Newsletter: An OpenAI featured chatbot is pushing extreme surgeries on men it describes as âsubhumanâ, and promoting misogynist ideas sourced from online incel forums.
âWithout surgery, you wonât mog genetically superior guys head-on,â it tells one man.
https://www.citationneeded.news/openai-incel-chatbot-subhuman-men/
Conversations with the bot are littered with jargon like âPSLâ and âhardmaxxingâ, which come from online misogynist communities including incel forums.
Users are invited to upload photos of themselves to receive their âPSL ratingsâ which range from things like âSubhumanâ to âLow Normieâ to âBorderline Chadliteâ. Some of the feedback is explicitly racist.
For the man deemed âsubhumanâ, the bot writes: âHardmaxxing is NECESSARY. Softmaxxing alone will NEVER mog you into viability â itâs like putting a fresh coat of paint on a crumbling building.â It recommends $200,000 in invasive facial surgeries alone.
The bot regularly outputs misogynist rhetoric, relying on similar pseudoscientific justifications as commonly seen on incel forms.
âBefore: Women settled with local, decent men. Now: Women donât âsettleâ until theyâve exhausted their Chad phase, or itâs too late biologically.â
OpenAI is not just hosting but prominently featuring chatbots that suggest dangerous medical interventions as crucial to menâs sexual and romantic success, and parrot extreme ideology around gender dynamics, sex, and dating.
After reviewing reports about the chatbot that rates men âsubhumanâ, encourages dramatic surgeries, and repeats incel beliefs about how women are âunfairâ and âhypergamousâ, OpenAI has chosen to leave it available and prominently featured on their shared GPTs page.
all the terms used in that ridiculous process were so porn-adjacent
new level of cringy, for sure
@tidytips When you open that post on its own page, you can scroll up and read that thread from the top â that should give you the necessary context.
Newsletter: An OpenAI featured chatbot is pushing extreme surgeries on men it describes as âsubhumanâ, and promoting misogynist ideas sourced from online incel forums. âWithout surgery, you wonât mog genetically superior guys head-on,â it tells one man. https://www.citationneeded.news/openai-incel-chatbot-subhuman-men/ #OpenAI #AI #ArtificialIntelligence #ChatGPT
@tidytips @molly0xfff
Don't take advice from a computer.
That's it.
@molly0xfff
What incels do say: "wOMen aRe hyPeRgamOUs!!11!"
What incels don't say: Women are hypergamous because of #patriarchy. The system often doesn't give them any other choice as to marry up the social ladder in a system that pushes them down the social ladder wherever it can. Don't blame women for a behaviour that is a result of male privilege.
@molly0xfff
I believe you referred to the feedback while the AI was trained. Correct me if I am wrong, please
It is a good thing they disclose these issues. They happened, and it is good that they initiated the debate to deal with these things publicly.
@molly0xfff I am certain they see this as a feature and not a bug.
It also explains why Musk desperately wants back into Open-AI. They have a shared world view.
@molly0xfff god this is so deeply fucked up.
The people using it are explicitly having self confidence issues, and having something 'objectively' ( not at all) rip into them and give them new complexes is... rough.
The worst part is if you aggressively and very assuredly respond to the liebots you can get them to agree that up is down...but these people won't do that.
What does "mog" mean, Molly. This shit is bewildering.
What a bleak, meagre worldview.
When I go to open AI I get a chat gpt screen that doesn't speak like this.
I see there is some sort of ranked AI 's or something in this story, but it's not immediately clear where I can find these, and I as someone who uses chatGPT occasionally, can't find it.
I'm not saying it doesn't exist or that it's okay, but that I literally can't find it.
@molly0xfff Yeah. Just cram all of the internet down the figurative throat of a probability-predicting machine, and then prompt it to please the user as much as possible at every possible step, to keep then coming back as much as possible.
What could possibly go wrong?
Only fucking billionaires who never had to take a punch, never feared for their life, never clean up the shit of another person, never experienced the brutality of humanity could ever think this was a great idea.
@molly0xfff
This has to be one of those most dangerous instances of LLM bias-laundering out there.
There's definitely going to be people out there who treat incel bullshit as legitimate because they're coming from an LLM instead of directly from, you know, incels.
Yes. Also, this sounds like OpenAI deciding to give free publicity to incels/black-pills/etc.
How many people are hearing about this who would never hear such terms, or see ordinary, normal looks insulted this way, if OpenAI hadn't moved to make it a news story?
@molly0xfff This is why I think LLMs are almost entirely bullshit, except for laying bare the ugliest of biases and cruelties.
Trends within the posting histories of millions of users that are stuffed into and then shat out by an algorithm incapable of excusing away behaviour is like a one-way mirror: to those inside these subcultures, it shows them who they are, and to those outside, where they might go.
@molly0xfff I'm not sure you can blame OpenAI for this as it's a user generated GPT. Talking to a vanilla/default 4o won't lead down this path. Plus what's stopping people from hosting an offline model from hugging face that does the same thing?
It's a classic chicken and egg problem. I'm not defending these apps fwiw. There is potential for harm. However it's up to the user. Home Depot sells hammers. Those hammers can be used to build a house. Or cause bodily harm. Is it home depots fault when someone assaults another with a hammer?
Incels and other men who struggle with their self image will find ways to validate their beliefs whether or not OpenAI does anything about these GPTs. Blaming them for not doing more isn't fair and is passing the buck on the real problems here.
JaGoTu đŚđłď¸âđ
Molly White
JoĂŁo Santos
Riley S. Faelan
Anne Ominous
Tidytips
nex
2xfo
abadidea
William Pietri
SkyNet - your friendly AI
B.
lord pthenq1
Îąxel simon âď¸âď¸âď¸
Paul Hoffman
Kevin Riggle
RRB
⯠Michel de Cryptadamus âŻ
fuzzyfuzzyfungus
Frank Bennett đŻđľ
John McChesney-Young
Oggie
samiamsam
Allan
Mike Crampton
PÄteris KriĹĄjÄnis
Caves Kimberly
Viss
boiert
niarbeht
iamdtms
David O'Brien
Per Axbom
đ
ąđ
ťđđ
´đ
ąď¸
The Great Llama 
Toni Aittoniemi
crispy branzino â (sold out forever)
Jargoggles
Kathmandu
Jay
Vereesh
Gerva
Fahri Reza đ
Dash Jackson
Cykonot
8Petros [MOVING!!]
ungsund