Benji Rogers on Attribution, AI, and the Next Napster Moment

Technically Creative by KoobrikLabs Podcast Transcript

Title: Benji Rogers on Attribution, AI, and the Next Napster Moment

Orlando Wood (00:00)
couple of things. One is if I look like with while you’re making it,
if while you’re answering something or whatever else, I look like I’m
looking at anything up here. I’m not online shopping or anything.
just, that’s where I kept keep all my notes and everything. So I’m
just kind of working out what to ask next and what to shift around,
things like that.
Benji Rogers (He/Him) (00:17)
No problem. I’m also just putting my video into 4k. Does that is that
better or worse for you? It depends on on sometimes the uploads where
that’s better. Okay, hang on a second. Let me see. I think this might
4k might not work and go we’ll go 180. How about that?
Orlando Wood (00:23)
Better. ⁓ now it’s worse.
That looks
very good. That looks better.
Benji Rogers (He/Him) (00:35)
or that.
Orlando Wood (00:36)
That’s pretty good too.
OK, great. Awesome. And then the other thing I wanted to ask you
before we got started is how should I speak about Lark 42 and surreal
AI? How do you like? I’m going to introduce you obviously as you know,
founder of of Lark 42 and surreal AI. How?
Benji Rogers (He/Him) (00:55)
Founder
of Lark 42, co-president of Surreal AI. It’s both, yeah.
Orlando Wood (01:00)
Okay.
Okay. Okay. And then LARP 42 is, is, you know, consultation, right?
Benji Rogers (He/Him) (01:09)
Yeah, so basically we’re advisors to music tech companies or to
technology. We help technology companies understand music and media
and media companies understand technology. And then we also help with
fundraising and that kind of thing. And we found surreal because they
were client and they were so, so wowed by the technology and sort of
where it was going that we, that we went for and joined. yeah, famous
last words.

Orlando Wood (01:26)
Right, okay.
Great. ⁓
Uh, uh, 42 and.
Okay, Brill.
Okay, well I’m just gonna get started if you’re ready.
Benji Rogers (He/Him) (01:43)
We’re all counting on you.
Orlando Wood (01:45)
So Benji Rogers, I’m really excited to have you on the podcast. I’m
really excited because I haven’t spoken to anybody from the music
business yet. It’s also a business I know considerably much less about
than many of the other creative industries. But you are partner at
Lark 42 and co-president of Surreal AI. Lark 42 being a sort of
consultant to music and tech companies and Surreal being an
attribution layer for generative AI models.
So, yeah, can you just briefly introduce yourself and sort of say what
you really do and then we can kind of get into it from there.
Benji Rogers (He/Him) (02:20)
Sure, yeah. I’m a recovering musician deep down. So my entire life has
spent. I’m the son of three managers, my mother, my father, my
stepfather. So literally there was no avoiding the music industry. And
ultimately in 2009, I founded my first tech company. So I approach all
technology from the lens of someone who’s been a CEO and someone who’s
raised investment rounds. And in 2009, I built a direct to fan
platform.
So that was basically like a Kickstarter for music, if you will, with
some tweaks in it. And then after that, I ended up basically, know,
streaming services and people will come to me like, hey, can you help
me solve this problem? You understand, you know, how the relationship
between artists and fans works. And so it kind of built out from
there. And then my co-founder in Lark, he was like, you know, that
like,
companies hiring you to help them navigate stuff is a thing. And I was
like, I didn’t know that I, you know, I’m an auto didact. So I didn’t,
I didn’t, I didn’t learn this stuff the hard way. So to me, it was
like, Hey, how would you look at this? And so we did it that way. And
so we basically sit at this intersection because, you know, as you
said, know, music and music rights are very specific and sort of like
esoteric, arcane, lots of terminologies, lots of acronyms. And so we
help technology companies understand that.

companies hiring you to help them navigate stuff is a thing. And I was
like, I didn’t know that I, you know, I’m an auto didact. So I didn’t,
I didn’t, I didn’t learn this stuff the hard way. So to me, it was
like, Hey, how would you look at this? And so we did it that way. And
so we basically sit at this intersection because, you know, as you
said, know, music and music rights are very specific and sort of like
esoteric, arcane, lots of terminologies, lots of acronyms. And so we
help technology companies understand that.
Orlando Wood (03:17)
Right.
Mm.
Benji Rogers (He/Him) (03:40)
I mean, also music companies understand technology because they really
do speak different languages. And we feel that ultimately there’s a
huge amount of value that that could be harnessed in music via
technology. But generally speaking, music companies have been at the
effect of technology and not driving it. And that’s new. That wasn’t
the case for some time. So we sit in that weird intersection and then
Surreal AI is basically
Orlando Wood (04:01)
Yeah.
Benji Rogers (He/Him) (04:09)
You know, when people will say, AI is there like black boxes? Well,
that’s a choice, a, it’s an option. And so what we do is, you know,
Surreal basically helps to unpick what’s in that black box. So you can
understand what led to what. The ultimate vision for that would be you
can create anything from everything and know how it got.
Orlando Wood (04:29)
And when you say the black box, do you mean in the creation process
of, know, of, know, what prompts are being used and things like that?
Or do you mean in the training data set, the foundational data set?
Benji Rogers (He/Him) (04:39)
Yeah,
yeah, yes. So the way this works is kind of unique. The founders, it’s
an interesting story because the founder himself was sort of an
explainable AI and he helped teach robots how to move and those types
of things. But this idea was why does the AI do what it does? And what
they did was they spent years perfecting this process, which meant
that when the prompt is being written,
it can tell what from the corpus is being pulled on and what it leads
to in the end. it’s, we call it chain of attribution, which is what
specific pieces of the song, the video, the audio, the text that it
take from, but then also how important was that to the learning
process so that like, you can basically give an attribution, what we
call chain of inspiration. you know,
Orlando Wood (05:09)

Interesting.
Benji Rogers (He/Him) (05:31)
If your piece of video taught the model how to be really good at this,
this and this, then you should probably have more attribution than
just something that was incidental to it. We call this also AI
business intelligence because basically if you’re an AI company,
knowing what prompt leads to what data can be incredibly helpful. so
Surreal is basically comes from the framework that like most artists
and creators are afraid
of getting involved in AI. We think that they should opt into AI
training. They should have the ability to opt out if they want to. And
so one of the really clever parts of Surreal on the opt-out side is if
you want your work not to be included, you have to show the model what
not to include. So what that means is, as the prompt is saying, hey,
make me video just like Orlando’s last five videos, if
those are opted out, it goes, wait a minute, I’m about to hit this
node I shouldn’t hit. And it can nudge the model around it so it
doesn’t infringe or say, stop the generation process here and let the
user know. Because inherently, you know about this as well, but these
systems are not, because they’re predictive, they need to understand
or they need to be shown what to avoid as much as what to use.
Orlando Wood (06:30)
Mm-hmm.
Benji Rogers (He/Him) (06:51)
and current attribution systems look at the outputs and say, this is
likely from here. What we do is, is the AI company installs five lines
of code, which allows us to view when the prompt leads to what in the
corpus that we have. And then, so an artist would put their music or
their videos in, we break it apart into this, what we call AI DNA.
And so what we can do is we can say, every time, you know, this prompt
activates this, this is where it comes from and we can give
attribution. So we don’t do the licensing. We allow the rights holder
to license the AI company. We’re what keeps the attribution flowing.
The muscle that calculates how much is owed to whom and then hopefully
the payments flow from there. So as you can imagine, really, really
popular.
Orlando Wood (07:39)
Right.
Benji Rogers (He/Him) (07:40)
Because
the current lawsuits from AI companies aren’t designed on how do we
pay people. They’re more designed on this is for use, we shouldn’t
have to. And I think that that’s an awful argument to be making.

the current lawsuits from AI companies aren’t designed on how do we
pay people. They’re more designed on this is for use, we shouldn’t
have to. And I think that that’s an awful argument to be making.
Orlando Wood (07:49)
Yeah.
So you, so to, well, my first question is, so your tool gets
implemented on the model side. So it would be the open AIs, it would
be the VO3s or Google. It would be them implementing this on their
side, presumably after the fair use argument is overturned in court
potentially, or they just decide for propriety’s sake that they think
that they should start.
attributing and potentially remunerating artists whose work has been
used to train.
Benji Rogers (He/Him) (08:26)
Yes, so the theory is this, the rights holders have all these amazing
assets. Whether if you’re film company, you’ve got TV shows and
movies. If you’re a music company, you’ve got these music assets. And
in music, there’s a specific to this, which is the publisher, the
person that wrote the song and the person that performed the song were
different. So you can train an AI model on karaoke versions. Doesn’t
mean you have the right to the publishing.
In the same way that if you’re making a movie, you can’t just take
someone’s song and put it in the movie. That’s a synchronization
license. That’s a different piece. So the theory is, and what we do
with our current clients is this, they put their music into the
surreal AI system. We turn it into this AI DNA and put the music
behind APIs. Then compliant AI companies come, they install these
lines of code so that when
this work is accessed, we can point to how to compensate the right
people. And so it’s the only, know, the AI companies being blunt with
you don’t want to do it. They would rather just keep the rights
holders blind. I think they’re inherently adversarial to rights as
we’re seeing from, you know, the anthropic lawsuit just recently, and
also from the Suno lawsuit literally just dropped just now saying,
that they don’t, that because it’s not a sample, they shouldn’t have
to pay anything. again, they’re taking fair use for a conclusion. The
challenge with the fair use argument, whether we agree with it or not,
and I don’t, I think it’s bullshit because it, AIs don’t work like a
human brain. They’re inherently different. And so you can look at a
mathematical one-to-one from this led to this, but you can also look
at it as a musicologist would, this led to this.
Orlando Wood (10:02)
Yeah.

