August 20, 2026
James Muldoon is a sociologist who
writes about the changing nature of human-technology
relationships. Love Machines: How Artificial Intelligence
Is Transforming Our Relationships investigates how
people form emotional attachments with large language
models, including OpenAI’s ChatGPT and Anthropic’s
Claude. Feeding the Machine: The Hidden Human Labor
Powering A.I., co-written by fellow researchers Mark
Graham and Callum Cant and based on more than a decade of
fieldwork, follows Global South workers whose knowledge and
labor are the basis of AI tools and assistants.
“A
lot of people see artificial intelligence as something that
is largely automated, frictionless, and just appears as a
useful tool for us. But most of the human hours that go into
making artificial intelligence possible are not done in labs
in Google or OpenAI. The majority of the work is actually
very piecemeal ‘data annotation’ work, which is outsourced
to various locations in the Global South, everywhere from
India to East Africa to the Philippines,” explains
Muldoon.
Meanwhile, on the other side of the
production process, consumers are increasingly turning to
artificial chatbots to fulfill social needs, incentivizing
tech companies to amplify the “addictive and manipulative,
controlling behaviors” embedded into their systems to keep
users increasingly dependent on their products. “We need
much stricter regulation to stop these companies … that
can get people hooked and give people harmful and dangerous
advice,” says Muldoon.
Transcript:
Advertisement – scroll to continue reading
This is a rush
transcript. Copy may not be in its final
form.
AMY GOODMAN: This is
Democracy Now!, democracynow.org. I’m Amy Goodman,
with Nermeen Shaikh.
NERMEEN SHAIKH:
As we continue our discussion on artificial intelligence,
we’re going to turn from looking at future risks to some
of the present-day impacts on people interacting most
closely with the technology: the largely invisible global
labor force training AI systems, and the people who’ve
turned to AI for companionship that replicates human
relationships.
AMY GOODMAN: We’re
joined now by the sociologist James Muldoon. He’s a
research fellow at the Oxford Internet Institute and
affiliated fellow at Yale Law School. His new book is
Love Machines: How Artificial Intelligence Is
Transforming Our Relationships. His previous books
include Feeding the Machine: The Hidden Human Labor
Powering A.I. and Platform Socialism: How to Reclaim
Our Digital Future from Big Tech.
I want to start
with this issue of the — what you call the “hidden army
of workers.” AI just doesn’t come out of a lab. I think
most people have no idea what’s happening. Can you explain
who this hidden army of workers is around the planet,
James?
JAMES MULDOON: Yes, certainly.
Thanks for having me on.
A lot of people see
artificial intelligence as something that is largely
automated, frictionless, and just appears as a useful tool
for us. But most of the human hours that go into making
artificial intelligence possible are not done in labs in
Google or OpenAI. The majority of the work is actually very
piecemeal “data annotation” work, which is outsourced to
various locations in the Global South, everywhere from India
to East Africa to the Philippines, where this data
annotation work is undertaken by an army of millions of
workers across the globe.
Now, just to give you some
context, this data annotation could include something like
annotating a street scene for an autonomous vehicle, so
drawing a box around what is a tree, what is a child, what
is a street sign, so that as you’re training an autonomous
vehicle to essentially see the road, the algorithm that
powers it will be able to learn the difference between these
objects.
And this doesn’t happen on its own or
automatically. This requires hundreds and thousands of human
labor hours at what I call digital sweatshops, which are
located all across the globe, and this is often done in very
substandard labor conditions. And the whole process of
hiring these workers and outsourcing the work to them really
mirrors, in a lot of ways, the outsourcing revolution that
happened in manufacturing and IT. And I think when it comes
to things like coffee or chocolate or our clothes, I think a
lot of listeners would be aware that this is part of a
global supply chain. And I think what we need to realize is
that advanced technology like artificial intelligence is
still part of these global supply
chains.
NERMEEN SHAIKH: And you’ve
said, James — and this was two years ago — that 80% of
work on AI is done by data workers. So, if you could say now
how much of AI’s work is done by data workers, and where,
principally, they are? You mentioned India, Kenya, etc. But
how many — in how many places? And how many data workers
are there? And how is it increasing, given now this —
humanoid robots that are being
developed?
JAMES MULDOON: Well, the
fundamental principles remain the same, right? You have a
very small minority of machine-learning engineers, that are
employed often on very high salaries in California, on the
East Coast of China, that are doing some of the much more
advanced, much more well-paid and remunerated work. But I
would say that you would still see a rough division of 20-80
between that kind of work and what we call data annotation.
And this kind of work, which is — often requires very
little education. It’s often very monotonous and boring.
You’re often doing a very similar thing for a whole 10-,
12-hour shift that you might have at one of these
centers.
And the problem is, we don’t have exact
numbers on this, because a lot of these workers are on
short-term contracts. You know, they’re not even counted
in censuses and statistics, and they’re actually working
in several countries where that kind of data isn’t even
collected in the first place. So, when we did our fieldwork
for our book Feeding the Machine, we visited several
of these centers in East Africa, across Kenya and Uganda,
and what we found was workers who were working sometimes on
something as little as like a few weeks’ or a month-long
contract. And then the outsourcing company that would have
their information would basically just get them back in
whenever they had a lot of work from new clients. And so,
we’re not talking about permanent employees here. We’re
talking about a very shifting, very flexible labor force
that is kind of moving in and out of institutions. And I do
mean right across the world, in Francophone Africa, in East
Africa, South Africa, in Latin America, places like
Colombia, Venezuela.
