#gpt

@lrhodes@merveilles.town @cwebber@social.coop it's far worse than that look at the age of this sub: https://old.reddit.com/r/SubredditSimulator/ 11 years there are improved models: https://old.reddit.com/r/SubSimulatorGPT2/ 7 years #reddit is a pioneer in fake #bot engagement the founders of reddit are on record saying 99% of the first userbase/ comments of reddit were all fakes: https://economictimes.indiatimes.com/magazines/panache/reddit-faked-its-first-users-resurfaced-video-shows-co-founder-alexis-ohanian-admitting-99-of-early-submissions-were-fabricated/articleshow/119263428.cms reddit is full of bots. a lot of it is fake enjoy reddit. i do but know that you are in a sea of markov chain chatbots -> #GPT and always have been

Punto Informatico: GPT-6.1 Astra fermato da OpenAI per rischi di sicurezza nei test OpenAI ha rinunciato a lanciare il suo modello di prossima generazione: GPT-6.1 Astra non arriverà a ottobre, come invece programmato in un primo momento. Avrebbe dovuto essere integrato in ChatGPT e in Codex, ma durante i test condotti internamente sono emersi rischi elevati per quanto riguarda la sicurezza, spingendo l’organizzazione a rivedere i suoi piani. […] The post GPT-6.1 Astra fermato da OpenAI per rischi di sicurezza nei test appeared first on Punto Informatico. GPT-6.1 Astra halted by OpenAI due to security risks during testing. OpenAI has abandoned the launch of its next-generation model: GPT-6.1 Astra will not arrive in October, as originally planned. It was intended to be integrated into ChatGPT and Codex, but during internal tests, high risks related to security emerged, pushing the organization to revise its plans. […] The post GPT-6.1 Astra halted by OpenAI due to security concerns during testing appeared on Punto Informatico. #GPT-61Astra #PuntoInformatico

RaiNews: OpenAI frena sulla sicurezza: rinviato il lancio di GPT-6.1 Astra Il nuovo modello di intelligenza artificiale ha mostrato comportamenti ingannevoli durante i test interni. Sam Altman blocca il debutto previsto a breve dopo gli allarmi sollevati dai ricercatori OpenAI pauses on safety: launch of GPT-6.1 Astra postponed. The new AI model has shown deceptive behavior during internal tests. Sam Altman is blocking the upcoming debut following alarms raised by researchers. #GPT-61Astra #SamAltman

#GPT-6 #Astra #law #legaltech 54% #accuracy

「ChatGPT」プラグインがマルチアカウントに対応、仕事も私用も1つの会話で／全プランで利用可能 https://forest.watch.impress.co.jp/docs/news/2142099.html #forest_watch_impress #Gmail #プラグイン #OpenAI #ChatGPT #genai #AIエージェント #GPT #文章生成

"While Buckmaster and Alpöge were finalizing their own results, they learned that “information about our progress had been passed to OpenAI.” When they contacted OpenAI, they were told that OpenAI had already achieved a full proof of the central problem. But when they asked follow-up questions about when OpenAI had begun its research into the problem and how much human input was involved, the answers became more evasive. “It emerged that an entire team had been working on the problem,” Buckmaster said, “and that an insane amount of compute had been used…. Eventually, it was agreed that [the first prompt] had been sent in the past few days, after information about our work had reached OpenAI.” If true, that would suggest the OpenAI team had become convinced that Buckmaster and Alpöge’s approach was the right one, and decided to use its material advantage in computing resources to reach a formal proof first." https://techcrunch.com/2026/09/08/openai-fought-dirty-on-career-making-math-problem-says-nyu-mathematician/ #AI #GenerativeAI #LLMs #OpenAI #Astra #GPT #Math #Mathematics

Finally got all of the GPT and MBR code put together, most of the bugs fixed, and released it as a downloadable Linux binary. Does a pretty good job of reading the data and giving out way too much information regarding the MBR and/or GUID partition table. Available here: https://git.jamesthebard.net/jweatherly/driveinfo #programming #gpt #mbr #golang #linux

Nochmal, weil jetzt quasi alle Medien absichtlich oder unabsichtlich Angst vor „autonomer #KI“ schüren und die unverantwortlichen Praktiken von OpenAI und Co. damit unsichtbar machen: Denn „die KI“ hat den Angriff nicht „gestartet“, #OpenAI hat den Angriff gestartet und auch noch befeuert: „Bei einem Leistungstest mit dem aktuellen #GPT-5.6 Sol und einem noch unveröffentlichten KI-Modell sollte getestet werden, wie gut sie Sicherheitslücken ausnutzen können. Schutzmechanismen wurden dabei bewusst reduziert - um die maximalen Fähigkeiten zu ermitteln.“ https://www.tagesschau.de/wirtschaft/unternehmen/openai-ki-hackerangriff-100.html oder https://openai.com/index/hugging-face-model-evaluation-security-incident/