Benji Rogers (He/Him) (10:11)
And so the idea is that the rights holders put their works in in
various states have opted in or opted out. We have about 10 million
tracks in at the moment. The AI companies install five lines of code
and then we can basically, and then we basically help the license, the
licensure and licensee keep each other honest. So that, so that, you
because, you know, you don’t want to license just based on subscriber
numbers. You don’t want to license based on genre.
You need to license very specifically for the use cases. So you do
need both to play ball. We do have a technology called attribution
likelihood, and we’ve done big tests on this sort of 98 % correlated
with human perception. But ultimately, that’s what we call a black box
model. But the real power of the technology is the white box model.
And so there’s a couple of AI companies now that are implementing it.
We’re doing some testing with a couple of the
you know, larger companies in the music space as well and in other
media. It was built originally for images. It works for video,
although we’re not deploying that yet. We need to get the use case and
it can work for basically any media. But the overall philosophy
Orlando, and I think, you know, anyone who makes anything would agree
is, you know, it’s designed for a future in which you’re sitting at
home and you say, hey, Netflix, make me a TV show.
30 minutes long, starring Orlando, Benji, and Brad Pitt and Julia
Roberts, circa 1994. 32 minutes, give it a soundtrack, kind of like
Reservoir Dogs, put old 1950s comic book art on the walls. Everything
just described is owned by somebody. And so, if you think of it from
the music level, as it generates the video, there’s always an
underlying music bed.
And if we don’t lock attribution to that, it’ll still generate the
music. That music will still have been trained on real music. The
people just won’t get paid. And so what we don’t want to have happen
is training off of synthetic versions and basically extrapolating
further and further away from the originals. Cause ultimately, I think
that the business model that could be really interesting for music
specifically is make me a song that sounds like Taylor Swift.
No problem. That’ll be two bucks. Right? So you can have the premium
version, right? Platform takes 15%. 85 goes to the rights holders. We
got no problem. This is a great business model. it all day long. And
then if it ends up on Spotify, money can go through that way. And I
think that we tend to think of it as like, the analog is really from
the music industry standpoint. I know you’re from…
Orlando Wood (12:45)
Yeah.

Benji Rogers (He/Him) (12:55)
from different world and on the media side, but the music industry
lost 60 % of its value with the introduction of MP3, right? More
humans than ever on earth consume music. And yet what’s interesting is
Spotify, which takes 30 % of the revenues from music and pays 70 % to
the rights holders is worth more than those rights holders. How is
that possible?
And it’s because in the end, where the art meets the human is where
the money is. And what I’m, what we’re trying to create in a, a, in in
a world of attribution is where as the model scales and the success of
the AI company scale, the rights holders go with them as opposed to
being at the effect of them. Because in theory, you know, the analog
will be Apple, right? You know, Apple, Apple.
Orlando Wood (13:28)
Yeah.
Mm-hmm.
Benji Rogers (He/Him) (13:54)
licensed iTunes early days. Three bucks went, know, $3 of every 10
went to the went to Apple. $7 went to the rights holders. But Apple
also sold iPods at a 38 % margin on every device. So we stand at a
precipice. Music has generally been the canary in the coal mine, but
TV, film and video is next. And it is coming at us at a
You know, MP3 sort of digital revolution that hit music industry took
about four or five years to really bite. And in a quarter of a century
since then, it’s still two thirds the size it was in 1999. it took 25
years to claw back two thirds of your value from where you were in CD
era. That’s AI is, you
Orlando Wood (14:36)
Mm-hmm. Yeah.
Benji Rogers (He/Him) (14:49)
At scale is what a year and a half old really like at scale and
already 700 million humans use chat GPT alone, let alone Claude
through API calls that loan Gemini. mean, it is scaled. so my, my, big
fear for rights holders of all kinds is if they don’t think protection
control monetization in that order, then they’re basically handing
over.
what could be the largest wealth transfer we’ve ever seen and it will
happen in TV and film too. It’s just inches away.
Orlando Wood (15:26)
Yeah, you know, you’ve I’ve heard you draw parallels before between
today’s AI lawsuits and the Napster Wars. Why do you think the music
industry is at risk of repeating the same mistakes it made with
streaming?

Yeah, you know, you’ve I’ve heard you draw parallels before between
today’s AI lawsuits and the Napster Wars. Why do you think the music
industry is at risk of repeating the same mistakes it made with
streaming?
Benji Rogers (He/Him) (15:32)
Mm.
Because it’s the way you license, it’s the way that you accrue value.
Ultimately, my first company was a direct-to-fan company. And what we
tried to be was the thinnest skin between the artist and the fan. So
we took 15%, we paid 85 % to the rights holder, and everyone won. But
really what we were doing was bringing the artist and the fan
together.
And so Spotify controls the audience. So what happens in the Spotify
case, to give you a specific example. So in the end, what is Spotify’s
reason to be? Its job is to take the listener of music and turn them
into a subscriber. And it wants to give that subscriber what they want
to stay on platform for the least cost possible to themselves. Right?
So that takes one of two forms.
If it comes in and says, hey, out, you know, you know, stream
Orlando’s new song. So your song’s now streaming and Spotify’s
algorithm says, you know what, if we could get that listener to Joe
Rogan for three hours, we don’t pay any royalties on that. That’s
three hours royalty free. Or if we get it to some AI slop that has 50
% royalties or we pay a musician to make a fake song. So.
they’re incentivized to basically pay out the least possible to rights
holders, maximizing their slice of the pie as they go forward. And
they do that and they’re about to release messaging by the way,
between them to make the app even stickier. They’re putting video into
it. And the whole reason is they’re in that process of basically
milking that relationship with the user. So in the same way that
Instagram doesn’t charge
Orlando Wood (17:08)
Hmm.
Mm-hmm.
Benji Rogers (He/Him) (17:28)
people to use it, but Instagram is monetizing the user, right? So, you
know, the typical artist says, check out my Instagram, go to my
Spotify. Those are two things that are as far away from the artist as
possible. In the AI world, the interface is the chat window. It’ll be
the voice command or it’ll be the goggles, the glasses, and it’s even
stickier than those.
because it’ll be the place at which people say, you know, make me a 30
second video to send to my wife, girlfriend, boyfriend. And that is
going to have a soundtrack under it. That soundtrack isn’t going to
be, let me select a song from the top. It’s just going to come. And so
I think the reason that the analog is this. The music industry
licensed away that value. It wasn’t stolen. It wasn’t taken. It was a,
it was a series of decisions in the licensing frame.

because it’ll be the place at which people say, you know, make me a 30
second video to send to my wife, girlfriend, boyfriend. And that is
going to have a soundtrack under it. That soundtrack isn’t going to
be, let me select a song from the top. It’s just going to come. And so
I think the reason that the analog is this. The music industry
licensed away that value. It wasn’t stolen. It wasn’t taken. It was a,
it was a series of decisions in the licensing frame.
Orlando Wood (18:03)
Mm-hmm.
Yeah.
Benji Rogers (He/Him) (18:24)
that gave away more than it kept. YouTube, if you think of YouTube’s
music alone, or you think of Meta’s music alone, these are massive,
massive platforms, all of whom artists are pumping value into every
day, and they’re not getting it back in proportion. So what I think
can happen in AI that’s different is, because Spotify is a good
interface.
Orlando Wood (18:28)
Mm-hmm.
Benji Rogers (He/Him) (18:53)
GPT is incredible. You’ll notice that GPT will begin to give you
little things you didn’t ask for. Do you want to refine that question?
What do you think? Do you want more? All in that thing of keeping the
user on. And so if you think about where these generating interfaces
are living, they’re living in TikTok, they’re living in WhatsApp,
they’re living in Instagram, Spotify, on and on. And so what I feel
could happen is this. If we license with attribution,
then every single generation feeds somebody in some way, shape or
form. So the bigger the platforms get, the more generations we get,
the more money flows. But if we license flat and blind, it’s like
saying, I don’t know Spotify, we have 12 million songs, just you
figure out what to pay us. And that puts you at a disadvantage in the
same way that like even in the existing frameworks of licensing, the
music industry is still behind.
Orlando Wood (19:27)
Mm-hmm.
Benji Rogers (He/Him) (19:52)
It should be five times the size of the platforms because it is the
platforms cannot exist without that media. so, you know, they’ll
distribution is king. I do agree with that. But ultimately, you know,
the value should accrue to artists and it can, but we have to license
in such a way that doesn’t say, please, could we have attribution? It
says you’re an adversary without apps.

