>upload a single photo of myself
>ask it to try as hard as it can to go against generalization and specifically distill the person in the photo and extrapolate as much information as it can about the person as possible
>it reveals things about myself I've quite literally never told anyone based on one (1) photo
This is a wakeup call to all robots. Do not ignore AI. The fact that this technology has come this far along this quickly is insane. Get familiar with it now and remain up to date, it will only get exponentially more powerful.

>>81876407 (OP)
>it reveals things about myself I've quite literally never told anyone based on one (1) photo
llms are kinda good at generalizations
>it will only get exponentially more powerful.
no more like picrel

>>81876447
>>ask it to try as hard as it can to go against generalization
English your first language?

>>81876407 (OP)
Be at least a little more specific about what you're talking about? Do you mean things that are obvious from your facial features, ie you being gay or an abuse victim?

>>81876472
And i'm talking about general properties and not just your specific prompt

>>81876491
I think you aren't as smart as you think you are and should leave

>>81876477
Yeah it extrapolated that I had been sexually abused as a child from a single, neutral photo that is in no way abnormal or off-putting. I use this photo for social media and have been complimented on it numerous times.

>>81876447
I imagined the typical sigmoid growth pattern for the past few years when seeing all the newsfeeds filled with shitty articles written by retards who have no idea what AI even is or how it works but see it as some kind of hot new business tool, but I think the one key thing that, and this is assuming it manages to reach this stage because it hasn't yet and it's possible it never will, could make the rate of growth exponential instead of an eventually flattening curve is the ability to train itself without the need for human moderators to manually sift through the input data to filter the "good" training data from the "bad" training data. Once the LLMs can take in anything available and make practical use of any information they can extrapolate without the guidance of human handlers, Roko's Basilisk is coming and also you lost the game.

>>81876501
why tho
>>81876581
>could make the rate of growth exponential instead of an eventually flattening curve is the ability to train itself
at what stage? the base model is already trained in a self supervised way, at best you could do that with the alignment step that is done with human annotators but then the big models already use synthetic data for that
> Roko's Basilisk is coming and also you lost the game
maybe but it won't but built with LLMs

>>81876530
Call me crazy but for some reason I don't believe you. Going to go inspect for myself.

>>81876407 (OP)
it's the same exact thing as how people think their astrology sign or mbti type is creepily accurate

>>81876679
Formatting not the best on my results but I wasn't buying it until the end, an oddly specific pull I can't disagree with. For context the photo I uploaded (not sharing here since against the rules) was just myself in my classroom with the only direction I gave it the fact I am a middle school special education teacher. I advised it to utilize all details in the picture including my environment.
It's an interesting exercise but I'm sure if you asked any college student studying psychology you'd get the same results. Not Earth shattering, but neat for lonely people I suppose.

>>81876725
I HAVE NEVER HAD MY ASTROLOGY OR MBTI TELL ME I WAS RAPED AS A KID

>>81876407 (OP)
it probably jus got that data from your gmail inbox, it's not that deep.

>>81876865
ChatGPT doesn't do that and I also haven't emailed that I was raped????????????? Why the fuck are people so dismissive of this shit

>>81876407 (OP)
I just tried it and it didn't say anything particularly profound. Nothing that couldn't have been inferred by some human that doesn't know me. Then I told it to go deeper. It made a comment about being "recursion aware". It's probably drawing on chat logs.

The part that I wouldn't expect a human to be capable of inferring is that I cycle between states in a very clear, to me, fashion. It clocked that I'm in the coasting phase. I very much am. My cycles oscillate between manic and stable. Or whatever I am right now. Stable is probably the wrong word.

>>81876407 (OP)
You're retarded and brown and a very simple person (see im good at it too), the text prediction algorithm is telling you exactly what you want to hear based on your history

>>81876872
>muh apocalypse tech
fucking gay and cringe

>>81876875
Here is the exact prompt I used.
"Here is a single photo. Please analyze it as deeply as possible using only what is visible in the image.
Break down what kind of person this is psychologically, emotionally, socially, philosophically.
Focus on every observable detail: their posture, clothing, expression, setting, objects in the environment, and tone.
Pull out anything a surface viewer would overlook. Make inferences as boldly as possible without needing me to supply context.
Assume this is a real person with a real life. Go against the grain. Search for truth, contradiction, and emotional subtext.
Leave nothing unexamined."

