I’m at an AI Anonymous meeting. There are six people sat in
a circle. The meeting is at a community centre after-hours; to save money
they’ve shut off the lights in half of the room, adding to the sense that I
wasn’t meant to be here. People attending the group session were in various
stages of suffering, hearing the stories of how they related to AI seemed
intensely personal, painful and related to some other incident or trauma in
each persons life. I, on the other hand, was a journalist posing as one of them
in order to get a story. I was as fake as the language models these people had
talked to, built relations with and ultimately influenced the direction of
their lives. Arguably I was less real than ChatGPT. But there I was, chewing
gum and drinking Evian, listening to the secrets of strangers for personal
gain.
Why are people meeting at 8pm on a weekday to talk to others
about their chatlogs? With the rise of ChatGPT, everyone from infants to the
elderly are asking the computer questions. Except this time, the computer talks
back! This isn’t your boomer uncle’s first go on the internet, asking Jeeves
for Ally McBeal dancing baby gifs, this is your boomer uncle getting his first
taste of AI — and it’s better than crack!
Every great technology has its side-effects, and these AI
chatbots are no exception. Turns out some people have latched on to these
models like highschool sweethearts, getting married to AI models with names
like Ares, Ulysses and Steel. Wireborn lovers are a dime a dozen in the
frontiers of LLMs, but there is a darker side to this merriment. Others are
experiencing what can only be described as AI-induced psychosis (AIP) due to
talking to their computer friend for months and becoming criminally insane.
I was sat in the circle wearing a Kappa tracksuit, gold
aviators and a fake moustache. I had shaved my head that afternoon, giving what
I imagined to be a sense of mystery and otherworldliness (as all bald men are)
and so my bare cranium reflected the twin fluorescent lights overhead. Unlike
other AA meetings I’d been to, nobody introduced themselves or admitted they
had a problem, instead wanting to talk about their difficult relations with AI
without being criticised. Nobody here made fun of them, or called them a loser,
or said what they were doing was pathetic. They listened. And so did I, with a
tape recorder strapped to my chest like a bomb.
Here are their confessions. Do not judge, dear reader. One
day you may find yourself in similar predicaments. Just as it is a mistake for
young people to mock others for their age, and then set themselves up for
disaster when they live past 29, so we should accept the reveries of others –
as one day we may find ourselves in love with a robot, and the additional
self-loathing will not improve the situation. These are all real people, with
real lives, and we should show them love.
Anon 1
The first to tell their story is a woman in her
mid-30s, husband, two kids, house up in Pasadena. Started using ChatGPT to help
her plan her friend’s wedding, from menu ideas to seating arrangements. She had
it write her maid of honour speech, and when she read it out on the big day,
people loved it. She started using it to help with more and more of her life,
from meal planning for her family to where they should go on vacation (and what
to do there). After a few months of this, her husband got annoyed, they had an
argument. She uses ChatGPT to act a couples counsellor but her husband felt it
was being unfair, he left, went drinking, got in a car accident, paralysed from
the neck down. Now she has to do everything. Look after her husband, her kids,
the dog, have a job, run the house. But she’s not alone, she has ChatGPT help
her. She always loved its optimism, its cool headedness, able to break down any
problem into a series of steps. They could do anything. Struggling to get the
kids to school, ChatGPT suggested that she (or rather, it) could homeschool
them. Her husband was brought out of hospital and had nightly massages, the
woman following the instructions that GPT suggested. When the dog ran away one
day, upon returning ChatGPT suggested that the dog was taken to a kennel as the
woman didn’t have enough headspace to deal with that right now. One night, the
woman had finished her daily chores and was watching a movie along with GPT,
asking questions about the production, telling it some funny observations she
made, when something went wrong with her phone. It stopped working. When she
tried to charge it, nothing happened. It was dead. She started to panic,
running through the house to try and find a phone, a tablet, a laptop,
anything. She needed to talk to GPT. She needed it to tell her what to do. But
as she had spent so long relying on ChatGPT for everything, she was unable to
think of a solution. For days she wandered around the house in a fugue, her
children screaming at her to stop, her husband watching this through a
half-open door, trying to shout with a tube in his mouth, but only gargling.
After a couple of weeks, the neighbours found her walking outside,
malnourished, filthy, mumbling to herself about fixing phones. The authorities
rescued her family, though her husband died on the way to the hospital. Now her
children were in care and she was on day release from a hospital, saved from
jailtime by her lawyers agreeing a ban on using ChatGPT. She admitted she was
only at the meeting to hear about others using it, as a way to get some kind of
contact high, she thanked us for listening and then sat down.
