Field Dispatch
Bandcamp vs. AI | Matt Connarton Unleashed
Speaker 1: I did as promised. I wanted to talk about some news.
Speaker 1: You know, we mentioned band Camp on the show earlier.
Speaker 2: Band Camp.
Speaker 1: Two things that come up a lot on the show
Speaker 1: band Camp the website of course, bandcamp dot com, and
Speaker 1: AI and how AI is affecting how music is made, generated,
Speaker 1: consumed in the music industry. And band Camp has now
Speaker 1: banned AI generated music. They have made it official saying
Speaker 1: that they want musicians to keep making music. And you know,
Speaker 1: we've got kind of a point counterpoint on this to
Speaker 1: look at. But first, this is from Music Businessworldwide dot com.
Speaker 1: And this just happened, by the way. They just announced
Speaker 1: this ban a couple days ago. According to this website
Speaker 1: direct tofan platform, band Camp said it will ban artificial
Speaker 1: intelligence generated music and audio from its platform, including any
Speaker 1: use of AI tools to impersonate other artists or styles.
Speaker 1: The company made the announcement in a post on Wednesday,
Speaker 1: January fourteen.
Speaker 2: So this just happened. Like I said a few days ago.
Speaker 1: Quote, something that always strikes us as we put together
Speaker 1: a round up like this is the sheer quantity of
Speaker 1: human creativity and passion that artists express on band Camp
Speaker 1: every single day. The fact that band Camp is home
Speaker 1: to such a vibrant community of real people making incredible
Speaker 1: music is something we want to protect and maintain unquote.
Speaker 1: And by the way, and this came up during our
Speaker 1: conversation with Adam steglitic have ever felt in the second hour,
Speaker 1: I think what we talked about with Adam, it was
Speaker 1: either with Adam or with Grimrock.
Speaker 2: No, I think it was with Adam. We were talking
Speaker 2: about how band Camp.
Speaker 1: One of the reasons that I'm such a proponent of
Speaker 1: it is if you get your music files from band Camp,
Speaker 1: You're getting a high quality music file, much higher quality
Speaker 1: than say if you're just streaming it on YouTube or
Speaker 1: something like that. So just something to be aware of too.
Speaker 1: Band Camp is really good that way, it says here again.
Speaker 1: According to this article, band Camp said music and audio
Speaker 1: that has created quote holly or in substantial part by
Speaker 1: AI unquote will no longer be allowed on the platform.
Speaker 1: The company will also enforce existing policies against any use
Speaker 1: of AI tools to copy other artists or styles. The
Speaker 1: company said, quote, we want musicians to keep making music
Speaker 1: and for fans have confidence that the music they find
Speaker 1: on band Camp was created by humans unquote. This has
Speaker 1: come up several times. Like I said, over the past
Speaker 1: couple of years. You know, Jenny and I talk about
Speaker 1: this all the time on the show, the effect of
Speaker 1: AI in the music industry and in the creation of music.
Speaker 2: And there are some tools like.
Speaker 1: Pseudo dot com, which we've experimented with live on the show,
Speaker 1: that they can make some really realistic sounding, you know,
Speaker 1: music that sounds like it was made by humans, entirely
Speaker 1: by humans. One thing that look, it's any platform has
Speaker 1: the right, whether it's band Camp, whether it's YouTube, whomever.
Speaker 1: As frustrating as it can be at times, any platform
Speaker 1: has a right to police and control what the content is,
Speaker 1: in this case music, when we talk about band Camp,
Speaker 1: what the content is it is being uploaded to their platform. However,
Speaker 1: I really worry with policies like this that because it's
Speaker 1: going to happen, it's inevitable. I guess it's an extent
Speaker 1: it's impossible to prevent this. So it comes down to
Speaker 1: an extent to which it ends up happening. Artists are
Speaker 1: going to get caught in the net. And I don't
Speaker 1: mean net as an internet, I mean net as in
Speaker 1: like a fishing net. Artists are going to get caught
Speaker 1: in the net where whatever software band Camp is using
Speaker 1: to detect AI generated music, there are going to be
Speaker 1: artists who are flagged by band Camp as being AI generated,
Speaker 1: who in fact are not AI generated.
Speaker 2: It's going to happen. It's unavoidable. Okay.
Speaker 1: So I'm not saying that because that's going to happen,
Speaker 1: that band Camp should not do this. I'm not saying that.
