article has encouraged people to listen to a lot of
AI music because it thinks that that will help you
notice more of those subtle patterns.

Speaker 1 (05:19):
And then you can discern what's real and what's not.
It's like AI radio show, who would do that? Oh
do that? He's a naughty boy, am thank you, thank you.
It's all about authentasty man.