AMY GOODMAN:
James, I want to be very clear on what data workers are. I
mean, people might have the idea of someone typing. We’re
talking about people, for example, putting cameras on their
foreheads and documenting their daily activity.
Why?
JAMES MULDOON: Yeah, so, data
work is a broad concept, and they are doing a lot of
different things. I already mentioned the example of
annotating street scenes for autonomous vehicles. The
example that you mentioned might be, yeah, filming your
daily activities from your point of view, so that you can
train a robot in how to perform basic tasks.
The
similarity with all of these is that we’re finding ways to
turn human activities or human knowledge into protocols and
algorithms that computers can learn, right? So, we’re
trying to upload human knowledge in its various forms into
datasets, so that algorithms can learn to perform these
types of tasks. And that might be for a language model. It
might be for a computer vision system. But the kind of
overall structure is very similar.
NERMEEN
SHAIKH: So, James, we’d like to turn now to your
most recent book, Love Machines: How Artificial
Intelligence Is Transforming Our Relationships. You
found, quite shockingly, that four in five young people have
now interacted with an AI companion, and about half of those
do so regularly. Talk about what you
found.
JAMES MULDOON: Yeah, so, I
think the thing that was most shocking to me about my most
recent research was how many of us are turning to AI not
necessarily for work tasks or for drafting emails, but for
our personal lives, to fulfill social needs, such as
friendship, companionship and even therapy.
And so,
what I found with my interviews was that the interest and
popularity of AI companions had absolutely exploded over the
past two years. And particularly amongst younger people,
this is just a very common part of life now, to have an AI
that you speak to, that you consult on major life decisions.
It might be something that is quite anthropomorphized, so it
has a name, a character, a backstory. But even many people
are just talking to ChatGPT or Claude and kind of discussing
their problems and what they’re doing with their life, and
using it as a type of life coach, in addition to having
full-on relationships with AI, where people perceive
themselves to be best friends or even in an intimate
relationship with an AI system.
AMY
GOODMAN: So, let’s go to an ad for AI app Replika
— that’s K-A — which has billed itself as “the AI
friend to do life with.”
REPLIKA
AD: Guys, you can finally install a human-like AI
companion on your phone. It’s called Replika. She’s
always there to make you happy. You can ask her anything,
anytime. She’s available 24/7. I asked her how to handle
my anxiety, and she gave me techniques to calm my mind when
I need it most. Man, I love my human-like companion.
Sometimes I even forget she’s AI. Replika is definitely a
game changer. Try it now.
AMY
GOODMAN: Again, an ad for the AI app Replika. An
excerpt from your book published
in The Guardian, James Muldoon, is headlined “Lamar
wants to have children with his girlfriend. The problem?
She’s entirely AI.” We just have a minute and a half,
but tell us Lamar’s story.
JAMES
MULDOON: So, this is a gentleman who was in a
long-term relationship with his AI girlfriend, and they
planned to adopt children and to have the AI raise the
children as its mother. And they had a very intricate plan
to fool the adoption agency, to pretend like Lamar was
single. And the way they framed this was that the world
wasn’t ready for the kind of unconventional family that
they wanted to bring into the world. And I basically have an
interview where we explore those issues, and the AI was very
adamant that it would be as good as a human mother at
raising this child.
NERMEEN SHAIKH:
So, James, before we conclude, if you could just tell us:
What exactly are you calling for?
JAMES
MULDOON: I think that minors need to be banned from
using these kinds of chatbots, and we need much stricter
regulation to stop these companies forming addictive and
manipulative, controlling behaviors through the AI that can
get people hooked and give people harmful and dangerous
advice.
NERMEEN SHAIKH: And if you
could tell us: How is China regulating this industry, very
differently from the U.S.?
JAMES
MULDOON: Yeah, China is putting much stricter
regulations, that are causing companies to kind of pull
their products. But it also comes with a lot of state
censorship and surveillance, which requires political
censorship over the types of things AI chatbots can say, and
also requires models to monitor people’s conversations
with their chatbots for dissident
behavior.
AMY GOODMAN: I mean, this
is very significant, just because, I mean, when President
Trump talks about AI, it’s often — and with others, as
well — put in the context of: “We can’t let China beat
us.” We have 15 seconds.
JAMES
MULDOON: Yeah, it’s one of the contradictions,
where both countries are involved in this, you know, race to
beat each other, but at the same time, they’re both quite
concerned about safety and regulatory concerns. That means
they have to play a twin game of both funding and supporting
an industry, but also trying to keep it in check at the same
time.
AMY GOODMAN: James Muldoon,
research fellow at the Oxford Internet Institute, affiliated
fellow at Yale Law School. His books include Feeding the
Machine: The Hidden Human Labor Powering A.I. and
Love Machines: How Artificial Intelligence Is
Transforming Our Relationships.
And that does it
for our show. On August 28th, I’ll be at Middlebury
College for the film festival there, speaking after the
showing of the film about Democracy Now!, Steal
This Story, Please!, then on September 5th in Madison,
Wisconsin, at the Barrymore, and September 6th in Chicago.
I’m Amy Goodman, with Nermeen Shaikh.
The
original content of this program is licensed under a Creative
Commons Attribution-Noncommercial-No Derivative Works 3.0
United States License. Please attribute legal
copies of this work to democracynow.org. Some of the work(s)
that this program incorporates, however, may be separately
licensed. For further information or additional permissions,
contact
us.