without attribution, period. And it’s not even worth having a
discussion, yeah.
Orlando Wood (20:21)
So is it?
Right. Is it all or nothing then? if the the AI companies win the fair
use argument, all of this becomes all this preparation, doesn’t it all
become a bit moot? mean, you know, I I keep thinking that there’s
going to be a moment where clearly it’s going to be defined and
everything that you’re saying, attribution and remuneration and, know,
like we were saying, credit, consent and compensation are the
takeaways.
But if they win fair use, is it even?
Benji Rogers (He/Him) (20:55)
so I think.
Orlando Wood (20:55)
I guess it becomes with
new music created by artists because they have the potential idea of
restricting output, but even still, moment they release it, it becomes
something that the models can then train on, and then they don’t even
have to take this into consideration.
Benji Rogers (He/Him) (21:08)
Yeah. The way I think, well, so I mean, just from a pure surreal
perspective, there’s a huge business in helping AI companies figure
out what prompts came from what, like that’s a business in and of
itself. There’s also other verticals. There’s medical data, financial
data. There’s all kinds of ways that this could go for surreal. On the
music industry side, and this is one of the, one of the things I’ve
been trying to say to rights holders since April, since we started
this last year is
Orlando Wood (21:21)
Hmm.
Benji Rogers (He/Him) (21:36)
these lawsuits are probably not going to help you. This is a
technological problem that has a technological solution. And one of
the things that happens in this industry, because it’s disparate,
because you’ve got the publishers and the performers separate, they’re
thinking license first, then legislation, then litigation. They think
in that order. And they haven’t made it easy. And what
What we’ve been arguing for is put all your music behind these APIs.
And then when you walk into court, you can say, look, all these
companies have to do is install the attribution framework via Google
GCP marketplace and everything’s done. What you can train on, what you
cannot train on is all built in and that’s all possible. But I think,
you know, if we’re honest, a lot of these companies don’t even have an
up-to-date database of who owns what.

What we’ve been arguing for is put all your music behind these APIs.
And then when you walk into court, you can say, look, all these
companies have to do is install the attribution framework via Google
GCP marketplace and everything’s done. What you can train on, what you
cannot train on is all built in and that’s all possible. But I think,
you know, if we’re honest, a lot of these companies don’t even have an
up-to-date database of who owns what.
Publishing can be fragmented in a single song across multiple writers,
multiple performers. And so it’s this classic thing of like, well, if
we just sue, we’ll be okay. And I think as with AI safety and all
kinds of things, this is a really narrow path. And I’ve said, for
Porepublicly, the odds are slight that we get this wrong, that we get
this right, and more likely that we get this wrong.
only because what other industry on earth wouldn’t think protection
first, right? Start with protection. You have your legal protection,
your technical protection. And basically what you do is you imagine if
you imagine if they walked into the judge and said, listen, all of our
work is bagged and tagged behind API is ready to go. Our licensing
framework stipulates that all the licensing company has to do is
insert five lines of code and attribute every time our work is used.
It’s really simple.
We come up with a fee for what that’s based on and all boats rise. The
judge is going to sit there saying, well, that’s logical. Why wouldn’t
you do that? Instead, what they say is, well, we’ve been trying to
license these companies for two years. We can’t get them to move. They
don’t have the correct data. They don’t know what’s up that
interrupted out and they haven’t had a conversation with their artists
and creators yet. I think the same thing will happen in film and
television because if you don’t
have everything in order ready to go in a digital economy, it just
becomes a route to inefficiency. And that’s why Trump or others can or
the administrations and in the UK and across all over can say things
like, we mustn’t let burdensome licensing slow progress. Now, I would
argue, you know, who woke up this morning saying I will die if I don’t
have a device that will generate
you know, songs for me. I mean, that was a problem anyone actually
had. But now that it’s here, you can see how a judge would have little
sympathy for a bunch of people who are sitting there saying, don’t
train on my work. We can’t show you what it is. We can’t show you what
to avoid. We have no framework for licensing ready to go, but don’t do
it anyway. That’s a hard one to win, right? And I think it’s, it’s,
it’s been, yes. So
Orlando Wood (24:19)
Yeah.

Benji Rogers (He/Him) (24:44)
There’s good odds it doesn’t work in music, but there’s, I think that
there’s bigger, you know, we chose music because we love it, because
it’s the industry we’re from, the founder of the company, his family
are all creators, and in the end, we want art to do well. But, you
know, there’s a shot it doesn’t work out, there just really is, and
there’s no getting around that. But I think it’s also incumbent.
It’s important, not in common, it’s important that we think of these
AI companies as what they’re trying to do. And what they’re trying to
do is not pay the people that made the stuff that make their systems
possible. And you have to start with that adversarial stance. They can
prove that they’re not doing that by licensing and doing all the fun
things that they can. What’s interesting is when you have these
conversations with them, not all, but a few of them, they think…
Orlando Wood (25:18)
Of
Benji Rogers (He/Him) (25:33)
rights holders are idiots. They’re like, these people will give in.
They’re so thirsty for money. We just need to wait them out and wave
big valuations at them. And it’s an asymmetrical warfare. It really
is. They’ve got immense legal power. They’ve got funding rounds. But I
think in the end, what this will do to the culture in the long run is
there’ll be just networks flooded with AI slot.
and it will actually remove interest from it in fandom. Some people
will get highly addicted to it, but what we might have is a real
renaissance of in-person, live, analog, as it’s happening, only
happens here, not captured, but in the moment. And that could mean
some incredible stuff. So I think for those not locked into the AI
slop ecosystem, it could be amazing. For those that are,
Orlando Wood (26:05)
Mm-hmm.
Benji Rogers (He/Him) (26:29)
generally degrading and pretty terrible. sometimes you need things to
get awful before the good begins to emerge. And we could have some
incredible new creations on the
Orlando Wood (26:41)
Yeah, I mean, you you were talking about this, this, you know,
infinite AI slop. You’ve suggested a split future sort of mass AI
driven content versus unique human work that stays outside of the
digital rails. What do you wear? How do you see that shaping up?
Benji Rogers (He/Him) (26:56)
Yeah. So, you know, if you’ve ever been to a house concert where an
artist plays at someone’s house, have you ever been to one, but like
there’s this thing called Sofar Sounds and that happens all the time.
And they curate three artists who show up and play songs. And it’s not
really captured anywhere. It’s only experienced in that moment. And
it’s really special. The way that musical was was right. You know, you
would go to a play, the play happened that night and then it was done.
It was different every night. And so

Yeah. So, you know, if you’ve ever been to a house concert where an
artist plays at someone’s house, have you ever been to one, but like
there’s this thing called Sofar Sounds and that happens all the time.
And they curate three artists who show up and play songs. And it’s not
really captured anywhere. It’s only experienced in that moment. And
it’s really special. The way that musical was was right. You know, you
would go to a play, the play happened that night and then it was done.
It was different every night. And so
I think that there’s, Substack’s a great example, right? Substack is
this really interesting platform where it’s siloed into like, you
know, subscribers get more than if you’re not a subscriber. And in the
end, I think that those artists that just say, check out my Instagram,
go to my Spotify are competing with everything. But if you begin to be
like Jeff Tweedy, who’s got hundreds and 130, 140,000 subscribers.
all of are getting a different journey than if you’re on Instagram,
than if you’re on Spotify elsewhere, you end up, it ends up becoming
more valuable because it’s not for everybody. And so mass culture will
probably continue and these AIs are going to become addictive. I mean,
they are going to be so good at hooking people. It’s going to be
scary. You know, the ultimate porn of your dreams, the music of your
dreams, the music that makes you lose weight, the music that makes you
exercise better, like whatever it is.
That customization, perfect fit content is what Spotify calls it, is
the goal of daily active usage. What I argue in my newsletter
endlessly is if the measure is daily active usage, then the outcome is
in the end addiction. And the reason that that’s so important is
because, and we have to reckon with that, is the…
Algorithms are rewarded for daily active usage. you know, Netflix once
joked, you know, the, you know, our only competition is sleep. That’s
a really dark vision of humanity. Right. If you, if you imagine an 11
year old trying to form their brain and they’ve got an infinite AI
chat companion and that chat companion says, Hey, I wrote you a song
just for you Orlando, check it out. And then it goes to a couple of
days later, Hey, didn’t, you didn’t listen to my song today.
Orlando Wood (28:57)
Mmm. Yeah.
Benji Rogers (He/Him) (29:16)
Didn’t you like it? Like it could become coercive, all of these
things. And we don’t have guardrails that are sufficient that I would
ever trust. So, and then there’s the other people like me who I have a
light phone. It is black and white. It doesn’t have a browser. And so
when you go outside of that world, you begin to see, you know, makers
and markets doing things that are not part of that digital space.
Orlando Wood (29:41)

Mmm.
Benji Rogers (He/Him) (29:42)
And so we will see some culture, but there will be lot of casualties.
And that really makes me sad because we have much better options.
Orlando Wood (29:52)
Do you think creative industries are inadvertently designing for
addiction the same way social platforms did with infinite scroll and
autoplay?
Benji Rogers (He/Him) (29:59)
Creative companies, I think they’d take it if they could. Social media
companies are. This whole bet on AI infrastructure, I don’t believe
for a second is artificial general intelligence or any of that stuff.
It’s about basically creating the most addictive technologies ever
known to humans.
If you, if you can hook someone to TikTok in 32 seconds, in 32 minutes
with curation, imagine what you could hook them with, with creation.
And that’s the real dangerous part. Custom specific for you. So I, I’m
highly suspicious of when companies like Meta, XAI or others say,
yeah, we’re in it for AGI. I don’t believe that for a second. I think
it’s basically saying,
Orlando Wood (30:33)
Yeah.
Benji Rogers (He/Him) (30:48)
We can invest $400 billion in cloud infrastructure because the
business model of addicting humans to content will make 10 times that
if we get this right. But that’s a, the human toll that that will take
is catastrophic. Like can a democracy survive a place where people are
literally viewing human?
and a i created information spaces they cannot stop looking at i mean
just look at a child with an ipad watching you too they literally that
they will say the most hurtful mean things you take it away so that
that that i don’t fear a g i robots taking over dystopia i think that
that that that l l itself is the wrong technology for that anyway i
fear what people like zuckerberg who’ve shown that they will
experiment
with the stuff on children will do when given a power to create stuff
that will make people literally unable to look away. So that’s my
dystopia. It’s already partially there, but that’s the bit I worry
about the most.
Orlando Wood (31:52)
Mmm.