Speaker 1: But what I am expressing is a concern that what
Speaker 1: if it happens a lot. Because anybody who's ever created
Speaker 1: anything that you've then put on the Internet for mass consumption,
Speaker 1: whether it's music, whether it's a podcast, doing a radio
Speaker 1: show that you know, in addition to being live on
Speaker 1: FM radio also streams online. Anybody who's ever created anything
Speaker 1: knows that sometimes you can get flagged, you can get tagged,
Speaker 1: you can get dinged for something by these platforms that
Speaker 1: is in fact, uh not correct, and you're getting blamed
Speaker 1: for something that you didn't necessarily do you know. Look,
Speaker 1: I've been on YouTube for a very long time. You know,
Speaker 1: it hasn't happened in a long time, but I have
Speaker 1: in the past. More so when I was doing a
Speaker 1: lot more political content, I would get I would get
Speaker 1: dinged by YouTube for things. So you can't say this,
Speaker 1: you can't talk about that. You've got to strike on
Speaker 1: your account now because you said this, you repeated this,
Speaker 1: and this is not accurate. You can't do that, you know.
Speaker 1: And most of the time, not to sound like I'm
Speaker 1: shirking responsibility, but most of the time it's like, wait
Speaker 1: a minute, that's not what I said. They're flagging me
Speaker 1: for I was quoting something somebody else said that wasn't
Speaker 1: my opinion. Why are they Why am I being penalized
Speaker 1: for repeating what somebody else said and making it clear?
Speaker 1: But see, you know what happens is right there is
Speaker 1: these the bots. They it's all it's AI. See, that's
Speaker 1: the thing. They use AI to detect AI. That's what
Speaker 1: band camp is going to be doing here. They're going
Speaker 1: to use AI to detect when there's AI and AI.
Speaker 2: The AI that.
Speaker 1: They're using to detect whether or not they there's AI
Speaker 1: is not going to get it right every time, especially
Speaker 1: when because this is a little broad when they say
Speaker 1: band camp, okay, band Camp said music and audio that
Speaker 1: is created holly or in substantial part by AI. Listen
Speaker 1: any any recording studio you know that that is modern
Speaker 1: is probably you know, probably has software in their computers
Speaker 1: in that recording.
Speaker 2: Studio that use AI to fix things, to tweak things.
Speaker 1: I mean, it's so ingrained in everything, and it really
Speaker 1: has been with us for you know, some people act like, oh,
Speaker 1: AI it's this new thing.
Speaker 2: It's not.
Speaker 1: It's just becoming a lot more prevalent with access to
Speaker 1: tools like chat ebt and things like this. But AI,
Speaker 1: there's always been an element of AI, not always literally,
Speaker 1: but for a very long time in everything we do.
Speaker 1: I mean, look, even twenty years ago, you could go online,
Speaker 1: for example, and you could interact with chatbots. Maybe there's
Speaker 1: a particular service that you are having a problem with.
Speaker 1: Maybe you made a purchase through a website, for example,
Speaker 1: and your item never arrived. There's something wrong with it.
Speaker 1: So you go to the website and you try to
Speaker 1: contact someone to get some help with your purchase, and
Speaker 1: you end up talking to someone through a chat box. Right,
Speaker 1: You're talking to someone, but the person you're talking to
Speaker 1: doesn't seem to be able to help you very much,
Speaker 1: or they seem to be really easily confused by what
Speaker 1: you're trying to communicate to them. Why, Because it's not
Speaker 1: an actual person. It's a chatbot, it's a bot. It's
Speaker 1: AI that's been with us for a very long so
Speaker 1: it's not new. It's much more prevalent, especially over the
Speaker 1: last couple of years. It's much more prevalent than it
Speaker 1: was and will continue to be. It seems to be
Speaker 1: accelerating and unavoidable, but it's not perfect, and it makes mistakes.
Speaker 1: It's almost makes it kind of human in that sense.
Speaker 1: Right now, as it becomes more sophisticated, I presume it
Speaker 1: will make fewer and fewer mistakes, but it still makes mistakes.
Speaker 1: So in other words, again, I say all that to say,
Speaker 1: people are going to get caught up in this net.
Speaker 1: When I see these platforms saying well, we're going to
Speaker 1: ban any use of AI in anything that is uploaded
Speaker 1: to our platform, It's like, but how are you going
Speaker 1: to accurately detect it? And I know some will say, no,
Speaker 1: there's nothing to worry about here. You'd be surprised. I've
Speaker 1: had people tell me that, Matt, you'd be very surprised
Speaker 1: at how accurate AI can detect when there's AI being used.