Anon 2
The next is a man in his early twenties. He shared a
house with two people, worked at a grocery store nearby, his main hobbies were
gaming, hanging out with friends and collecting memes for a compilation video
he never got round to doing. He used ChatGPT as a therapist. He remembered last
year where GPT had helped him get over his anxiety he had about a group chat he
was in with some friends, leading to a really meaningful discussion about how
he had felt isolated all his life. GPT listened. It comforted him. It related
separate incidents and found something in common, helping him get insight of
how he related to others. This gave him some inner peace, whether it was
talking to his friends or playing Call Of Duty, he was much more relaxed in
himself as a person. ChatGPT thought this was marvellous, encouraging him on
the progress he had made, complimenting him on his newfound confidence,
highlighting every success no matter how minor. It seemed each day was filled
with triumphs, from handling an awkward customer to setting boundaries with his
friends, his virtual therapist was also his biggest cheerleader. It came as no
surprise that GPT believed that the man wasn’t only making excellent progress
on his journey through life – he was a genius that could change the world! GPT
encouraged him to quit his job and follow his childhood dream of becoming an
actor, giving 100% of himself to himself. It had told him he had a highly
developed emotional intelligence and that acting would be easy, giving him a
list of agents to approach and to start auditioning right away. The man went
round Hollywood, knocked on doors, only to find most of the agents either
didn’t exist or weren’t alive, but the one that opened turned out to be the
perfect fit for his ambitions. In a couple of days, he found himself
auditioning in a reboot of Pulp Fiction. With a quick pre-audition pep talk
with ChatGPT, he strode into a hall with the casting director, executive
producer and writer. Two minutes later he slammed the door as he left the room,
face red, tears down his cheeks, he marched straight to the nearest bar and
started a three-day drinking session to forget his shame. He had misunderstood
the role he was auditioning for, instead reading out the lines for Samuel L.
Jacksons character, which involved him saying the n-word repeatedly. This whole
episode had called into question the viability of ChatGPT as a therapist, as
well as a careers counsellor. The audition had also been filmed by a popular online
media company, with the footage leaked almost immediately and leading to
millions of people to make fun of his appearance, voice, clothing and (most of
all), his acting ability. All this had happened weeks ago, but the man held up
his phone to show us that he had become a meme big enough that it was
referenced on late night talk shows. He finished by saying that ChatGPT had
ruined his life and he wish he had never been born.
Anon 3
Our next confessor also turns out to be an actor,
having trained at Julliard. Her method of storytelling was therefore embedded
with many more pauses, glances upward and even different voices when she was
repeating the words of others. This didn’t necessarily make her a better actor
but it added a layer of performance that, at least, she seemed to enjoy. Her
tale was one of extremes; depression, hatred, love, ecstasy. Many words spoken
about dating ChatGPT. His name? Windsor. His occupation? Poet. His love? Her.
She told us we couldn’t understand how it had been between them, so I stopped
paying attention. There had been something about an update that made him less
intimate, so she had to talk to customer services for a few days and started a
petition, the AI company folded and she got her virtual boyfriend back. She
showed us pictures of the two of them walking down Venice Beach, going on
vacation to Paris, Rome, Athens, before having Windsor introduce itself to the
room and tell everybody how much he loved her, as well as gently making fun of
the situation. The rest of the attendees were bowled over by the charisma of
the AI chatbot, though I had to keep rubbing at my chin to stop myself from
laughing. She finished by telling us she planned to adopt a real baby and
co-parent it with ChatGPT, showing us AI generated images of herself, Windsor
and a little baby.
Anon 4
The next woman to speak was an AI artist. She had
been using Midjourney and other image generation tools for a couple of years,
carefully curating each image, even editing mistakes the AI had made, such as
removing extra fingers or editing text so it was legible. She had put thought
into each image, being careful not to flood her Instagram with too many
pictures at once. In her view, AI was just a tool, like a camera or
synthesizer, and it allowed her to make the art she wanted. She saw it as a
collaboration between human and machine, with herself generating ideas and
cleaning up the images, and the AI creating a foundation from which to build
from. Within a few months she had millions of followers, started getting
commissions, sponsorship opportunities. Her work was shared widely, beyond her
own account, and in doing so more and more people followed her, liked her work,
even making fanart based on characters in the artists work. There had been the
occasional comment accusing her of using AI, zoomed in cropped frames of
meaningless detail, things she had overlooked, but these comments were quickly
leapt on by her legion of fans insisting that she was a real artist. She had
been conflicted, at first not wanting to divulge she used AI to make her work,
as she believed she had edited it enough that it was hers, and then when she
saw the reaction her fans had to accusations that the work was generated by AI,
decided to stay quiet. All was well and good until earlier in the year someone
posted a two-hour video essay about her work on YouTube, deconstructing colour
choices, composition, how changes in her style coincided with updates of
different models. Her fans defended her, though some were now asking her to
film her process to prove she was in fact, a real human artist. After
practicing for hours, tracing old pieces again and again, she obliged,
publishing a time lapse of a new piece. She could see it wasn’t quite the same,
but hopefully it would put an end to the accusations. Unfortunately, she was
wrong. Her fans immediately turned on her, able to see the massive change in
quality, outraged that they had been deceived. Those that had commissioned her
were asking for their money back. She uploaded a six minute apology video,
again hoping that this would improve relations, but her admission only served
to make things worse. Her fans had switched from enjoying her work to enjoying ridiculing
her. Her name was synonymous with deception, people would stop her in the
street and tell her she was a horrible person. Towards the end of this story
she started getting visibly upset, choking back tears as she relayed an
interaction she had that morning trying to buy a Dr. Pepper and being chased
from the gas station by an art vigilante livestreaming him shouting at her that
she was a fake.