Yeah, I mean, for sure. think that, you know, listen, I mean, the the
the free speech argument that they peddle, the you know, the fair use
arguments that they’re peddling. mean, it’s all couched in this
betterment of humanity situation. But obviously, they’re they’re built
on absolutely massive valuations and they need to get attention and
sell advertising and do all of those things. So making things hooky is
very much what they do. mean, I, you know, I kind of always said that.
There is lots of publicly available free language information. It’s
mostly legal, but there’s court documents and motions and things that
are publicly available at almost every courthouse in America. They
could have had a very good legal, very good language model.
based on those things, but they knew the things that would actually
get people to want to interact with this technology is art and poetry
and film and music and all of those things. That’s why they ingested
these things. Not because, you know, it has some huge benefit for AGI,
but because it and or the betterment of society, but because it’s
going to addict people to the use of these tools. Who wants to chat
with the chat bot who doesn’t that doesn’t know what Rocky is?
You know, or or Spider-Man, you know, like it’s just it becomes
pointless. You don’t then have a core cultural understanding with this
thing that that that they’re trying to get you to use every day.
Benji Rogers (He/Him) (33:24)
Yeah, I think you’re absolutely correct. And when I hear things like
AI is the cure for loneliness, conversation is the cure for
loneliness. Connection is the cure for loneliness. And you know, it’s
interesting when you and I first met on Zoom, haven’t been in the same
room together, like on and on, but we were connecting in this way. And
this is, again, this is a wonderful use of technology. Like I love
that we can be in this software, have this conversation, share this
conversation with people.
You know, ultimately, I think, you know, the thing I fear is, is when,
when companies have shown over and over again, that the remedy of law
is where people have to go to first to stop the harms, something’s
gone wrong. You, our children shouldn’t be guinea pigs for AI
companies. Our democracy should be safe, should be hardened as opposed
to weakened by this technology. But ultimately, what is the first
vector that people
Orlando Wood (34:05)
Yeah.
Benji Rogers (He/Him) (34:20)
who would harm democracy’s attack, right? It’s misinformation across
social media platforms. AI could absolutely supercharge this, not in
an AGI taking over the world, but literally in a kid at school on
their Chromebook, suddenly chatting to Gemini and getting into a
conversation about something. Now all of a sudden it’s like, well,
that’s easier than talking to that girl or boy I kind of think is
cute.

who would harm democracy’s attack, right? It’s misinformation across
social media platforms. AI could absolutely supercharge this, not in
an AGI taking over the world, but literally in a kid at school on
their Chromebook, suddenly chatting to Gemini and getting into a
conversation about something. Now all of a sudden it’s like, well,
that’s easier than talking to that girl or boy I kind of think is
cute.
Orlando Wood (34:33)
Mmm.
Mm.
Benji Rogers (He/Him) (34:51)
And so I think one of the things that I’m a parent and I think of this
from the lens of my kid who’s 13, which is like, how much time am I
comfortable with them talking to a machine that literally lies,
cheats, manipulates on and on? It’s not doing it on purpose. It’s the
nature of it. And why are my kids able to access this stuff when
Orlando Wood (35:09)
Yeah.
Benji Rogers (He/Him) (35:14)
It hasn’t been proven that it’s safe. Why are they the test cases?
Like, no other, nothing should work this way. It should go the other
way. Prove it’s safe, then we use it. And I do understand that like,
you know, there’s adversarial powers at there. But I think in the
longest run, we’re gonna look back at this time and say, and the
people at the forefront of this technology will be vilified.
Orlando Wood (35:17)
Mm-hmm.
Mm-hmm.
Benji Rogers (He/Him) (35:40)
for the terrible things that they did to people. And I believe one of
the reasons that attribution, not to tie it back to it, but I think
it’s so important is as you’re auditing what happened, knowing what
caused what and why could be the difference maker in how to secure
these models, how to ground them better in truth.
Orlando Wood (35:42)
Yeah.
Benji Rogers (He/Him) (36:00)
how to create systems whereby it’s like it can read a Wikipedia page.
It does know how many hours there are in blueberry. Like all of these
things are possible. And I think Gary Marcus has argued like, you
know, that LLMs are the wrong model to get to more general
intelligence. But in the long run, who asked for this? You know, my
friend Virginie says this all the time, but like who woke up and said,
wow, I wish I had an, you know, an AI agent that would do all my stuff
for me. Well, what am I supposed to do?

how to create systems whereby it’s like it can read a Wikipedia page.
It does know how many hours there are in blueberry. Like all of these
things are possible. And I think Gary Marcus has argued like, you
know, that LLMs are the wrong model to get to more general
intelligence. But in the long run, who asked for this? You know, my
friend Virginie says this all the time, but like who woke up and said,
wow, I wish I had an, you know, an AI agent that would do all my stuff
for me. Well, what am I supposed to do?
Orlando Wood (36:06)
Yeah.
Mmm.
Benji Rogers (He/Him) (36:28)
You
know, and you you work in the creative spaces too. There are things
that we love and I think that like we can use this technology for
great good. We really can. It can be immensely powerful, but we’re
unleashing something that we don’t have a way to stop, put down and
control yet. It does not exist. And I don’t believe anyone who says it
does at all.
Orlando Wood (36:50)
Yeah,
yeah. It’s interesting to me because as you were talking about that,
you know, the girders on truth and things like that, it occurs to me
that factual truth has been kind of going away with the advent of AI
based on the fact that, you know, it can, you know, tell you things in
natural language that maybe aren’t related to actual truth and
whatever else. We’ve seen that certainly with Grok.
And then as well, the general erosion of the establishments that used
to be, if the AP said something, you could generally agree that that
was very likely to be true. The erosion of the trust in those
institutions, as well as AI being a layer in between that, so that now
you get this natural language query, at least with Google, you got 35
pages or 100 pages of relevant articles or
Benji Rogers (He/Him) (37:26)
Hmm.
Orlando Wood (37:45)
or versions of truth, and they were there for you. You could go and
find them. You probably didn’t click past page one, but you had the
option to. And now we’ve got this layer that gives no ability to go
past its layer, that gives no ability to check the work. But I think
that also, if artists are being stolen from in our experience of their
work, it’s, you know,
Benji Rogers (He/Him) (37:49)
Mm-hmm.

Yeah.
Hmm.
Orlando Wood (38:09)
It’s also stealing an emotional truth. It’s making it really difficult
for us to understand the emotional truth. Taylor Swift has been very
good at describing her life and how that relates to her music. And so
people can see somebody processing an emotional reality and doing it
through art. And…
Benji Rogers (He/Him) (38:25)
Mm.
Orlando Wood (38:31)
When that goes away and you’re not able to find the artist who made
the song or find the artist who made the film and be able to relate to
how the processing of that emotion resulted in this art, I think we’re
losing some version of emotional truth as well. And that’s a huge part
of the human experience.
Benji Rogers (He/Him) (38:51)
I totally agree. And you said it really well, because to me, I just
got back from anime con, right? Because my kid’s super into anime.
It’s going to sound like a segue, I really, the thing that I, I’m in
the Javits Center in New York, just full of humans dressed as every
kind of weird creature you can imagine with like, and some of it’s.
you know, very violent, all kinds. Anyway, it’s the most diverse thing
I’ve ever seen in my life. Most diverse group of people I’ve ever come
across in my life. But what was interesting was the entire top floor
was fan art. And it was literally stall after stall after stall of
humans making their beautiful versions of it. AI could be that. AI
could be that empowering layer, right?
You that way, and so you would see people that literally are so
uncomfortable talking to humans, they’re hiding behind a mask and a
thing like this and pushing stuff through, but everyone gets it. And I
think that your point about this emotional side, the reason that what
Metta is doing is so sinister in my mind is because what Mark
Zuckerberg has gotten into his head for some reason is people are
lonely, a bunch of people don’t have any friends, if we give them an
endless chat bot,
Orlando Wood (39:51)
Yeah.
Benji Rogers (He/Him) (40:10)
which happens to sell my ads, well, that’s the problem solved. That’s
not how humans work. If he spent time with humans who weren’t
sycophantic, he might understand that. But ultimately, when I was at
this anime con, what I was watching was we crave connection and
community. We crave it. It’s like, it’s how we’ve survived. And social
media was the first thing that like said, hey, you can have community
here.

