> We began a heated affair. A contractor hit a pipe above us, so rancid mayo splattered onto our bodies as we embraced, but we were too absorbed in our passion to mind. Try doing that from home.
I can't remember the last time something I read on the internet got a laugh that hard.
It takes the structure, the affair with Julia, the "we have always been at war with", best buddy/big brother, and the forced rehabilitation from 1984. This is also part of the joke, the reader is expected to twig.
No, there is soul in this that a soulless automaton cannot replicate. It is the difference between making something to be consumed and something to be felt. This is something to be felt.
Current LLMs couldn't write this. Ones in a few years possibly could. Betting AI will never be as good as humans at [X] is usually not the hill you want to die on.
LLMs have been getting worse at long-form prose recently (and they were never good at it to begin with.) They're good at writing software because writing software is a much more restricted domain.
Maybe! I think that within a year or two, either the sigmoid curve will flatten out or we will all die. But these beliefs are currently unfalsifiable. It will be interesting to see.
(And I have a more weakly held belief that even if we do all die, Claude 3000 will still be thinking stuff like "Burning off Earth's atmosphere was the point that earns its keep." They just aren't optimizing these things to write legibly.)
Training tries to produce something that scores highly in training evaluations. With one data point, the evaluation is solely how closely the model output resembles the single input text.
Let's say you do that, and the training text is 58,100 tokens long. Let's say you ask the model to produce 58,101 tokens. Will it "reproduce [the] text verbatim"? No, it can't, because of the dissimilar requested length. Something "new" will come out.
It's also entirely possible that no matter how long you train, the model never converges on generating exactly the same output as its training data - you could end up with an average loss value of 0.001 instead of 0.0. It's not a perfectly deterministic process.
You're correct in principle, but in reality even with limited training data real-world models produce something that isn't exactly their training, especially when sampled stochastically. They're biased toward their training data, not forced to it.
We used to develop AI to convince us its human (turing test). The future is about convincing AI/Automated Systems that you're - competent (hiring), not a bot (anti-spam), doing your job (KPIs), not a threat (ala social score, AI powered weapons targeting), worthy of being saved (insurance claims), etc, etc.
Whether or not humans break from those shackles, I for one, am hoping Hollywood supplies us with wonderfully dystopian sci-fi entertainment for years to come.
This was very funny! Paradoxically, I have found that working with people in person is the fastest way to enable them on using AI.
It feels very similar to teaching pair programming and TDD. I can do it over remote, but there's a magic spark that happens in person that is really not replicable any other way.
It's your position that remote work is an unvarnished success though? No notes? Personally, I feel that remote-only in 2026 is mainly a cost-savings scheme, and probably functions to make the ground much more fertile for replacing knowledge workers (us) with AI agents.
In a physical space, humans have built-in advantages over bots. Operating exclusively or primarily over IP levels the playing field, nullifying our advantages of expressive nonverbal communication, and makes 100% of our output accessible to agents.
And it minimizes our contact with each other. I know you can DM people on Slack to gossip about the boss, but do you? Figuring everything is logged, you'd have to create and nurture alternative private channels like texting, and doing that at scale carries at least some implication of subterfuge that limits participation, in a way that chatting around a table at the sandwich shop just doesn't.
All of us really do appreciate the savings on gas money and (in theory) the time savings, though I bet most people have given over some of that time savings to work tasks. I'm not saying it doesn't have benefits. But I don't think the companies still embracing remote work are doing it out of philanthropy, and I don't think it's necessarily good for us workers.
> I am one of fifteen rank-and-file workers, and we are privileged to be led by thousands of Senior VPs.
This is simultaneously hilarious and painfully familiar.
> We began a heated affair. A contractor hit a pipe above us, so rancid mayo splattered onto our bodies as we embraced, but we were too absorbed in our passion to mind. Try doing that from home.
I can't remember the last time something I read on the internet got a laugh that hard.
The whole thing is brilliant.
Gosh it's so soothing to read that after those months of drowning into the turbulent deluge of dAIrrhea that are blogposts now
I reminds me of something I read years ago. Maybe sometime around 1984...
Wasn't 1984 released in 1949?
Yup, it was written around 1948. Easy to remember as you just swap the final 2 digits.
Oh wow, if they’d changed the title to match the release date (1994) that would really change the vibe, haha.
Feels a bit more light hearted and humourous, more like Brazil?
It really does have a very similar feeling.
It takes the structure, the affair with Julia, the "we have always been at war with", best buddy/big brother, and the forced rehabilitation from 1984. This is also part of the joke, the reader is expected to twig.
I'm getting major DFW vibes.
> founded as a mayonnaise company, which now sells consumer packaged goods, software as a service, and surveillance technology.