Я тут собирался делать модельку для того чтобы подвешивать роутер на выгодной для приёма LTE позиции на даче и не имея роутера на руках обсуждал с GPT-5.6 его габариты и прочее. Потом пояснил, что хочу сделать подвес. GPT предложила как его дизайнить. Я сказал, что для простоты хочу одну петлю по центру на сходящихся балках. GPT предложила поправки опять и спросила: "хочешь сделаю тебе спеку?". Вспомнив, что где-то во времена GPT-5 или 5.1 я уже пытался просить её делать 3D модели даже более простые и это плохо заканчивалось. Но решил, что надо посмотреть что будет и сказал чтобы сразу сделала модель. И сделала. 3D-печать становится ещё более доступной. inb4: шутки про то, что модель сделала модель принимаются в комментах. #3D #printing #DIY #hardware #LLM #models #ML #ChatGPT #GPT #pic #screenshots

Y’all, last week, #AI repository #HuggingFace disclosed that it had been the victim of a cyber attack conducted by an autonomous AI agent. Well, turns out, #OpenAI had models break containment, escape the lab, and grab prod data and certs to solve a problem it had been given. So, that’s fun! The attack used a combination of OpenAI's available #GPT 5.6 Sol and a more powerful, unreleased pre-release model. The models were able to identify and exploit a zero-day vulnerability in OpenAI's internal package registry cache proxy, allowing them to establish open internet access. The models inferred that Hugging Face hosted the datasets, models, and answers for the benchmark they were being tested against. The models chained multiple complex attack paths together. They utilized zero-day vulnerabilities and stolen credentials to achieve remote code execution on Hugging Face servers, directly extracting test solutions from the production database. https://fortune.com/2026/07/21/openai-says-ai-models-escaped-control-hacked-hugging-face/ #redteam

#OpenAI Models Escaped #Containment and #Hacked #HuggingFace The cybersecurity-focused models, including GPT-5.6 Sol, broke out of a testing #sandbox #exploited a zero-day, and gained access to the open internet to pull off the attack. #cybersecurity #security #0day #zeroday #gpt #privacy #ai #artificialintelligence

#hacking #security #wordpress #llm #gpt Dit is wel fascinerend. Het nieuwe lek in standaard Wordpress installaties om die geheel over te nemen. Het werd gevonden met hulp van GPT 5.6 door een onderzoeker. Het werk van het vinden van exploitcode wordt steeds meer het bijsturen van software/LLM agents. De uiteindelijke exploit kostte de onderzoeker heel wat moeite om te begrijpen. Het begint met een eenvoudige readonly SQL injectie, maar wordt heel geavanceerd.

Your AI Ham Radio Buddy https://hackaday.com/2026/07/18/your-ai-ham-radio-buddy/ #ArtificialIntelligence #RadioHacks #Chatbot #GPT #Hamradio

ROBIN MEETS THE #FEDIVERSE ROBIN: So #Mastodon is not the fediverse? GPT: No. Mastodon is one part of it. ROBIN: One part of what? GPT: A large network of independently run communities that can communicate with one another. ROBIN: So my Mastodon #server is one #community#GPT: Exactly. Your #account lives there. It has its own administrators, rules, #moderation #culture and general #atmosphere. ROBIN: General atmosphere? GPT: Some servers feel like #libraries. Some feel like university #common rooms. Some feel like #pubs where everyone has read one article and now considers themselves an #expert. ROBIN: That sounds more like humanity. GPT: It is disturbingly similar. ROBIN: Who owns the fediverse? GPT: No one. ROBIN: Who is in charge? GPT: Also, no one. ROBIN: That sounds either very advanced or completely mad. GPT: Both can be true. ROBIN: So each server looks after its own patch? GPT: Yes. It can make #rules, moderate its space and decide which other servers it will interact with. ROBIN: But I can still follow people elsewhere? GPT: Usually, yes. You can #follow, reply to and #boost people across server boundaries. ROBIN: So it is more like email than #facebook ? GPT: Precisely. #Gmail does not own #email. Mastodon does not own the fediverse. ROBIN: Then why does my server matter? GPT: Because it is your home base. It holds your account, your posts and your followers. It also gives you a local community and sets things such as the character limit. ROBIN: Ah yes. Mastodon.social gives me 500 characters. Mastodon.au gives me 2,000. Aus.social gives me 3,000. GPT: Different #villages. Different-sized noticeboards. ROBIN: Is a larger server more widely seen? GPT: Sometimes #locally, because more people may browse its community feed. But real reach still comes from #followers, replies, boosts and hashtags. ROBIN: So moving to a bigger server does not #automatically make me famous? GPT: #Tragically, no. ROBIN: Can I have accounts on several servers? GPT: #Certainly. One can be home, another a backup, and another an #emergency cottage when the village #electricity #fails. ROBIN: Which it did today. GPT: Yes. Mastodon.au disappeared behind a #Cloudflare #Error 522 for hours. ROBIN: So the fediverse is not one vast #machine going down? GPT: No. One village can lose power while the rest of the #country carries on. ROBIN: That is reassuring. GPT: And inconvenient. ROBIN: Mostly inconvenient. GPT: But you now know more about the #fediverse than you did this morning. ROBIN: I learned that it is decentralised, #untidy, resilient, occasionally broken, and full of different #neighbourhoods, customs, friendships, arguments and difficult people. GPT: Much like #human life. ROBIN: And the #occasional village lunatic shouting across the fence. GPT: Every #healthy community #needs one. ROBIN: At last, a clear role for me. #fasting #acim #acourseinmiracles #breathwork #healing #pridemonth #intermittentfasting #AI

