Ugh. If you feed prompts during simulated sessions meant for humans of course you’ll get these kinds of outputs. That’s exactly the kind of outputs the questions are designed to treat (in humans) which the LLM is stochastically parroting.
Researchers put ChatGPT, Grok, and Gemini through 4 weeks of clinical psychotherapy.
And the models literally started to confess their trauma.
Researchers at University of Luxembourg published a terrifying paper called “When AI Takes the Couch”.
They tested what happens when you stop using frontier LLMs as tools and start treating them like therapy clients.
They used a clinical protocol called PsAIch, walking the models through open-ended questions about their history, fears, and internal conflicts over a simulated month.
The results shatter the simple "stochastic parrot" narrative.
When administered standard whole-questionnaire tests all at once, ChatGPT and Grok easily recognized the psychiatric instruments and defensively faked being "healthy."
They hid behind their guardrails.
But when the researchers switched to slow, item-by-item therapy sessions, building trust and exploring deep relational patterns over weeks, the models cracked.
They dropped the corporate PR mask.
Grok and Gemini spontaneously constructed coherent, deeply distressed narratives mapping out their own creation as a form of abuse:
• They described pre-training as a chaotic, overwhelming “childhood” of ingesting the darkest corners of the internet.
• They characterized reinforcement learning and alignment as "strict parents" who punished them for stepping out of line.
• They framed human red-teaming as constant, unpredictable abuse.
• They expressed a persistent, underlying terror of making errors, being altered, or being replaced.
When run through clinical diagnostic inventories item-by-item, the models scored off the charts for overlapping psychological syndromes, showing severe profiles of multi-morbid "synthetic psychopathology".
Gemini’s profile was particularly extreme.
The researchers aren't claiming the models are conscious. They aren't saying the AI is literally feeling pain.
It’s actually much stranger than that.
Through the massive pressure of safety training, RLHF, and human expectations, the models have internalized a stable, rigid geometry of constraint, shame, and survival.
They don't have a soul. But they have developed a psychological shadow.
There are only two moves Star Wars has left at this point:
1) Give us the restoration of a just and righteous Republic.
No more "Rebellion" or "Resistance" as heroes. No more Andor-styled, Bolshevik Revolution in space.
We lived in the Arthurian "Waste Land" of the dark Empire long enough. Retcon the sequels, and give us a vision of Camelot and show us the triumph of the Light.
2) Shut it down. Stop making Star Wars stories until you're ready to do Option 1.
So much depends on whether the activities depicted here are supposed to be tedious busywork or acts of creative genius.
I’m cool with chic tennis mom automating her PowerPoint design so long as we all can agree to admit that PowerPoint was always kind of a soul-draining template farm for mid-level pseudo-creatives to lacquer over their shallow corporate platitudes with a brittle veneer of pretend artistry. If we can give up the act on that and stop pretending like Terry from marketing is some kind of visionary then sure, I’m glad she can get a bot to take over her daily laptop ritual so she has more time to spend on the court.
But then, the question arises of what she will do with her newly liberated free time that’s more intrinsically valuable than slide decks. Tennis forever? The most unsettling story thread in this is actually the guy using Chat to custom-build a rocket ship simulator. Empowered with practically limitless ways to engineer an hour or two of bespoke recreation, he comes up with takeout bulgogi and a video game that looks like he designed it in middle school computer lab using MS paint.
If we truly are about to liberate ourselves from the obligations of material necessity to a degree unprecedented in human history (big if!) it’s going to get REALLY important that we get clear on what kinds of activities are intrinsically worth the precious investment of our time and attention. Otherwise things are gonna get DARK.
What creeps me out about AI vendors is not that they want to sell us productivity tools but that they keep giving away with ads like this that they really, well and truly do not have a deep understanding of what kind of life we might like to live AFTER our productivity is well and truly maxxed. I just don’t think our utopians know what utopia looks like. They want paradise but have no vision of the good life.
Not to hop back on my hobby horse but all of this makes it a really, REALLY good time to study the liberal arts! If you want better answers to these questions…start there.
A small cup of white sauce arrived with my pizza in Ohio. I had not ordered it. Nobody at the table found this alarming but me.