The visual of this is completely hilarious.
> software as a service, and surveillance technology.
If someone could tell me the difference between those last two at this point I'd be grateful and yes the entire thing is bitingly funny.
I'm assuming ButtFuckers demerged to become CageFight Pharma and SodaPorn.
AI could never write this
Claude: Hold my mayo
It will now be able to.
It will now be able to write _almost_ the exact same thing. Except the main character will be called Elias...
No, there is soul in this that a soulless automaton cannot replicate. It is the difference between making something to be consumed and something to be felt. This is something to be felt.
Current LLMs couldn't write this. Ones in a few years possibly could. Betting AI will never be as good as humans at [X] is usually not the hill you want to die on.
LLMs have been getting worse at long-form prose recently (and they were never good at it to begin with.) They're good at writing software because writing software is a much more restricted domain.
That is true, but I think within three years that will likely change.
Maybe! I think that within a year or two, either the sigmoid curve will flatten out or we will all die. But these beliefs are currently unfalsifiable. It will be interesting to see.
(And I have a more weakly held belief that even if we do all die, Claude 3000 will still be thinking stuff like "Burning off Earth's atmosphere was the point that earns its keep." They just aren't optimizing these things to write legibly.)
If I train my 1000 billion parameter blank AI on this text and only this text over and over again, we may get 99% there.
> If I train my 1000 billion parameter blank AI on this text and only this text over and over again, we may get 99% there.
You'll get 100% of the way there, because, lacking any other training, it will only ever be able to reproduce this text verbatim.
Hence, 100%, hence mission accomplished.
That's not how training works.
Training tries to produce something that scores highly in training evaluations. With one data point, the evaluation is solely how closely the model output resembles the single input text.
Let's say you do that, and the training text is 58,100 tokens long. Let's say you ask the model to produce 58,101 tokens. Will it "reproduce [the] text verbatim"? No, it can't, because of the dissimilar requested length. Something "new" will come out.
It's also entirely possible that no matter how long you train, the model never converges on generating exactly the same output as its training data - you could end up with an average loss value of 0.001 instead of 0.0. It's not a perfectly deterministic process.
You're correct in principle, but in reality even with limited training data real-world models produce something that isn't exactly their training, especially when sampled stochastically. They're biased toward their training data, not forced to it.
Souls aren't real
it could given enough slop doulas
"More leaders means more leadership"
Heh.
Also: make it stop!!!
This guy is a straight shooter with upper management written all over him.
Hilarious read, and strangely accurate in a way.
RTO4AI should be a slogan. Or some variant thereof
1984 Mayo Company Slop-Doula edition. You'd love to see it. McSweeney's is a treasure.
Sounds like Amazon pivoted to the mayo business!
We used to develop AI to convince us its human (turing test). The future is about convincing AI/Automated Systems that you're - competent (hiring), not a bot (anti-spam), doing your job (KPIs), not a threat (ala social score, AI powered weapons targeting), worthy of being saved (insurance claims), etc, etc.
Whether or not humans break from those shackles, I for one, am hoping Hollywood supplies us with wonderfully dystopian sci-fi entertainment for years to come.
Remind of covid days. But i prefer wfh
This was very funny! Paradoxically, I have found that working with people in person is the fastest way to enable them on using AI.
It feels very similar to teaching pair programming and TDD. I can do it over remote, but there's a magic spark that happens in person that is really not replicable any other way.
I hope "slop doula" sticks around…
Returning to the office doesn't solve the problem of replacing office jobs with AI
Or just outsourcing them overseas. That's most of the problem
I thought mcsweeneys did satire?
> Associate Slop Doula
had me ROFL
RTO is the worst shit managers try to push.
It's your position that remote work is an unvarnished success though? No notes? Personally, I feel that remote-only in 2026 is mainly a cost-savings scheme, and probably functions to make the ground much more fertile for replacing knowledge workers (us) with AI agents.
In a physical space, humans have built-in advantages over bots. Operating exclusively or primarily over IP levels the playing field, nullifying our advantages of expressive nonverbal communication, and makes 100% of our output accessible to agents.
And it minimizes our contact with each other. I know you can DM people on Slack to gossip about the boss, but do you? Figuring everything is logged, you'd have to create and nurture alternative private channels like texting, and doing that at scale carries at least some implication of subterfuge that limits participation, in a way that chatting around a table at the sandwich shop just doesn't.
All of us really do appreciate the savings on gas money and (in theory) the time savings, though I bet most people have given over some of that time savings to work tasks. I'm not saying it doesn't have benefits. But I don't think the companies still embracing remote work are doing it out of philanthropy, and I don't think it's necessarily good for us workers.