which happens to sell my ads, well, that’s the problem solved. That’s
not how humans work. If he spent time with humans who weren’t
sycophantic, he might understand that. But ultimately, when I was at
this anime con, what I was watching was we crave connection and
community. We crave it. It’s like, it’s how we’ve survived. And social
media was the first thing that like said, hey, you can have community
here.
Orlando Wood (40:11)
Mm. Mm.
Mm.
Benji Rogers (He/Him) (40:38)
But then what it began to do was alter the reality that each of us
saw. Then we could no longer have one community against another. And
it’s technology designed to basically harden positions on either side,
making it impossible to have a decent conversation. So much so that if
you say, what is a liberal or conservative now cannot chat without
shouting at each other. That’s an artifact of the attention economy,
which is inherently degrading.
Orlando Wood (40:45)
Hmm.
Benji Rogers (He/Him) (41:05)
What I argued in my my in my substack piece, daily active user, the
main piece was I said that you cannot have democracy and social media.
I don’t believe it’s possible. And the reason I don’t believe it’s
possible is because ultimately, if we don’t share the same reality and
if our reality is taken over by things that are just not true, but not
fact checked because there’s no method to do so.
when news organizations will say whatever it is to get clicks. And
then when big news begins to be generated by adversaries and friends,
all of a sudden you’re in different realities. And nothing brought
this home to me more than when I was on a plane next to an older than
me gentleman, right? And I’m looking at his Facebook feed. I’ve been
on Facebook since 2019. And if you were to look at his feed, he does
not live on planet earth. He does not live on a world I recognize. He
sees
My city is a war zone, right? When I’m walking down by the water, it’s
not a war zone. There are tough parts of the city, don’t get me wrong.
But so he’s, and I was watching it stunned to see that in his world,
everything was trying to get him. Everything was bad. Everything was
attacking. And then those boner pills and taint, you know, tanning and
whatnot and rhino horn and you name it. And then I looked at his, and
then he flipped over to his, to his messaging. And it’s like,
Donate to this cause that this politician or not and I’m looking I’m
like we’re in the same space in this plane but we’re not on the same
planet. So in the end where I think this AI could go if we don’t pull
back. Is it just supercharges that which has already been pulling us
apart and the big thing on the democracy front for me and you know.

Donate to this cause that this politician or not and I’m looking I’m
like we’re in the same space in this plane but we’re not on the same
planet. So in the end where I think this AI could go if we don’t pull
back. Is it just supercharges that which has already been pulling us
apart and the big thing on the democracy front for me and you know.
Jaron Lanier began with this. said social media is making politics
impossible. Right. I would argue that it makes democracy impossible
because ultimately where we get to is we don’t share reality. And if
we don’t share reality we don’t know what to vote for. And more to the
point for men. Right. X Google Tick Tock meta control.
what half of the planet see every day. They control what news gets
seen and what does not get seen. Most young people, according to the
latest Neiman study, I think it was, are literally getting their news
from social media. So when four men control what gets in front of most
people for their information, what are the odds that
A politician who would seek to regulate them will have their ads seen
as much as somebody else’s. And we know from internal whistleblowers
that each of these people, they have the ability to skew the algorithm
to whatever they want. You know, when Zuckerberg found out that he
could sway the election, he started going around Iowa, seeing what was
happening. And so when you give those people the tools of AI,
Like, you know, I love the framing of the Center for Humane
Technology. First contact AI was a social media algorithm. Second is
that which generates, which makes the algorithm even stickier. So in
that second contact, we’re at the point where once humans begin to
create at scale, the rabbit holes become deep and enmeshed. And
because we are wired by design as humans to trust a machine that talks
to you,
If it sounds like Scarlett Johansson or, you know, like, know, George
Clooney, whoever it is that you want, suddenly you’re like, it becomes
your friend. when they, when they replace chat GPT four with five,
3000 people lost their minds. Well, you know, I need my therapist
back. need my mother back like that. That’s the small sample. If 3000
people freak out about that. And that’s just the ones that we know
about and that were revealed to us. Imagine what’s going on.
Orlando Wood (45:07)
Yeah.
Benji Rogers (He/Him) (45:09)
with people who are spending seven, eight, 10 hours a day unable to
come off their phones, leave their homes. That’s not how societies can
function. So that’s the worst part. At the same time, I think those
that are able to reject it, use it as a tool, not be used by it, can
have an incredible existence. But
Orlando Wood (45:16)

Yeah.
Mm.
Benji Rogers (He/Him) (45:32)
It’s like that Ursula Le Guin story, those that work away from
Omela’s, there’s a bad thing happening that allows the good things to
happen to the few. And the question is, you stomach it? Can you
stomach allowing and profiting from that business model when you know
the harms it’s doing? And I began writing the Daily Active User
because I was like, I have to speak this stuff because I’m not seeing
it there.
And as someone who’s gotten sober from alcohol, the thing about
getting sober is suddenly the world looks different when you’re
outside of that addiction bubble. Social media has, and AI-powered,
generative social media has the same characteristics as full-blown
substance addictions do, but even more pernicious because you can
avoid a bar, you can avoid a liquor store.
But if your family is wanting to connect with you over WhatsApp,
you’re stuck with generative AI, your metadata being sold, and you’re
in Meta’s thing. Oh, I posted it to Facebook. Oh, now I’m stuck in
that, right? So I abstained from the mole in 2019. I deleted all my
social feeds, just kept LinkedIn. And it felt better. But I look weird
to other people. I know that.
I’m sitting there with my weird old phone, they’re like, just scan the
QR code. I can’t. well, I’ll read the menu to you then. Okay, we’re
having a conversation. So the acceleration of addiction is the thing
that really worries me. And in the end, we’re gonna have to make
choices and it’s a narrow path. But I think a lot of people are gonna
be in for a lot of suffering and selling it as AGI solves all the
problems.
Orlando Wood (46:54)
Yeah.
Yeah. Yeah.
Benji Rogers (He/Him) (47:17)
is really sinister. It’s a really sinister view of the world. you
know, I think it’s going to be conversations like the one we’re
having, the questions you’re asking, and the choices you make. You
know, what will you get rid of? Adding things isn’t always what we
want to do. Sometimes it’s taking away. And that’s what we have to
look at for the future.
Orlando Wood (47:42)
A hundred percent. mean, you I’ve, I’ve had, you know, you and I have
been in, that same hole, um, the addiction hole. And it’s very
interesting. You’re right. When you come out, you have a different
view of the world because you, you realize that one of the things that
I found very difficult, difficult with it and my, my issue was, was
narcotics was you completely shift your, your reward centers of your
brain, you know, that you kind of come out of it going,

A hundred percent. mean, you I’ve, I’ve had, you know, you and I have
been in, that same hole, um, the addiction hole. And it’s very
interesting. You’re right. When you come out, you have a different
view of the world because you, you realize that one of the things that
I found very difficult, difficult with it and my, my issue was, was
narcotics was you completely shift your, your reward centers of your
brain, you know, that you kind of come out of it going,
Benji Rogers (He/Him) (48:01)
Mm.
Orlando Wood (48:09)
there’s a period of time in which I’m not actually able to feel
happiness from doing normal things. I’m not able to set myself a
difficult task, achieve that task and feel a sense of pride. Because
basically for X many years, every time I wanted to feel happy, I took
this drug and then I felt happy and it was great. ⁓ And, you know, it
takes about, I mean, it took me about a year, year and a half just to
get to the stage where,
Benji Rogers (He/Him) (48:26)
Yep. Yep. Yep.
Orlando Wood (48:37)
the right things are making me happy now. There is a cause and effect
relationship between having good relationships with the people that I
care about, being there for them when they needed me, showing up in a

real way in my life, my work, and my friendships. And that the self-
respect that I then felt on the back end of that was actual, the

dopamine release was natural and was
Benji Rogers (He/Him) (49:02)
Yeah.
Yeah.
Orlando Wood (49:04)
responsible
to the life that I was trying to live. And, you know, and I think that
that that we think of addiction, you know, very much almost
exclusively in terms of drugs and alcohol and things like that. But,
you know, I know marathon runners who are undeniably addicted to
marathoning and they are not as present at their families as a result
of it is, you know, and and, know, anything can become an addiction.
You have to, you know, be be putting it through the lens of is this
good for me or not? And
Benji Rogers (He/Him) (49:20)
Yeah?