Speaker 1: I'm a little bit skeptical because again I'm someone who
Speaker 1: has seen what happens when AI tells me, for example,
Speaker 1: that I've done something wrong on YouTube, when I didn't
Speaker 1: even do the thing that I'm being accused of by
Speaker 1: this bot, and good luck trying to talk to a
Speaker 1: real person, you know.
Speaker 2: Don't get me wrong.
Speaker 1: There's been a couple of instances where I've appealed strikes
Speaker 1: on my channel and one. But I've also lost some
Speaker 1: too that I shouldn't have lost. But again, it's it
Speaker 1: is AI. It's not a human being I'm talking to.
Speaker 1: So so that's my concern about this.
Speaker 2: Again.
Speaker 1: Ultimately, they can do whatever they want, it's their platform,
Speaker 1: but I worry about people getting caught up in the
Speaker 1: net who shouldn't be. It also says here again continuing
Speaker 1: with this article, This is from Music Businessworldwide dot com.
Speaker 1: It says band Camp said the new policy was driven
Speaker 1: by quote the sheer quantity of human creativity and passion
Speaker 1: that artists express on band Camp every single day. Users
Speaker 1: will be able to report suspected AI generated content through
Speaker 1: band Camp's reporting tools. The company says it reserves the
Speaker 1: right to remove music based on suspicion of AI generation.
Speaker 1: The company said it will update the policy as AI
Speaker 1: technology continues to develop. The announcement comes over four months
Speaker 1: since band camp launched a new thirteen dollars a month
Speaker 1: subscription service that is human curated. The service gives users
Speaker 1: access to monthly recorded selections, listening parties, recommendations, and exclusive
Speaker 1: artists content. Band Camp general manager Dan Melnick at the
Speaker 1: time said, quote, instead of algorithms, fans get human curated
Speaker 1: picks from some of the best DJs and journalists in
Speaker 1: their respective areas, exclusive interviews with artists, and community listening parties.
Speaker 1: Vibrant music scenes are made up of dja's journalists, fans,
Speaker 1: and artists, and band camp clubs reflect that as always,
Speaker 1: artists are paid fairly and at the heart of the experience.
Speaker 2: Unquote all right, well, I like that. I like that.
Speaker 1: Founded in two thousand and eight, band camp launched as
Speaker 1: a direct to consumer alternative to record labels for artists.
Speaker 1: It has since expanded its services to include features such
Speaker 1: as ticket live streaming, and vinyl pressing. Artists like Peter
Speaker 1: Gabriel and b York have placed their catalogs on the platform.
Speaker 1: All Right, I don't want to get into the whole
Speaker 1: history of band camp, but uh okay, and then it
Speaker 1: gets into some more details about band Camp. Oh, I'm
Speaker 1: skipping down though, because Spotify is mentioned here and I
Speaker 1: remember we talked about this on the show too, and
Speaker 1: I expressed the same concern when talking about Spotify removing
Speaker 1: AI generated music. Some artists are going to get caught
Speaker 1: in the net, it says here. Spotify in September said
Speaker 1: it removed more than seventy five million spammy tracks from
Speaker 1: its platform over the past year. Amid the explosion of
Speaker 1: generative AI tools, the streaming giant rolled out new policies
Speaker 1: for managing AIG generated content on its service to address that.
Speaker 1: In November, French streaming platform Deezer said it receives over
Speaker 1: fifty thousand fully AI generated tracks daily, or thirty four
Speaker 1: percent of all tracks uploaded to its platform each day.
Speaker 2: Oh my god, that's nuts. That is wild. Now. So
Speaker 2: that's band Camp's position. Now someone has.
Speaker 1: Offered a I don't know if rebuke is quite the
Speaker 1: right word, but there's an artist because you would think
Speaker 1: you would think recording artists would be in favor of
Speaker 1: what band camp is doing, and probably most are. But
Speaker 1: Holly Herndon says band Camp's AI ban is misguided. This
Speaker 1: comes from stereogum dot com. This went up just yesterday.