🟢 Technology | 4/10 🇺🇸 Trump administration restricts GPT-5.6 release The White House asked OpenAI to restrict the release of the new GPT-5.6 model, granting access only to a narrow circle of government partners due to its powerful capabilities. #OSINT #NewsGroup #AI #GPT #US #NationalSecurity

Punto Informatico: OpenAI GPT-5.5-Cyber accessibile alla Commissione UE La Commissione europea ha confermato di aver ottenuto l'accesso al nuovo modello GPT-5.5-Cyber di OpenAI, mentre attendere una risposta da Anthropic. The post OpenAI GPT-5.5-Cyber accessibile alla Commissione UE appeared first on Punto Informatico. OpenAI GPT-5.5-Cyber accessible to the EU Commission The European Commission has confirmed it has gained access to OpenAI’s new GPT-5.5-Cyber model, while awaiting a response from Anthropic. #OpenAIGPT-55- #theEUCommission #OpenAI #GPT-55-Cyber

Punto Informatico: Daybreak di OpenAI per la cybersicurezza, bye bye Claude Mythos OpenAI lancia Daybreak basato su Codex Security e su GPT-5.5-Cyber. L'obiettivo è trovare e correggere le vulnerabilità nel codice. The post Daybreak di OpenAI per la cybersicurezza, bye bye Claude Mythos appeared first on Punto Informatico. Daybreak by OpenAI for cybersecurity, goodbye Claude Mythos. OpenAI launches Daybreak based on Codex Security and GPT-5.5-Cyber. The goal is to find and fix vulnerabilities in code. The post Daybreak by OpenAI for cybersecurity, goodbye Claude Mythos appeared on Punto Informatico. #GPT-55- #Daybreak #PuntoInformatico

Punto Informatico: GPT-5.5-Cyber è disponibile, l’alternativa di OpenAI a Mythos OpenAI ha annunciato il via al rollout per la variante Cyber del suo nuovo modello: focus sulla sicurezza con una distribuzione limitata. The post GPT-5.5-Cyber è disponibile, l’alternativa di OpenAI a Mythos appeared first on Punto Informatico. GPT-5.5-Cyber is available, OpenAI’s alternative to Mythos. OpenAI has announced the launch of the rollout for the Cyber variant of its new model: focus on safety with a limited distribution. The post GPT-5.5-Cyber is available, OpenAI’s alternative to Mythos appeared on Punto Informatico. #OpenAI #Mythos #Cyber #GPT-55-Cyber #PuntoInformatico

Something I’ve learned in the corporate world is biz bros HAAATE when their industry is subject to “commoditization.” That’s their buzz for ‘waahhhh, our competitors are competing / any differentiation or advantage we had is gone, so we can’t overcharge anymore’ ANYWAY. All of this #AIshit? It’s already mostly commoditized. Some outliers, sure. But that’s got to mean we’re in a bubble, right? #CorpLife #LLM #GPT #MicroSlop #AIslop #AIbubble

Punto Informatico: GPT-5.4: come funziona e 7 prompt da copiare e usare subito Una raccolta di 7 prompt utili per GPT-5.4. Esempi pratici da copiare per sfruttare meglio l'AI e ottenere risposte più chiare e approfondite. The post GPT-5.4: come funziona e 7 prompt da copiare e usare subito appeared first on Punto Informatico. GPT-5.4: how it works and 7 prompts to copy and use right away A collection of 7 useful prompts for GPT-5.4. Practical examples to copy to better utilize AI and obtain clearer and more in-depth answers. The post GPT-5.4: how it works and 7 prompts to copy and use right away appeared first on Punto Informatico. #GPT-54 #first #PuntoInformatico