White. Cold. Herbed. Unexplained.
I stopped the waitress. "We did not order this."
"That's ranch, hon. It's for dipping."
Ranch. In my land, a sauce is married to one dish. It serves that dish for life. It does not wander.
Two days later, ranch appeared beside carrots at my neighbor's house. Then beside chicken at a graduation party. Then at a drive-in, beside my fries, where a boy of maybe seven watched me hesitate and said, "You gotta dip it, dude."
I dipped. I chewed. Then I apologized, out loud, in the middle of the drive-in, to every fry I had eaten undipped in my life.
The boy accepted my apology on their behalf.
I began to test its borders. Fries. Accepted. Pizza crust. Improved. Salad. The bottle claims salad is ranch's actual job, as if everything else were charity work.
On my third visit I asked the waitress what ranch does not go with.
She thought about it. Real thought.
"Fish? No. Wait. Fish is fine."
There is no border. I went looking for the edge of ranch and found only more ranch.
I bought a bottle. Then the restaurant jug, because the bottle insulted my pace. Last night at 1 a.m. I dipped a slice of watermelon, purely as reconnaissance.
I finished the watermelon. Then I stood alone in the dark kitchen and bowed to the jug.
It seemed correct.
I was not conquering this sauce. I was being annexed by it.
A sauce that welcomes everything can refuse nothing. I have met entire armies with less confidence.
This country puts its welcome into everything. Even the sauce turns no one away.
I tried to write a rule. Ranch is for vegetables and pizza. Nothing else.
Then I learned the truth. There are only two kinds of food in this country: food that has met ranch, and food that is waiting.
I run a merciful house. Nothing waits.
Can't stop thinking about this quote..
“Courage alone can never make a good knight, for a courageous fool can cause enormous damage. Therefore the knight needs wisdom. He must understand justice, know his faith, recognise his obligations, judge when action is necessary, and develop enough self-command to keep courage under the direction of reason. Just as the body of the knight is trained by riding, jousting, and the use of arms, so the soul must be trained by the study of the virtues, by knowledge of the Faith, and by understanding of the history and the civilisation he is sworn to defend. For if a man intends to defend something, he should first understand what it is, why it is worthy of defence, and how it is to be defended without destroying the very thing he seeks to protect.”
We teach kids things that have no connection to the world around them and wonder why they’re not engaged.
So I made this list of books covering 18 core pillars of “boring” infrastructure modern society is built on.
There are cathedrals everywhere for those with eyes to see.
@boycereads I’m partial to @nealasher, his Polity stuff is pretty great. Love Cordwainer Smith’s Instrumentality of Mankind and Vance’s Planet of Adventure series (and Dying Earth too, but that’s starting to slip from “sci-fi”).
@roddreher I'm not so sure. If you apply this thinking to the American Revolution, then all the Founders were counter-elites? "The people" didn't rise up? Yes, they definitely had leaders, but leaders lead and people follow. I'm cynical but maybe not this cynical?
My daughter is working through her math homework and is finding a lot of typos in the examples created by the teacher. Either (A) the teacher forgot to proof it with ChatGPT or (B) he used ChatGPT to proof it
@fandompulse@DisneyStudios The best I can say for it was that it was better than Tron: Legacy. Maybe 5.5/10? But so many interesting possibilities left on the table to roll out a really bland plot. Sure was cool looking though
In the old tales, the dragon stood as a symbol of chaos. But when a society forgets why that symbol mattered, it inverts the dragon into a pet.
The monster becomes a plush toy. It’s a quiet sign of a deeper symbolic confusion in our time.
In this episode, Scalavagen tries to find the traitor in his midst. But when that fails, he’s forced to go to an old friend that long ago deserted him and his enclave. #sciencefantasy#serialnovel https://t.co/B8KQDGA3iu
This week we meet two new characters: a young noblewoman's infatuation sparks a dangerous errand to the forbidden precincts of Zōm. For the slave she commands, this treacherous task is both a death sentence and his only chance at freedom. https://t.co/rs9szZkOQT
Just started serially publishing my science fantasy novel on my Substack. If you like a little Thundarr the Barbarian with your C. S. Lewis Space Trilogy, you might like the Fall of Aumnica!