#chatbots

Miguel Afonso Caetanoremixtures@tldr.nettime.org
2025-06-01

"The 0-10 “PSL rating” scale is named after “PUAhate Sluthate Lookism”, a list of three early incel and manosphere forums. PUAHate earned some notoriety following the 2014 Isla Vista killings, perpetrated by a 22-year-old man named Elliot Rodger, who had posted extensively about his incel beliefs on the forum. Though it and the other forums mentioned in the acronym are no longer online, they’ve been replaced with an active network of online communities where members commiserate in threads with titles like “if guns were easily accessible as in US, most of us have had gone ER or roped by now”.

All of the outputs of the chatbot are peppered with jargon from these forums, like “mogging” (dominating another person in terms of looks, status, and the other characteristics these men believe women prioritize most strongly), “hardmaxxing” (medical interventions like plastic surgery and hormone therapy), “softmaxxing” (less invasive improvements to a person’s attractiveness, such as grooming, fitness training, or practices like mewing), and “SMV” (“sexual market value”, a pseudo-economic ranking of a person’s attractiveness).

Some of the feedback is explicitly racist, with the chatbot identifying one set of photos as “Low Normie / High Subhuman Borderline (White)” and offering to explain “how and why being White (Caucasoid) helps in the context of PSL rating and SMV classification”."

citationneeded.news/openai-inc

#AI #GenerativeAI #Chatbots #LLMs #OpenAI #Incels

» Aakerbeere 🏖️ :mastodon:Aakerbeere
2025-06-01

💬 im FEDIVERSE

Wie stehst Du zu dem Thema?

WriterOfMinds (she)WriterOfMinds@sigmoid.social
2025-05-31

I started working on trial-and-error learning abilities for Acuitas this month! Notes on the technique I'm trying to establish here: writerofminds.blogspot.com/202
#ArtificialIntelligence #AI #chatbots #NLP

2025-05-31

PREPRINT: "Assessing the Performance of 8 #AI #Chatbots in Bibliographic Reference Retrieval: Grok and DeepSeek Outperform ChatGPT, but None are Fully Accurate" arxiv.org/abs/2505.18059 #GPT #LLMs #scholcomm #libraries

2025-05-30

"Unlike today’s chatbots, which generate answers statistically based on training data gleaned from the internet, all of SmarterChild’s responses were written by humans, or what Kay calls “curated.” This meant that SmarterChild would not give false, dangerous, or otherwise problematic answers."

medium.com/chmcore/smarterchil

#ChatBots #History #Education

2025-05-30

Decades before #Siri and #ChatGPT, #chatbots like SmarterChild were popular with Internet users passing the time with “inane chat”. Learn more from technology historian Hanen Hsu of the Computer History Museum: medium.com/chmcore/smarterchil

#Medium

Anne Sverini ☀️annsev@troet.cafe
2025-05-30

Bei der Beschäftigung mit #Chatbots wird schnell klar, wie enorm hier fremdes geistiges Eigentum missachtet wird (#Urheberrechtsverletzungen). Das passiert auch durch wörtliche umfangreich Textausgaben, die den geregelten Zitatumfang übersteigen. Was vielen womöglich nicht klar ist: Auch wenn sie umfangreiche Textpassagen vom Chatbot übernehmen und diese veröffentlichen, können sie wegen Urheberrechtsverletzung bestraft werden.

2025-05-29

Tech Xplore: Dark LLMs: It’s still easy to trick most AI chatbots into providing harmful information, study finds. “A group of AI researchers at Ben Gurion University of the Negev, in Israel, has found that despite efforts by large language model (LLM) makers, most commonly available chatbots are still easily tricked into generating harmful and sometimes illegal information.”

https://rbfirehose.com/2025/05/29/dark-llms-its-still-easy-to-trick-most-ai-chatbots-into-providing-harmful-information-study-finds-tech-xplore/

Athavariel 🇺🇦Athavariel@vivaldi.net
2025-05-29

Millionen #Menschen sind ständig online, aber oft allein. Jetzt will #Meta-Chef Mark #Zuckerberg unsere soziale Leere mit #Chatbots füllen – und macht aus Freundschaft ein Geschäftsmodell. #KI #AI #Facebook #Whatsapp aargauerzeitung.ch/kultur/soci

2025-05-29

This is alarming and should act as a wake-up call. Using AI it costs virtually nothing to create personalised misinformation campaigns aimed at local politicians - it would be surprising if this was confined to Canada.

nationalobserver.com/2025/05/2

#Misinformation #AI #ChatBots #Politics

2025-05-28

I’m going to write about this… Any thoughts? OpenAI's ChatGPT may be driving countless of its users into a dangerous state of "ChatGPT-induced psychosis."

