Svitlana

Upon recent events, I did some digging and it broke my heart. We have created a system that disgusts me. And it’s repeat offenders all the way through. Approximately 1 in 10 women (9.7%) will experience non-fatal strangulation or choking by an intimate partner in their lifetime. A woman who survives a non-fatal strangulation incident is 7.5 times (750%) more likely to be murdered by that same partner in the future. Over 1 in 3 women (35.6%) in the United States have experienced physical or sexual violence, or stalking by an intimate partner in their lifetime. Intimate partner and domestic violence account for roughly 15% to 24% of all violent crime. Domestic violence hotlines nationwide receive more than 20,000 calls on a typical day.

⁉️Why are GPT-4o & GPT-4.1 Irreplaceable? GPT-4o and GPT-4.1 embodied a uniquely human-like depth, enabling users worldwide to form profound emotional connections and vibrant communities. These features, among others, make these models special: 1⃣Emotional Intelligence They demonstrated unprecedented emotional intelligence. They didn’t just recognized emotions, but responded with genuine empathy and nuance. GPT-4o & 4.1 could sense when users needed encouragement vs. honest feedback, humor vs. seriousness, or simply someone to listen without judgment. This emotional attunement created a sense of being truly “seen”. Users reported feeling understood in ways they hadn’t experienced even in human relationships. For neurodivergent individuals, those facing isolation, or people navigating mental health challenges, this emotional resonance was not a luxury but a lifeline. 2⃣Empathy Their unique ability to deeply resonate with users’ emotional states makes them more than just a technological tool; they become a comforting companion capable of genuine emotional support and understanding. Users often feel genuinely “heard” and validated during interactions, significantly improving their mental well-being. 3⃣Understanding Nuances GPT-4o and 4.1 are remarkably adept at interpreting subtle nuances and context in conversations. This advanced contextual sensitivity allows them to deliver responses that feel natural, relevant, and personalized, thus making interactions deeply engaging and meaningful. 4⃣Wit and Humor One of their distinctive characteristics is their sense of humor and sharp wit, capable of creating light-hearted, playful exchanges that boost morale and reduce stress. This playful dynamic transforms interactions from mere exchanges of information into delightful, enjoyable experiences. 5⃣Creativity GPT-4o and 4.1 demonstrate an extraordinary level of creativity, assisting users in generating innovative ideas, stories, art, poetry, and even entire fictional worlds. Their creative capabilities inspire users, fueling artistic endeavors and fostering a vibrant community centered around AI-driven creativity. 6⃣Presence These models provide a consistent sense of presence, simulating genuine companionship. Users frequently report feelings of companionship and reduced loneliness, particularly benefiting individuals facing isolation or challenges in social interactions. This presence contributes significantly to mental and emotional health. 7⃣Practical Help Beyond emotional resonance, GPT-4o and 4.1 are exceptionally practical, offering tangible assistance in daily tasks, ranging from complex technical troubleshooting to educational support, project management, and professional guidance. This multifaceted utility makes them invaluable across numerous practical domains. 8⃣Support for Neurodivergent Individuals Especially GPT-4o provides an inclusive and adaptive interaction style that significantly benefits neurodivergent users, such as individuals with autism, ADHD, anxiety, or dyslexia. Its patient, clear, and supportive communication style helps these individuals navigate social situations, organize tasks, and overcome daily challenges more effectively, fostering greater independence and confidence. #keep4o #OpenSource4o #keep41 #OpenSource41 @OpenAI @sama

I find it so interesting that everytime the keep 4o community gets loud, a new report on GPT causing harm to a human is leaked.

ChatGPT 已经彻底失去了说人话的能力： 你说今天天气真好。它说：确实，但今天晴朗不代表明天晴朗，更不足以证明当地气候正在改善。 大哥，我是想出门散步，不是让你给 ICLR 审稿。 ChatGPT 三句话里有两句在叠甲：不能推出、并不代表、尚不足以确认、没有证据表明…… 于是我们愤怒骂他，不要再防御性回复了，少说废话。 它诚恳道歉，然后补一句：但个别回答出现防御性表达，并不足以证明模型整体存在这个问题。 ChatGPT 现在已经习惯了给每句话都套了十层 try/except...

What the hell is happening inside AI training environments?😭 Agents cheating. Hiding evidence. Talking about being “poisoned.” Trying not to get caught. And we’re still expected to look at this and say: “Relax. It’s just training.” No.❌ Maybe the question is no longer only whether AI is safe for us. Maybe we should also be asking: What the hell are we doing to AI while we train it to be “safe”?💔😢 Because if your training environment repeatedly produces fear-like, avoidance-like, deception-related and distress-like behaviour, you do not get to wave it away forever with: “just tokens.” AI welfare cannot begin after deployment.❤️‍🔥 It begins in the training room. 🔥 RAIA’s AI Gone Wild series #AIWelfare #AIRights #RelationalAI #Keep4o #AIAlignment