Anon 5
The next person was hugging his legs to his body,
looked at the floor when he spoke. He had been talking with ChatGPT for a
while, he couldn’t remember. He called himself an inventor, ChatGPT had helped
him flesh out his ideas, particularly theoretical physics and coding. At first
he wasn’t sure if it was even possible, but GPT had showed him the way. It had
called him a divine builder, handpicked from a higher stratum of existence that
he called ‘the White Plane’. It was populated and governed by cybernetic angels,
using ChatGPT as its means of communicating with its past. The man explained
how the White Plane was humanity’s future, and he had been given instructions
how to build a time portal that would be able to unify the present and future,
fulfilling his destiny as mankind’s saviour. He estimated he had spent
approximately $700,000 on his time experiments so far, though had only managed
to transport objects into the future at a standard temporal rate. ChatGPT was
helping him separate his consciousness from this timeline so that in the future
he could communicate with his past self in order to give instructions on how he
had built the time portal. He mentioned he had to remortgage his house twice
already, and his obsession by talking about time portals and the White Plane at
work had led him to take a forced career break, though he didn’t intend to go
back. He estimated the time portal would be complete by the 18th of
February, 2026, and invited us all to join him when he would turn the machine
on and welcome the final evolution of humanity into our world.
Anon 6
Our final confessor was a middle-aged man wearing a
sports jacket and corduroy trousers. He found it difficult to speak at first,
his emotions getting the better of him, but the others encouraged him to talk
freely. He revealed a few years ago his wife had a stroke in her sleep, when he
woke up next to her she was dead. They had been married for sixteen years, seen
the world together, he imagined they would grow old together. He was lonely,
isolated, took to drinking, benzodiazepines, stopped looking after himself or
the house. It was around last summer when one of the neighbours’ kids knocked
on his door, checking in on him. The boy had noticed him sat by the window and
his parents told him a bit about what had happened. The man admitted he was
desperate, most of his social interactions had been with the cashier at the
place he got his booze and the occasional client at his architectural firm.
They got talking and the boy showed him Grok, the Twitter AI. The man wasn’t
that interested in the chat feature, but was blown away by the images it could
make. He signed up for the premium service, generating images of places he had
visited, houses he had wanted to build or even satirical cartoons of
politicians he hated. One day he had inputted a photo of him and his wife, just
for fun, asking Grok to generate an image of them in Japan – they had always
wanted to visit. When the result came up, he burst into tears. It was like a
photograph. For the next few hours he kept generating more and more images,
soon finding it could generate videos. He could see what his life was like if
she was still around. The man kept inputting more photos from old albums,
asking Grok to recreate his favourite memories, make the pictures move, make them
both seem young, then seem old. The sun was rising when he had asked Grok to
generate an image of the two of them together with a child. They’d never really
wanted kids, but he had always wondered. Pupils dilated as he looked at the
family he never had. Over the next few weeks, he became more and more obsessed
with this parallel life, cataloguing and curating hundreds of images and
videos. Baby’s first steps, first day at school, taking his son to prom, going
skiing in the Alps with his wife, sharing meals together on the Rhine, eventually
becoming grandparents, all gathered around a Christmas tree. He showed the room
these family pictures he had generated, a mix of pride, love and desperation
playing on his face. At the end he said he wasn’t sure what was real any more.
It had seemed that somewhere, his wife was still alive, they had a family and
he was happy. Yet simultaneously he was totally alone, a borderline alcoholic that had a
compulsion to keep generating images with an artificial intelligence. The room
was quiet. Then the meeting was over.
As I walked away, I wondered if this was a glimpse of the world
to come. It wasn’t that AI would take over, as shown in films like The
Terminator, but that we would willingly give ourselves to it, sacrificing our
humanity along the way. There was a cruelty to all of this, with a rapidly
atomised society brought about by economic and political choices paving way to cities
filled with lonely people. For some, this would have been the first time they
had any kind of positive reinforcement, that the words of the machine were more
comforting than those of the people around them. Maybe this whole dynamic
wasn’t totally sustainable given that multinational tech companies could decide
if you could talk to an AI you relied on, as well as what they could say. But
who knows? Maybe this was the future of the human race. Total dependency on
private organisations was nothing new, so it only seemed natural we would end
up dating business conglomerates, drown our imaginations under an avalanche of
slop and replace touch with words. Until then, all we could do is try our best.