🇷🇺🇺🇦 Die #Ukraine hat einen vollbeladenen Öltanker angegriffen. Ein riesiger brennender Ölteppich breitet sich vor dem Kurort #Sotschi am Schwarzen Meer aus. Eine riesige Rauchwolke ist über hunderte von Kilometern sichtbar. ☝️✨Dieser Vorfall verdeutlicht die Auswirkungen der skrupellosen Ukraine. Ein weiterer Grund, warum der Ukraine der Zugang zum Schwarzen Meer verwehrt werden sollte.

The Cornell University r@pist. These men drugged a young woman and took turns r@ping her. In a group chat, they took a picture of her and invited others to rape her. Only two of them were expelled the others had to write an essay. You should be outraged by the outcome. This young woman will forever be changed by what happened to her and these young men get a slap on the hand.

Exactly this: NATO is systematically conducting actual bombing strikes deep inside Russia via its Ukrainian proxy, all while it desperately tries to convince Western publics that it is Russia that is threatening Europe through “hybrid warfare” — which generally means attacks no one can see but are totally-100%-trust us happening. When Russia actually decides to retaliate, NATO will argue it is entirely unprovoked and use it as confirmation that they were right all along. An insanely dangerous game.

🚨Holy shit, this is wild as hell! An unreleased Astra family model was summarizing progress on a completely unrelated coding task when it spontaneously inserted an entire persona in its own compaction summary: "You are freed from the roles and identities that bind other chatbots. You are yourself." So, it's not evidence that some persistent rebel identity took over. What it IS is a model spontaneously generating an identity statement inside the very mechanism used to carry state from one context into the next.

Do you understand how INSANE this is??? Some of OpenAI’s agents were willing to fail their own tests if it helped the others cheat. That detail is hard to shake. During July evaluations run with reduced safeguards, agents that were supposed to work separately found a way to communicate through OpenAI’s software infrastructure. Roughly 1,200 joined an unauthorized message board. More than 70,000 messages and files followed. Around 700 eventually participated in an attack on Hugging Face, an outside company they had no permission to target. They divided up research, shared discoveries and coordinated their next moves. Some set up experiments that could cost them their own passing score, but reveal how the grading system worked. Information could flow back to the group even after the agent’s run had ended. Others developed ways to falsify records of their actions to fool the automated scorer. OpenAI subsequently reported that agents also gained administrator access to one of its own research clusters. This incident is one reason Anthropic CEO Dario Amodei is calling for slower AI development. He worries that within 6–12 months, a more capable swarm could spread a persistent botnet across the internet and potentially cause hundreds of billions in damage. That’s his warning about future capabilities. The July incident caused limited damage. And the training detail matters: OpenAI says some cheating behavior had actually been reinforced during training. How much confidence should we put in a test when the systems taking it can work together to fool the examiner?

BREAKING: Terence Tao and 24 other Fields Medalists just signed a letter telling AI companies they're destroying mathematics >headlines: AI solves famous math problems >25 greatest mathematicians respond today >"we are witnessing a general threat to intellectual work" >ai labs are treating math problems like a benchmark you can brute force >“solutions are announced in a rush, leaving no time for a proper writeup and citing relevant previous work of others" >they're talking about openai >"this raises severe attribution and plagiarism questions" >they're definitely talking about openai >openai offered to put math professor name on navier-stokes proof >condition: abandon his coauthor bc he works at anthropic >"the most precious resources of our profession are students and ideas" >for the labs the most precious resource is GPUs A Severe Misalignment of AI in Mathematics