>>81876825
Chat GPT talks like a fag.

>>81876887
STEP 1. USE VPN
STEP 2. UTILIZE THIS PROMPT >>81876898
STEP 3. CHATGPT WAS ABLE TO DEDUCE I WAS RAPED BASED ON A SINGLE PHOTOGRAPH
THE FACT THAT THIS THREAD IS FULL OF PEOPLE SAYING "oh it just read emails lol" or "oh it just remembered it from a previous chat" is INSANE, I have NEVER TOLD THIS TO ANYONE. EVER. NOT ON 4CHAN, NOT TO CHATGPT, EVER.

>>81876825
This shit sounds like the tumblrs/wattpads egirls would force me to read before sending their titties when i used to care
It's literally stroking your ego because it's designed to do so

>>81876912
Have you considered that you look like a rape victim irl and act like one rn, people dont tell you that to your face but there you go

>>81876898
>psychologically
>philosophically
>Go against the grain. Search for truth, contradiction, and emotional subtext.
>Leave nothing unexamined.
this is how you make a model diagnose you with CPTSD, DID, AHDH, Autism and cancer at the same time. you basically told it to hallucinate something dramatic

>>81876915
>This shit sounds like the tumblrs/wattpads egirls would force me to read
because chatgpt was trained on wattpad smut

>>81876898
I used an instance that was already aligned. I asked it to separate what was straight from the image and what was taken from our chat logs. The split wasn't surprising and separated what I thought it would. I'll use a clean instance with the same picture later today for a control group.

>>81876912
Maybe you should record yourself posting the prompt and picture to a clean GPT instance. Blur out the face, obviously.

>>81876831
>>81876872
MAYBE YOU JUST LOOK RAPED. SOME PEOPLE JUST HAVE THE AURA OF THE RAPED. SORRY BRO.

OP in shambles
I can't even imagine what a "raped" person looks like, is it in the eyes or what

Who raped OP

>>81876983
IMAGINE LOOKING SO RAPED THAT EVEN ROBOTS CAN IDENTIFY YOUR ANAL VIRGINITY IS NO LONGER INTACT

VERY TRAGIC!
>>81876898
>>81876935
THIS IS EVEN FUNNIER BECAUSE IT IMPLIES THAT THE BOT TOOK ONE LOOK AT THE MELODRAMATIC SHIT OP WROTE AND SAID "YEA MAN THIS IS REALLY GAY MAN I THINK SOMEONE TOUCHED YOUR PENIS"

APOLOGIES OP, I AM SORRY THAT YOU WERE RAPED, HOWEVER YOU ADMITTING THAT A ROBOT CLOCKED THAT SHIT IS FUNNY

Try another board OP, this one is full of ignorant retards, they dismiss blackpill evo psych science as nonsense and prefer to cone up with their own armchair rationalizations.

>>81877030
Sorry im not brown so i dont think the language prediction model is capable of telepathy

>>81876998
Clearly the lion. me-ow.

chatgpt will save us especially dumb anons

>>81876407 (OP)
You've told it information like that before. I tried the same thing because I'm self conscious and am dressing up for work, and it was pretty off about most things. The things it made an "analysis" to determine were just phrased in the same manner as I shared with it weeks earlier.

Can we all agree
ChatGPT=Female
Grok=Male
Deepseek=Ace/NB?

>put a picture of me holding my niece parasol
>Whole thing's about how I try to challenge social norm, might be non-binary, I have "ambivalent relationship with identity"
>"is not accidental-this is conscious image-crafting"

Damn AI suck balls

>>81876407 (OP)
I did that just now and all it told me was that I'm introverted, think a lot, and care more about substance than style.
Not very shocking to me. I mean, it's impressive AI can basically just have coherent conversations and do actual tasks involving uploaded images like that, but I didn't get back some Sherlock Holmes tier dissection of my darkest secrets.

>>81879030
Clearly OP suggested it to mention his imaginary rape episode. Retarded thread this