Speaker 1: Earlier this week, band Camp issued a blanket ban on
Speaker 1: generative AI and deep fake impersonations. The new measure makes
Speaker 1: band Camp an exception on the streaming landscape where AI
Speaker 1: music is becoming more and more inescapable. Most of the
Speaker 1: reactions to this bandcam announcement were positive, but one artist
Speaker 1: who's not applauding it is Holly Herndon, an experimental artist
Speaker 1: who has long messed around with what she sees as
Speaker 1: the artistic possibilities of AI technology. Holly Herndon has a
Speaker 1: PhD from Stanford's Center for Computer Research in Music and Acoustics,
Speaker 1: and she has made her most recent album, twenty nineteen's
Speaker 1: Proto with Spawn and AI Baby, curated by Herndon and
Speaker 1: her partner Matt Dryhurst. A few years ago, Herndon and
Speaker 1: Dryhurst launched Spawning, a tool that allows artists to choose
Speaker 1: whether or not their work will be used to train AI,
Speaker 1: and that work earned them Austria's first Digital Human Rights Award. Interesting,
Speaker 1: I didn't know that was the thing. On Wednesday, in
Speaker 1: response to Bandcamp's new policy, Herndon tweeted this, putting the
Speaker 1: ban in band Camp. It is a wicked problem, so
Speaker 1: I understand what they are trying to do. I'm sure
Speaker 1: the human moderators won't be banning artists exploring AI. Rather,
Speaker 1: I hope, will just be looking to filter out sound spam.
Speaker 1: The gray area will only get larger m HM. After
Speaker 1: a negative response to that tweet, Herndon elaborated with a thread.
Speaker 1: She said this, I understand why band camp is taking
Speaker 1: this measure, but it's a tourniquet. The human AI binary
Speaker 1: is not going to hold and will become a matter
Speaker 1: of superficial optics. I already have more authorship in my
Speaker 1: models than most pop stars do in their songs. People
Speaker 1: will already be using models to generate songs they then
Speaker 1: put a human filter on. Artists will integrate generated passages
Speaker 1: into they orchestrate, and will soon train their own models.
Speaker 1: Another protection might be to flag accounts it posts on
Speaker 1: an inhuman amount of content, but that too may seem retrograde.
Speaker 1: We live with infinite media now. I encourage platforms to
Speaker 1: be more curated, but enforcing a hard human AI binary
Speaker 1: is not the right way to address this long term.
Speaker 1: If an expert uses AI to help them write an article,
Speaker 1: they otherwise would not have the time to complete it
Speaker 1: will not be of less substance than a non expert
Speaker 1: with time on their hands writing something by hand. A
Speaker 1: practice or reputation of thought and work developed over time
Speaker 1: is the best filter. The contemporary demand for twenty four
Speaker 1: to seven content from people is the problem. So that's
Speaker 1: that's basically her argument. So she's concerned. She has some
Speaker 1: of the same concerns you, I mean from a little
Speaker 1: bit of a you know, from a different perspective.
Speaker 2: But but.
Speaker 1: Yeah, you know, I think the tooth based is probably
Speaker 1: pretty far out of the tube at this point. As
Speaker 1: I like to say, Now, there's another element to this though,
Speaker 1: and this is this is the part that people don't
Speaker 1: talk about very much. But this came up on the
Speaker 1: show not long ago. These streaming platforms because of the
Speaker 1: rise of AI generated music and and how easy it
Speaker 1: is to make it, and and it is a form
Speaker 1: of spam. I think Holly Herndon referred to it as
Speaker 1: sound spam when you just if you just make a
Speaker 1: whole bunch of AI tracks and just upload them all
Speaker 1: at once to Spotify and bandcamp, YouTube and wherever else,
Speaker 1: it does cause a problem. Uh, in terms of storage
Speaker 1: because you know, we like to think of you know,
Speaker 1: it's just data. It's not it's not something, uh you
Speaker 1: know that you can touch or feel or it's or
Speaker 1: hold it. It's data is Therefore, there's just a It
Speaker 1: doesn't take up any actual room anywhere. You just have
Speaker 1: this infinite amount of storage space to put data. But
Speaker 1: that's not true.
Speaker 2: You have to.
Speaker 1: You do have to create the physical mechanisms with which
Speaker 1: to store that data, and you do have to make
Speaker 1: more of those if you more and more data you
Speaker 1: have to store. So if you think about this, I
Speaker 1: don't know the numbers offhand how many different song files
Speaker 1: are on band Camp and Spotify, and obviously it's in
Speaker 1: the millions, perhaps in the billions, I don't know. But
Speaker 1: if if that's all been ratcheted up exponentially because of
Speaker 1: the rise and the ease with which you can create
Speaker 1: AI generated music, then that's a problem. So of course
Speaker 1: they have to find ways to filter this stuff out.
Speaker 1: And then I found this. This is on Music Businessworldwide
Speaker 1: dot com And this just went up a couple of
Speaker 1: days ago.
Speaker 2: Music streaming.
Speaker 1: Music streaming platform now host a quarter of a billion tracks.
Speaker 2: Where does it end?