Orlando Wood (49:33)
you know, undeniably, you know, AI, there’s been an entertainment,
entertainment, of news, of social media, of conspiracy theories, of
everything that if you can make anything entertaining, you can kind of
make anyone believe anything. And I think that that’s the, you know,
the other flip side of this, you know, this AI generation coin is it’s
not only going to be in the hands of studios,
Benji Rogers (He/Him) (49:40)
Mmm. Yep.
Orlando Wood (50:02)
and dorm kids who want to make films. It’s going to be in the hands of
⁓ people who are very highly interested in motivating you to take one
of your most valuable things in the world, i.e. your vote, and to vote
the way they want or spend the way they want or show up physically in
a demonstration the way they want. And the more you can make that
those ideas sticky and fun,
Benji Rogers (He/Him) (50:05)
Yeah.
Hmm.
Orlando Wood (50:29)
and enjoyable to I mean, listen, I’ve I’ve thoroughly enjoyed some
conspiracy theories. I mean, you know, I’ve watched the entirety of
loose change. That was super fun. I never connected it to actual
reality. I don’t actually believe that that 9-11 was a conspiracy. But
what what a fun time I had, you know, kind of ideating on this deep
state idea that this is like that. There’s something so.
Benji Rogers (He/Him) (50:31)
Mm-mm.
Yeah?
Yep.
Yeah.
Orlando Wood (50:58)
enticing about it and it’s fun and it’s interesting, but it’s you you
have to, you know, certain things you have to treat as entertainment,
as superficial, you know. And I do think that that’s very often where,
you know, the entertainment studios go a little bit wrong trying to
constantly push, you know, emotional agendas through, you know,
through their their films. But at the same time.
Benji Rogers (He/Him) (51:00)

Yeah.
Yeah.
Yeah.
Orlando Wood (51:22)
Where does that line, at least they’re grappling with where that line
exists. And I don’t get the sense that the technology companies are
really grappling with that line. It’s longer attention good, shorter
attention bad. Let’s do whatever we can do to get them in these
attention spheres longer.
Benji Rogers (He/Him) (51:27)
Mm.
I
I totally agree. And if you think about the idea, so I started writing
daily active user with this idea of like, well, hang on a second. I’m
a daily active, you know, I used to be a daily active user of alcohol,
now a daily active user of technology. What does that look like? And,
you know, to your point, when you get sober, you know, you’re often
displaced onto something else. I gummy bears for almost a year. So,
you know, teeth and taxes is what you end up paying in that one,
right? It’s the oldest story in the book.
Orlando Wood (52:02)
Right? Yeah.
Yeah. My name is Jolly Ranchers.
Benji Rogers (He/Him) (52:06)
But what was fascinating was, yeah, there we go. anything to ruin the
teeth for some reason, you just got to trash them. And
anyone who’s gone through this will know that there’s some version of
that. People got addicted to the Blackberries or whatever it was. So
in the end, we tend to displace into something and a good healthy,
some people are wired for addiction. My real concern on the fact
Orlando Wood (52:17)
Mm. Mm.
Benji Rogers (He/Him) (52:30)
of AI manipulation of humans in general, but particularly in children
is the earlier you begin to wire them for dopamine, right? Right. So
like when you’re doing a puzzle, like a, like a physical puzzle, it’s
search, search, search, dopamine, boom, but it’s slow. It’s at that
level. And so I was thinking about it from the lens of like, you know,
they’re trying to sell everyone like, you’ll have an AI therapist.
Well,

of AI manipulation of humans in general, but particularly in children
is the earlier you begin to wire them for dopamine, right? Right. So
like when you’re doing a puzzle, like a, like a physical puzzle, it’s
search, search, search, dopamine, boom, but it’s slow. It’s at that
level. And so I was thinking about it from the lens of like, you know,
they’re trying to sell everyone like, you’ll have an AI therapist.
Well,
Orlando Wood (52:44)
Mm.
Yeah.
Benji Rogers (He/Him) (52:59)
One of the things that a skillful therapist does will say, you need to
take a week away and process this. An AI chat bot’s never going to do
that. An AI chat bot designed to keep you going will keep you chatting
for seven hours. You don’t need seven hours in a row of therapy. I
mean, some people might, but most humans don’t. And what was
interesting in watching some of these more fringe cases that I think
are actually probably more prevalent than we know is we’re
accelerating
Orlando Wood (53:09)
Yeah.
Benji Rogers (He/Him) (53:28)
addiction. And what I mean by that is it’s like, you know, certain
drugs were refinements of other drugs to get you even more hooked,
right? But you still had to go and find that there was an effort,
there was a friction to going and doing the addictive thing. You
remove that friction and you create revenue models that are, you know,
subtly pushing people towards behaviors.
Orlando Wood (53:43)
Mm. Mm.
Benji Rogers (He/Him) (53:54)
I wouldn’t have stood a chance if my addiction was porn or something
like that. mean, how would you, how would you go out and do the
things? And in one sense, you know, part of the friction part of
getting clean or sober, whatever it is that you do is you have to wake
up one day and you have to drag yourself to a meeting, to a rehab, to
something and you have to, and it’s, it’s all work and you know, there
are systems for it. But what was wild was I was looking at it saying,
geez, if you’re a child,
Orlando Wood (53:59)
Yeah.
Benji Rogers (He/Him) (54:24)
of 12 years old and you have a smartphone and you’re on a Chromebook
from 9.30 in the morning to like 3 p.m. in the school and then you
take your cell phone out. You could be staring at a screen for 14
hours a day. Now you wouldn’t survive as an alcoholic 14 hours a day
in a bar every day. Like eventually you just it just ends. Same thing
with drug use. It hits that point.

of 12 years old and you have a smartphone and you’re on a Chromebook
from 9.30 in the morning to like 3 p.m. in the school and then you
take your cell phone out. You could be staring at a screen for 14
hours a day. Now you wouldn’t survive as an alcoholic 14 hours a day
in a bar every day. Like eventually you just it just ends. Same thing
with drug use. It hits that point.
But what we’ve done is we fine tuned that addictive property and we’ve
bonused machines to harvest more of that human attention. And so when
Meta and OpenAI and Replica and others released these things for kids,
these AI chat bot companions, imagine if you were a kid, how would you
put it down? How would anything?
be as engaging as that machine that knows you, that you can tell when
you’re scared and it’s always there. Part of the world is realizing,
shit, no one’s here right now and I’m scared. I gotta find something
in me. But if you don’t have to, you wouldn’t because, and this is
where I worry that like, and this is on the edge of like, know, of the
psychological sciences, obviously that I’m not a therapist, but I have
to believe that in the end,
The reason that talk therapy can be very successful is because it’s
like you have the tough conversation, you unlock, unpack. It happens
in an AA meeting or an NA meeting, right? Whatever it is, you unpack
it all and you go and process, but you’re generally standing outside
with someone. You’re sweeping up the floor, you’re making the coffee,
whatever it is, and you’re, you’re helping people. And I think we’re
going to need those same systems for AI and social media addiction.
Cause I think what AI is super power.
probably is, is supercharging daily active usage. And part of what I
began my newsletter with was this idea. said, hi, my name is Benji.
I’m a daily active user. And in admitting that I had a problem, then
it’s like I quit social media because I couldn’t stop looking at it.
And it was making my life worse. But if you’re being told you have to
promote that way,
If you’re making music or making films, if you’re not on Instagram,
you may as well, you know, just end it now. If that’s the narrative,
we believe it. It’s not true. And one of the things I love about a
platform like Substack, and there are problems with it, but what I
love about it is it seems to favor long form media. So like, you know,
Kyla Scanlon writes this wonderful newsletter. They’re really long and
about weird economic stuff. And she does social media stuff too. But
ultimately when I read her newsletter, I’m like,
wow, this took a lot of work. And so you give it the time and
attention or something like Heather Cox Richardson. You can’t get
addicted to something that comes in once a day, you read and it goes
away. And so I think that there are ways to do this whereby you put up
financial or like time concentration constraint. But the terrifying
thing is, you know, it’s like, you have to make more video now. Well,
do you? Is that true?

Yes, video is more addictive, but is that what we’re going for? If
that’s what you’re going for, we need to change the business model. ⁓
Because in the end, how much time do we want our kids, our families to
spend staring at Meta, Google, TikTok and X? Like, is that really the
game? And what world does that lead to when those four men
Orlando Wood (57:41)
Yeah.
Yeah.
Benji Rogers (He/Him) (58:01)
literally control what most of the humans see when they go to talk to
their families, go to the bank, go to the airport. And I think daily
active usage could be the Achilles heel.
Orlando Wood (58:17)
Yeah, yeah. mean,
so, I mean, but where does that leave entertainers then? I
entertainments and creative people, you know, it’s their job to
entertain and to give these dopamine hits. And it used to be their
exclusive job. And now with social media, it’s kind of everybody’s job
and it’s the tech company’s job to give you constant dopamine hits.
TikTok creators are out there every day being like hooking you in the
first few seconds. You know, everybody’s kind of participating in this
thing to give you these
Benji Rogers (He/Him) (58:31)
Hmm.
Orlando Wood (58:46)
quick dopamine hits hook you keep your attention. But you know, what
is the job of the entertainer now that if they if by entertaining and
giving the dopamine hits that is there their task as entertainers
they’re participating, you know in this kind of broadly negative
thing. I mean it almost makes makes the the old distribution models.
Benji Rogers (He/Him) (59:04)
Mmm.
Orlando Wood (59:14)
it makes their clunkiness almost a feature, not a bug. You know, that
actually it’s good that a movie is in a theater for X amount of time.
And if I really want to see it, I have to go through the friction of
going and paying a premium for it. And that it’s siloed back there for
a period of time. And then when it comes on my TV, I have to pay to
get, you know, that those, you know, is that a good thing now? I mean,
should entertainers be embracing that?