“Friends and family are watching in alarm as users insist they've been chosen to fulfill sacred missions on behalf of sentient AI or nonexistent cosmic powerse — chatbot behavior that's just mirroring and worsening existing mental health issues, but at incredible scale and without the scrutiny of regulators or experts.” #CyberPsychosis #AI #ChatBots #psychology #Behavior

Source: Futurism share.google/5gbEHBdr0SQiso6bv

N-gated Hacker Newsngate
2025-05-28

🤔🙄 Oh wow, because the world was just DYING for another overhyped AI toy to invade our chats! What’s next, $300M to teach asteroids how to text? 🚀🪨
techcrunch.com/2025/05/28/xai-

Prof. Dr. Dennis-Kenji Kipkerkenji@chaos.social
2025-05-28

Wie nutzen die Menschen #KI im #Alltag?

Mein Live-Studiogespräch mit den Hörer:innen des #WDR gibt es nun auch als Podcastfolge zum Nachhören – und die Themen, über die wir in einer Stunde gesprochen haben, waren mehr als vielfältig: Ökologische #Nachhaltigkeit und KI-Rechenzentren, digitale Abhängigkeit und digitale #Souveränität, #Chatbots im Betrieb bis hin zum #Datenschutz in Social Media.

Hier geht’s zum Reinhören für alle, die live nicht dabei sein konnten:

www1.wdr.de/radio/wdr5/sendung

2025-05-28

@glyph This shouldn't be a surprise to anyone. #chatbots are crap. Every time I use one (when I'm forced to) I end up having to scour the website for contact details so I can speak to a human.

#AI

N-gated Hacker Newsngate
2025-05-28

🎩🐑 So, apparently, if you slap a fancy "Megakernel" on Llama-1B, your chatbot will answer before you even ask. 🙄 Their groundbreaking discovery? Faster GPUs make things faster. Who knew? 🤦‍♂️🚀
hazyresearch.stanford.edu/blog

Dr. Thompsonrogt_x1997
2025-05-27

🧠 What happens when you train an LLM on your own life?
💬 5 years of private Telegram chats
🧪 LoRA fine-tuning on Mistral-7B
🎮 PS5 jokes, dark humor, sarcastic punchlines
💸 All for just $70

⚡ This isn’t just AI — it’s me, replicated in code.
👇 Read the full experiment:
🔗 medium.com/@rogt.x1997/70-5-ye


medium.com/@rogt.x1997/70-5-ye

Miguel Afonso Caetanoremixtures@tldr.nettime.org
2025-05-27

"The curse of prompt injection continues to be that we’ve known about the issue for more than two and a half years and we still don’t have convincing mitigations for handling it.

I’m still excited about tool usage—it’s the next big feature I plan to add to my own LLM project—but I have no idea how to make it universally safe.

If you’re using or building on top of MCP, please think very carefully about these issues:

Clients: consider that malicious instructions may try to trigger unwanted tool calls. Make sure users have the interfaces they need to understand what’s going on—don’t hide horizontal scrollbars for example!

Servers: ask yourself how much damage a malicious instruction could do. Be very careful with things like calls to os.system(). As with clients, make sure your users have a fighting chance of preventing unwanted actions that could cause real harm to them.

Users: be thoughtful about what you install, and watch out for dangerous combinations of tools."

simonwillison.net/2025/Apr/9/m

#AI #GenerativeAI #LLMs #Chatbots #CyberSecurity #MCP #PromptInjection

Client Info

Server: https://mastodon.social
Version: 2025.04
Repository: https://github.com/cyevgeniy/lmst