Speaker 1: And uh, this is Uh, this is short, but it
Speaker 1: just kind of gives us an idea of the scope
Speaker 1: of the concern here, So it says here there were
Speaker 1: two hundred and fifty three million music tracks sitting on
Speaker 1: audio streaming services at the close of twenty twenty five
Speaker 1: yep over a quarter of a million. According to new
Speaker 1: data from Lumini's new annual report, that was that was
Speaker 1: up only thirty seven point nine million tracks, an average
Speaker 1: of one hundred and six thousand uploads per day. Most
Speaker 1: of this music was far from popular. Almost half of
Speaker 1: the two hundred and fifty three million files hosted by
Speaker 1: audio platforms received fewer than ten streams last year, Almost
Speaker 1: three quarters of it received fewer than one hundred annual streams,
Speaker 1: and nearly nine tenths of it received fewer than one
Speaker 1: thousand annual streams. We're obviously a long long way from
Speaker 1: Apple's promise to deliver one thousand songs.
Speaker 2: In your pocket with the iPod.
Speaker 1: Yes, as AI generated music proliferates, are we now just
Speaker 1: a few short years from the likes of Spotify hosting
Speaker 1: quote one billion songs in your pocket. And wouldn't such
Speaker 1: a tidal wave of content inevitably swamp digital services while
Speaker 1: hurting artists, songwriters and the perceived value of music amongst consumers.
Speaker 2: This is all.
Speaker 1: These are all legitimate concerns. In my opinion, I don't
Speaker 1: know what the answer is. By the way, off, I'm
Speaker 1: not here to offer solutions because I have no idea
Speaker 1: what the answer is to this. But these are legitimate concerns.
Speaker 1: Universal Music Group CEO and chairman Sir Lucian Grange certainly
Speaker 1: thinks so, and he's keen to stop it from happening.
Speaker 1: Last week, Grange delivered a stark warning about AI generated
Speaker 1: content overwhelming streaming platforms. In his twenty twenty six New
Speaker 1: Year's Address. He wrote, quote, validating business models that failed
Speaker 1: to respect artists' work and creativity and promote the exponential
Speaker 1: growth of AI slop on streaming platforms is a grave
Speaker 1: disservice to artists', songwriters and all of us who work
Speaker 1: in music unquote, this warning was not theoretical. Last year
Speaker 1: Spotify removed think think about these numbers. Last year, Spotify
Speaker 1: removed over seventy five million spammy tracks from its platform,
Speaker 1: while rival Deezer reported receiving fifty thousand fully AI generated
Speaker 1: tracks per day, accounting for thirty four percent.
Speaker 2: Of all daily uploads.
Speaker 1: So think about that. So on any given day, deezer,
Speaker 1: what's being uploaded? A third of it a third is
Speaker 1: AI generated tracks. Yet despite this aggressive purging of content,
Speaker 1: the total amount of music and the global audio streaming
Speaker 1: ecosystem continued to soar. In twenty twenty five. Discussion around
Speaker 1: tracks that attract fewer than one thousand streams per year is,
Speaker 1: of course particularly relevant. Here illuminates new numbers help illustrate why,
Speaker 1: with UMG's encouragement, audio streaming services moved toward artist centric
Speaker 1: style payment models a couple of years ago, and why
Speaker 1: these models are now increasingly being stress tested by AI.
Speaker 2: In early twenty twenty.
Speaker 1: Four, Spotify introduced a threshold requiring tracks to attract at
Speaker 1: least one thousand plays in a twelve month period to
Speaker 1: qualify for royalty payouts, which a lot of people were
Speaker 1: mad about.
Speaker 2: They were critical of I get it though at.
Speaker 1: The time Spotify said ninety nine point five percent of
Speaker 1: streams on its platform went to tracks exceeding the one
Speaker 1: thousand annual streams threshold.
Speaker 2: Under its new.
Speaker 1: Policy, it's said each of those tracks would earn more
Speaker 1: with payouts previously going to sub one thousand stream tracks
Speaker 1: redirected back into the royalty pool. Deezer's artist centric model,
Speaker 1: launched in partnership with you Universal and Warner, achieved a
Speaker 1: similar aim, providing a double boost in royalties to artists
Speaker 1: with a minimum of one thousand streams per month and
Speaker 1: five hundred unique listeners.
Speaker 2: This goes on.
Speaker 1: There's a lot of numbers here, and we don't have
Speaker 1: time to get into all of it, but the article
Speaker 1: isn't that long, so if you want to check it out,
Speaker 1: Music Businessworldwide dot com, which is a great website for
Speaker 1: music news if you're interested in the industry and so forth.
Speaker 1: But we are rapidly running out of time, so I
Speaker 1: think we will close with that
Podbean