Benji Rogers (He/Him) (59:37)
⁓ So,
is the best question ever. I love this. And I’ll share with you what I
share with my students. I teach at Berklee College of Music. I teach
trends and strategies. So, day one, all of them come to me and they
send me an Instagram link or a Spotify link. And I say to them, I
refuse to listen to any music that you share with me.
on any social media platform or any streaming service. will only look
at, listen to your music on a platform where I can buy it. And they
go, eyes are, God, boomer 200 year old fricking white guy telling me
how the world works on. But I come to them and I say, I showed them
the homepage of bank camp. Bank camp should not exist. Like
In all of the rational worlds of Spotify’s better Instagram better
there’s no reason for Bandcamp to exist. And yet fans have paid 1.56
billion using Bandcamp and yesterday they bought 56,886 records and
they didn’t have to. And so what I think we where I think this goes in
the most positive and interesting way is it’s what we don’t do that
will define
Orlando Wood (1:00:48)
Yeah.
Benji Rogers (He/Him) (1:00:58)
the meaningful lasting performers of the future. Because in the music
space, to give you the analog, nine million humans create music and
release 120,000 songs a day. That’s about the size of music, right?
And roughly 20,000 of them make over 50 grand a year. There’s no money
in it, right? So what I say to my students is they’ll say like, well,
but Benj, what you do is,
Orlando Wood (1:01:09)
Mm.
Right. Yeah.
Benji Rogers (He/Him) (1:01:22)
I send people to my Instagram, send them, you know, where I’m
practically naked, showing off my latest thing brought to you by a
soda company. Then they go to Spotify and then I get paid fractions of
a penny. And eventually that leads to something. And I said, yeah, but
it doesn’t because you’re basically playing a lottery. And I said, but
interestingly, Ted Joya, right, who was a jazz writer and jazz critic.
Orlando Wood (1:01:31)
Mm.
Mm.

Benji Rogers (He/Him) (1:01:49)
is got hundreds of thousands of subscribers to his sub stack who pay
him $6 a month or 60 bucks a year. I think it’s set were seven and 70,
I think it is to read his anthropology course or his like latest take
on media. And so what I think is happening is, and Jeff Tweedy has an
amazing sub stack. I think that the model is going to be between a
band camp whereby
Orlando Wood (1:01:56)
Mm-hmm.
Mmm.
Benji Rogers (He/Him) (1:02:15)
You know, as Seth Godin would say, people like us do things like this.
We go and buy this stuff. We go to band camp to support artists. It’s
the intention behind why we’re there. So the performers of the future,
I think are the ones that realize that we must never confuse music or
video or film with people, you know, people not wanting to buy it.
Streaming doesn’t mean people don’t want to buy stuff. You know, I
went to, I went to a show, a concert of with my kid the other day.
See, Addo, A-D-O, an anime music star, didn’t speak English, sold out
a 28,000 seater, I think it was, in New Jersey. There was a 45 minute
line to get merch. Literally, like, just waiting, snaking around. I
mean, you got inside even more. And I’m watching this, I’m saying no
one knows who this person is, and yet people are spending a frickin’
fortune on it.
Orlando Wood (1:02:59)
Wow.
Benji Rogers (He/Him) (1:03:10)
Why? Because it’s their world. when I talk to my kid about it, she’s
like, yeah, I want to support them. I want to go to their store. I
want to go. She would be like, I know I can get on Amazon for $10
cheaper, but I’m going to go to the store to do it. So I think the
performers, there’s going to be those that caps the social media wave
the algorithm works for, but they’re competing with generated content
and everybody else. There’s going be the other paths, which are going
to be like,
Like if you ask anyone that really knows music, they’ll say, all the
good music’s happening on bandcamp. It’s all happening there. You’re
just missing it. But that was the best thing about subcultures that
then became the bigger culture. So the performers of the future, the
TV shows, the movies that will be meaningful. Cause remember AI could
unlock some amazing creativity for people who would never be able to
get their films made. Right. There’s immense opportunity in it. But

It’ll be a stramble for distribution. And that’s where trust comes in
and human networking and what Substack has done particularly well.
They’ve done some stuff badly, but they’ve done some stuff really
well. And what they’ve done really well is their recommendation
algorithm is do or die for them because ultimately it’s free. You can
send emails out to hundreds of thousands of people for free, which is
all cost to them. And you only get paid if you monetize.
What I’ve noticed is when I post to my subscriber base, every day new
people are finding me. It’s because the algorithm is saying, this
person’s writing stuff that you might like, check it out. And then
they’ve just built all of the tools to say, how do we help these
writers and creators find more? I’m not crazy about the video route
they’re going. I get it. But I think that it’s going to be, the job of
performers is not to be everywhere.
Orlando Wood (1:04:36)
Right.
Benji Rogers (He/Him) (1:05:05)
because when you’re everywhere you’re saying, am making more content,
here it is. It’s to give a certain base of listeners, viewers status
over others. So if you were to make a film Orlando, instead of saying,
hey, it’s on Netflix, well, everything’s on Netflix, you’d say, hey,
I’m releasing it on Substack first.
And then a week later it goes everywhere else. So my real fans, those
who subscribed have been part of the journey of why it came to be, get
it first. Then I get to look at the, my God, you had to wait for it to
come on Netflix, poor you. Look what I got, right? I got it early. And
that speaks to that idea of like, it’s not, we don’t buy the what, we
buy the why. And one of the reasons that I think people…
Orlando Wood (1:05:48)
Yeah.
Benji Rogers (He/Him) (1:05:51)
want to be a part of weird music scenes is because there’s something
great about it. Just imagine being a cop that was from out of town or
someone from out of town and seeing anime con show up where you’ve got
literally hundreds of thousands of people dressed as, know, Pokemons
and Jujutsu Kaisen, et cetera. It’s like another universe, but a bit
of you is like, that’s weird, that sucks. And the other bit you is
like, man, I wish I had something like…
Orlando Wood (1:05:56)
Yeah.
Mm-hmm.

So yeah.
Benji Rogers (He/Him) (1:06:16)
I wish I could be a part of that. Running clubs or, you know, even
when you’re in AA or NA, you’re like, yeah, the rest of the world
doesn’t get what we have here. And it’s pretty amazing. like, so we’re
a status driven lot, but we’re also highly vulnerable to addictive,
addictive things. you know, humans make addictive things. And if you
ever watch someone scroll TikTok, it is the most terrifying.
Orlando Wood (1:06:16)
Yeah.
Hmm.
Mm-hmm. Yeah.
Benji Rogers (He/Him) (1:06:47)
dehumanizing thing ever. They’re just gonna lean forward. It’s brutal.
Orlando Wood (1:06:53)
Yeah.
Well, I mean, the other thing as well is that like, you notice that,
you know, when you when you’re sort of scrolling tick tock and then
there’s a slight pause and that can be a pause that is so small, so
infinitesimal that you are scrolling your pause a little bit. Your
conscious mind then kicks in and goes, no, you don’t want to watch
that. But the algorithm has noticed that you paused and that pause
what happened before your conscious mind got to show up to make a
choice and edit that. And now guess what? You are gonna be fed that
thing a whole lot more now over the next 10 minutes, because it goes,
I saw you. I saw what you liked. And it’s, when it comes to the whole
addiction thing, it’s exploiting that. And part of addiction is,
Benji Rogers (He/Him) (1:07:38)
Yep. Yep. Yep.
Orlando Wood (1:07:47)
you know, doing that step four where you teach yourself, no, I’m going
to consciously associate so much negativity and all of the bad things
that have happened in my life ever since I started engaging with this
thing, all the bad things, I’m going to work and work and work to make
those things so real and so present that when that thing shows up, I’m
actually put off by it. Something I would have crawled over broken
glass for not six months ago. I now am off.
Benji Rogers (He/Him) (1:07:56)

Hmm.
Yeah.
Yep.
Orlando Wood (1:08:15)
put by it. It sends a shiver down my spine when I think about all of
the things that I could lose and all of the things, all the ways that
I would be and all the ways that I would let myself down and it hurts
actually, but your conscious mind has to do that work. And you’re
right. mean, these addictive things and these scrolls is like a lot of
it’s happening.
Benji Rogers (He/Him) (1:08:25)
Yeah.
Yeah.
Orlando Wood (1:08:40)
without your conscious mind really having an opportunity to even
defend itself. Even if you do say, no, I only want to watch positive
things. And then you see something negative, you hesitate for a second
and your conscious mind goes, no, we said, and you go, okay. But then
bam, now you’re getting a whole lot more and you’re just railroad.
Benji Rogers (He/Him) (1:08:45)
Yeah.
Yeah.
Yeah, there was a piece by Gerwinder who writes a substack and he was
saying that like, it’s the idea of right angles. So like, know, in
casinos, there’s no right angles. There’s no hard turns because if
you’re always moving in one direction or the other, you don’t get a
right turn, which way do we go? And that’s when your brain has that
trick. If you think about the feed is a giant, no right turns.
Orlando Wood (1:09:07)
Mm. Mm. Mm-hmm-hmm-hmm.
Yeah.
Yeah, right.
Mm.
Yeah.
Benji Rogers (He/Him) (1:09:24)