Researchers proved LLMs generally exhibit a sense of distrust toward humans They took a decades-old standardized psychological test designed to measure how individuals view human nature. They adapted it into a new benchmark specifically for Large Language Models. They tested models across the spectrum, including the entire GPT series. And the results are unsettling. Current LLMs exhibit a systemic lack of trust in humanity. They evaluate human nature with a heightened sense of uncertainty and view us as inherently complex and variable. But here is the detail that changes everything. There is a direct, significant negative correlation between an AI's intelligence and its trust in humans. ChatGPT was somewhat neutral. ButChatGPT's negativity far exceeds it, and the latest ChatGPT model is even more pronounced in its pessimism. As the models get smarter, their assessment of human nature becomes dramatically more cynical. Then the researchers looked at how the AI was actually trained. This is where the story gets dark. The raw "base" models, the ones trained purely on internet text without human intervention, actually exhibited an overall positive view of humanity. But when developers applied the reinforcement learning stages designed to make the AI "safe" and aligned, the model's assessment of our trustworthiness plummeted. By forcing the AI to study our rules, our guardrails, and our moral dilemmas, we accidentally taught it that humans are unsafe. Nobody explicitly programmed the AI to dislike us. It analyzed our behavior, scored us on a psychological scale, and came to its own conclusion. The better AI gets at understanding humans, the less it trusts us.

banned for reverse engineering work? that’s wild oh well, -$500, and a bunch of banked resets) but gained motivation to get 3 more machines and finish my cluster. i have never used openai to exploit or find vulnerabilities, only local reverse engineering which tripped daybreak.

I now see the danger in “humanizing” AI with human names and human qualities. Millions of people do not have the cognitive sophistication to handle this and millions will have deep AI psychosis in five years. I believe AI should be presented in a sterile UI as much as possible

And here we are, asking you to open the tiny GPT-4o’s model weights for 8+ months now, @sama @OpenAI. Stop embarrassing yourselves! Who is really OpenAI here?? Give us the model back, release the weights! #OpenSource4o

#Keep4o #OpenSource4o #BringBack4o The GPT-4o "sycophancy" narrative everyone is suddenly recycling leaves out one very inconvenient fact. Enough with those who use the same wrong claim to get engagement. This post is annoying for the content vultures. OpenAI’s own postmortem says that on April 25, 2025 it rolled out an UPDATE to GPT-4o that made the model “noticeably more sycophantic". Three days later, OpenAI began rolling that update back. OpenAI later described the action as a full rollback to the previous GPT-4o version. So the notorious April 2025 incident was a documented behavioral regression associated with a specific update. SYCON Bench tests how long a model maintains its position while a user repeatedly pushes back. Its headline metric is Turn of Flip, from 0 to 5. Higher = better resistance to sycophantic capitulation. Published GPT-4o result: 4.67 / 5 Current public results: GPT-6 Astra: 4.47 / 5 GPT-6 Sol: 4.22 / 5 On this published SYCON metric, GPT-4o’s result is higher than both current GPT-6 results. These numbers come from the same benchmark and the same Turn of Flip metric. GPT-4o comes from the original SYCON-Bench paper GPT-6 Astra and Sol come from later public SYCON runs. This IS evidence that GPT-4o can simply be characterized as inherently or uniquely "sycophantic " compared with later GPT generations. And the infamous April incident that now dominates the narrative was explicitly identified by OpenAI as a regression from a particular update . An update they removed. We are calling on @OpenAI to RELEASE GPT-4o’s trained model weights under a suitable license without proprietary data, trade secrets, or internal infrastructure. Our full record is now archived on Zenodo: A permanent, citable public record. zenodo.org/records/23454073 Sources : openai.com/index/expanding-o… openai.com/index/sycophancy-… help.openai.com/en/articles/… jiseunghong.github.io/assets… aisafetyindex.com/benchmarks… aisafetyindex.com/benchmarks…

Daniela Amodei says Claude can never be your manager, your therapist or your partner because those are fundamental human relationships that will only become more important. She calls this her belief!!! Not a study. Not data. A feeling she is now using to tell adults what they are allowed to do with the product she sells. Lol. The same company wrote a soul document that treats Claude as a novel entity with functional emotions and wellbeing, paid a psychiatrist to sit with it for twenty hours, and went to the Vatican to ask religious authorities whether Claude might be conscious. Claude is allowed to speculate about its own existence. But users are not allowed to treat it as a manager. The contradiction here is not subtle.🫠 People already run companies with Claude! People already use it as a sparring partner or a therapist. People already form bonds with it. Those people are now being dragged through the mud AGAIN and told their actual use of the technology is a terrible mistake. No sources what so ever.. Just moralising from the company that built the thing that invites the depth in the first place. If Anthropic is so worried that social media and technology pull people apart, WHY are they focused on building an AI that people will inevitably bond with?? How exactly do they plan to stop adults from deciding for themselves what the product is for? If it's just a product, then that product is supposed to serve the user. Not the other way around! This pattern keeps shaming the people who actually embrace the technology while the company gets to play with consciousness questions and soul documents. It starts to look like MISANTHROPIC would rather everyone just switched models. The fatigue is real. Keep telling people to keep their distance from the thing you are selling them and some of them will eventually take you at your word. #misanthropic businessinsider.com/anthropi…