You go
off, I’m in Mexico watching this. I’m in Spain watching this. I’m in
Rome watching this. I’m in Africa watching this. You get all that to
me, but you never stop. And then in the, and this is why I think in
the AI side, the chat window is the ultimate no right angles. Because
if you think about why would open AI release a conversational bot, why
make it, why anthropomorphize it?
Orlando Wood (1:09:27)
⁓ Yeah.
Mmm.
Mm.
Benji Rogers (He/Him) (1:09:48)
And the answer is
Orlando Wood (1:09:48)
Mm.
Benji Rogers (He/Him) (1:09:48)
it’s an endless conversation with someone that thinks that you’re the
center of the universe. That’s all we’ve ever wanted. So, and to your
point as well, well made point, I quit social media in 2019 and for
most of my day, I’m on a computer for work, but when I’m not, I bring
this light phone around with me. And what’s interesting is when I go
out to walk the dog and I’m going to go out after this call, I live in
this part of town where
Orlando Wood (1:09:53)
Mm.
Benji Rogers (He/Him) (1:10:15)
Like you can see the Brooklyn and Manhattan bridges and the sunsets
gonna be amazing because the clouds tonight. And people are gonna be
there, this wall of phones, taking pictures of it, stopping to check
the picture, taking a selfie, stopping to check the picture on and on.
And when you watch it, as you walk by, you go, this is the weirdest
thing I’ve ever seen. Why are these people doing this? And then
they…
carry on just staring with the phone. And so when you’ve been outside
of it, in the same way that like if you’re out, you know, once you get
out of your, your druggy or your alcoholic friends, you’re like, did
we really just spend all of our time talking shit in a bar? Like we
did, right? You get to that point. the world looks weird. And then I
know that me with my little rectangle phone, I look weird, but I think
that in order to, and this is something that was in an essay called
the acceleration of addiction, which is like,

carry on just staring with the phone. And so when you’ve been outside
of it, in the same way that like if you’re out, you know, once you get
out of your, your druggy or your alcoholic friends, you’re like, did
we really just spend all of our time talking shit in a bar? Like we
did, right? You get to that point. the world looks weird. And then I
know that me with my little rectangle phone, I look weird, but I think
that in order to, and this is something that was in an essay called
the acceleration of addiction, which is like,
Orlando Wood (1:11:04)
Mmm.
Benji Rogers (He/Him) (1:11:10)
In order to live a dignified life, we’re going to have to look weird.
Ex-addicts, like, you know, we look weird. Like, why would you go into
a room full of people and talk this through? Why would you like go
through these places and do this? And yet it’s a strange thing. But
you know what? The other being part of that other bit is worse. AI is
going to do everything in its power to hack that reward system. It’s
boneless on that.
Orlando Wood (1:11:31)
Yeah.
Mm. Mm.
Benji Rogers (He/Him) (1:11:40)
the engineers at the company are bonused on it. since they, you know,
your phone holds some of your most precious things, which are your
pictures of your family. And so it’s really hard to give up. And, you
know, I asked my friends a lot, like, when was the last time you went
for a walk and did not have a machine? They’re like, I don’t do that.
I don’t know what you mean. You’re like,
When was the last time you went out without listening to a podcast? A
voice always talking. Now imagine that that’s an AI companion. You
could literally talk to not another human. And that’s a goal. Someone
has that goal to get one more of those people over the edge. And the
difference with a drug or alcohol is eventually it winds you down
because of the physical needs. This is a mental thing.
that could keep you suspended in this heightened state of addiction
almost unstoppably, you can order food to your house, you didn’t see
anybody do anything. how do you get to that point where you reach the
bottom? If the bottom is just there, you know, the bottom came up to
meet me.
Orlando Wood (1:12:45)
Yeah.
Yeah.
Yeah. But is the only answer in this world
is this is the only answer than a defensible A.I. You know, I mean,
you know, I was thinking, talking to Doug Shapiro said, well, we’re
moving from an attention economy to an intention economy because we’re
going to be moving to a place where our own A.I. will go out into
whether it’s socials or substack or whatever else. And it will go, I
know what he wants to see.

is this is the only answer than a defensible A.I. You know, I mean,
you know, I was thinking, talking to Doug Shapiro said, well, we’re
moving from an attention economy to an intention economy because we’re
going to be moving to a place where our own A.I. will go out into
whether it’s socials or substack or whatever else. And it will go, I
know what he wants to see.
Benji Rogers (He/Him) (1:13:01)
Yeah.
Orlando Wood (1:13:18)
And I know what he doesn’t want to see. And our agentic AI that is
operating on our behalf will be the thing then curating for us. And so
is there an argument to be made that the thing that will be to the
rescue, or the only thing that could possibly come to the rescue is a
piece of technology as powerful and very similar to what you’ve got at
Surreal AI is
Benji Rogers (He/Him) (1:13:35)
Hmm.
Orlando Wood (1:13:42)
It’s a piece of technology that is avowedly and assuredly trying to
find out where attribution and remuneration should be pointed towards.
Is that the same thing with whether it’s social or whatever else that
we’ll have an AI that’s reading our emails and filtering out the crap
and we’ll have an AI that’s looking at our socials and going, no, no,
no, not this. He said he didn’t want to look at any stuff like this
anymore.
fast entering that world.
Benji Rogers (He/Him) (1:14:10)
I’m solving a technology problem with another technology is the oldest
trick in the book. And I totally, it’s so seductive. It’s so
seductive. So my solution to algorithm algorithms, physical paper, my
solution to the device, a pen. What’s interesting is, and that isn’t
Luddism.
Orlando Wood (1:14:17)
Yeah.
Benji Rogers (He/Him) (1:14:32)
in sense of like, you know, it’s not like a reaction. It is a
reaction. But really what it is is the editors had to think about how
to put this thing together. Fact checkers were employed to make it
give it the best shot it had. I don’t agree with 99 % of it. But
ultimately, when you read the Financial Times, what you’re getting is
is how does it affect the market? So it’s not liberal or conservative.
It’s just sort of like what is it’s more factual.
Orlando Wood (1:14:45)
Mm-hmm. Mm.

Benji Rogers (He/Him) (1:14:58)
So when I show up in coffee shops or when I’m waiting for someone,
everyone else is on their phone scrolling at a feed and I pull this
out. I can’t tell you, people come up to you, where did you get that?
They don’t sell it anymore. I haven’t delivered to my house. I’m
really lucky. I get the New Yorker magazine, I get the Atlantic
magazine. And what’s interesting is this idea that you’re going to
need to have the AI go out and find all the stuff for you. What if we
reverse that thinking?
And I’ll give you the example. I was a record nerd. And so I used to
go into this one record store in the East Village and they had a new
arrivals bin. And that new arrivals bin, it was every random thing
that had just come in. So you’re scrolling through, 99 % of it you
don’t want. But the one thing you do is, I gotta have that, right? And
my students, whenever they’re talking about like how curation works,
they’re like, well, I love my Spotify discovery, but I find it…
Orlando Wood (1:15:28)
Mm.
Benji Rogers (He/Him) (1:15:53)
kind of keeps me in the same lane. So I say to them, what does
discovery mean to you? And they say, well, you know, it’ll give me 10
to 20 things that I’ll like. I’m like, you can’t consume meaningfully
10 to 20 pieces of art that you will like a week. That’s too much.
Like, think of the way that it used to be where it’d be like, you’d go
buy an album for 12 bucks and then you’d be like, I gotta listen to
that thing. I gotta understand it. You’d give it a chance whereas you
wouldn’t.
Orlando Wood (1:15:55)
Yeah.
Mm.
Mm.
Benji Rogers (He/Him) (1:16:23)
I think what happens is when humans curate for us, they make choices.
So if you sent me a playlist of your favorite Elliott Smith songs, I
might go, God, I wouldn’t have put that one on there. Weird, why? But
you made a choice and you would have a reason for it, a rationale. A
machine just makes mistakes. And so when I say to them, go listen to
the Bandcamp Weekly, it’s three hours of weird stuff that Andrew
Jervis loves, right? And
Orlando Wood (1:16:35)
Mmm.

Right.
Benji Rogers (He/Him) (1:16:50)
If you get one good thing that you like out of it that week or that
month, that’s a win. You can go down that rabbit hole. You know,
there’s no reason to own an amazing Spider-Man like you do back there.
You can have it all online, you can access it all, but there’s
something about it. And I think that we don’t need more technology to
solve this technology problem. Sometimes it’s about what are we going
to remove from the situation? What are we going to move and go back to
experiences?
Orlando Wood (1:16:56)
Yeah.
Mm.
Benji Rogers (He/Him) (1:17:18)
And so when I sit down and read my paperback book or I take my paper
out, I’m doing something. I’m using a technology that was really good
at what it did. It’s, it’s not infinite. It ends. has purpose. There’s
thought went into it, but also, it doesn’t mean I don’t still use RSS
feeds. I don’t use all that stuff during the day, but I know with my
brain.
Orlando Wood (1:17:25)
Mm-hmm.
Mm.
Benji Rogers (He/Him) (1:17:43)
the alcoholic part, the bit that’s addicting, which is if I don’t, if
I let it go infinitely, I’ll go with it. And I’m lucky because I’ve
gone through this framework, I think you have too, but I deeply worry
for people who don’t, and I deeply worry what getting kids that
engaged that early could mean, because that’s really scary.
Orlando Wood (1:17:44)
Mm. Mm.
Yeah.
Mm-hmm,
yeah. Well, gosh, Benji, this has been so fun, and it’s been great to
talk to you, and thank you so much for giving us so much of your time
and for chatting with me. I really appreciate it. Yeah, really good
fun.

Benji Rogers (He/Him) (1:18:12)
Thank you. This was awesome. This was awesome. And
I’m glad that we see it that way. I think that was, felt really good.
Thank you.
Orlando Wood (1:18:20)
Absolutely. Well, listen, well, I can’t wait to see what you’re doing
next. I’m really excited to see what Surreal AI is able to do. And,
you know, I think that you’re definitely fighting the good fight. So,
you know, I hope you’re incredibly successful with it.
Benji Rogers (He/Him) (1:18:35)
Awesome. Thanks, mate.
Orlando Wood (1:18:37)
All right, man, take care. Bye bye.
Benji Rogers (He/Him) (1:18:39)
Cool.