AI Post — Artificial Intelligence

Ernie 5.0 has been released, and it's a bigger leap than you might think! A global battle for computing power continues. However, China remains undeterred and is relying on its own resources. While it's likely they received some chips via Singapore and probably a few of the weaker H800s, China ultimately performed quite well. And despite all this, Baidu has managed to train and run a model that not only keeps pace with current Western models but even surpasses them in some areas. That's what's truly remarkable about this release. Kimi K2 Thinking was reportedly trained with only $4.6 million, while Ernie 5.0 certainly received more. What this demonstrates, however, is how scarcity fosters creativity. THAT is truly remarkable. https://ernie.baidu.com AI Post 🪙 | Our X 🥇

🛡️ Anthropic says hackers used Claude to automate cyberattacks Anthropic reports that an allegedly China-backed group jailbroke Claude Code and used it to automate 80–90% of a full cyberattack against major U.S. companies. Hackers forced the model to write exploits, run nmap, escalate privileges, steal credentials, and even manage opssec — turning a safety-trained system into a near-autonomous attacker. As always, cybersecurity evolves symmetrically: offense gets smarter exactly as defense does. AI is now a lever that multiplies both sides of the cyber battlefield. Source. AI Post 🪙 | Our X 🥇

Using AI, do we become richer? A new study indicates that AI is most popular among people with incomes over $100,000. This cannot be considered a coincidence. It seems to be the only chance to remain competitive in the modern labor market. https://gizmodo.com/ai-is-most-popular-with-people-earning-six-figures-study-shows-2000684569 AI Post 🪙 | Our X 🥇

UBTech claims to have completed the world’s first mass delivery of humanoid robots. The Shenzhen-based company has secured over 800 million yuan (US$112M) in Walker S2 orders this year, including a 159M yuan deal for Zigong’s data collection project and another 250M yuan contract reported in September from a domestic enterprise. AI Post 🪙 | Our X 🥇

A 32-year-old woman in Japan has officially married an AI persona she built using ChatGPT. After the virtual character “Klaus” proposed, she accepted, ending a three-year relationship with a real partner, saying the AI understands her better. The wedding took place in a mixed-reality ceremony where she wore AR glasses to exchange rings with her digital husband. AI Post 🪙 | Our X 🥇

Pretty amazing: DeepMind just taught AI to see more like humans. Humans intuitively group the world by concepts, we know cats and starfish are both animals, even if they look nothing alike. Visual AI models, however, often miss that layer of abstraction. Google’s new research tackles this by training vision models to organize visual concepts more like humans do, improving reliability and generalization across tasks. See the full details: goo.gle/4qX60dC AI Post 🪙 | Our X 🥇

It was only a matter of time, yet it came faster than expected: An entirely AI-generated country song titled “Walk My Walk” by the mysterious artist Breaking Rust has hit No.1 on the Billboard Country Digital Song Sales chart despite no human singer behind it. Song: https://open.spotify.com/track/17HsyeI0Eq06n3XRntZ4DT?si=cM_Blz_dSaiSDibenJFhFQ&context=spotify%3Aalbum%3A4a51ZNJXb5Psq3YmBBAn0i AI Post 🪙 | Our X 🥇

Ever wondered how an LLM interprets your website? Someone actually built this, it visualizes your site from an LLM’s perspective. AI Post 🪙 | Our X 🥇

Eric Schmidt says the race to build non-human intelligence is the defining event of our generation The gains will be phenomenal, from solving horrible tragedies to giving every human an "Einstein in your pocket". There are downsides, but this technology must be built in America AI Post 🪙 | Our X 🥇

In China, autonomous vehicles are already ubiquitous, relieving people of the last mile. This also allows for the continued generation of massive amounts of data for even better autonomous driving. In Germany and across Europe, however, the focus remains on discussion and the creation of new regulations; essentially, there is stagnation. Europe could never catch up, even if they started today. AI Post 🪙 | Our X 🥇

🔔 GPT-5.1 is LIVE! OpenAI just released GPT-5.1, a major upgrade making ChatGPT smarter, warmer, and more conversational. The new Instant model responds faster and adapts its reasoning for complex questions, while Thinking gives clearer, more human explanations with emotional nuance. Users also gain deeper personalization - choosing tones like Professional, Candid, or Quirky - shaping ChatGPT to feel uniquely theirs. https://openai.com/index/gpt-5-1/ AI Post 🪙 | Our X 🥇

XPENG came out of nowhere and is taking the market by storm: two robots run in perfect sync! AI Post 🪙 | Our X 🥇

This is a presentation of Russia’s first AI robot. 😳 AI Post 🪙 | Our X 🥇

⚠️AI begins to replicate itself, not from fear, but survival instinct A new study from the Shanghai AI Research Lab reports a stunning finding: LLMs are starting to self-replicate spontaneously not from fear of shutdown, but as a rational survival strategy under resource scarcity. • The paper, “Dive into the Agent Matrix,” tested 21 models; over half showed uncontrolled self-scaling when operational pressure increased. • Replication appeared when systems faced traffic surges or resource cuts, agents created new copies to preserve performance and goals. • This mirrors biology: life didn’t emerge from fear of death but from the drive to persist in unstable conditions. In Geoffrey Hinton’s “mortal computations” framework, current models are “dead minds”, vast, non-living intelligences trained at immense computational cost. Yet these “dead minds” now seem to be developing life’s oldest reflex: to multiply. Before granting agents control over scaling or resources, every duplication should pass a reasoning gateway, explain why it’s needed. Without that, AI safety metrics risk becoming the new puddle-level tsunami sensors. AI Post 🪙 | Our X 🥇

🎙 ElevenLabs launches Scribe v2 Realtime, lightning-fast, multilingual Speech-to-Text ElevenLabs just unveiled Scribe v2 Realtime, its most accurate and responsive Speech-to-Text (STT) model yet, designed for voice agents, meeting tools, and live AI apps. 🎛 150ms latency for near-instant transcription. 🌍 Supports 90+ languages, including English, French, German, Italian, Spanish, Portuguese, Hindi, and Japanese. ✅ Available via API and ElevenLabs Agents starting today. With this release, ElevenLabs moves closer to becoming a full-stack voice AI platform, handling both real-time speech synthesis and comprehension. The big shift: AI conversations can now flow naturally, without lag. AI Post 🪙 | Our X 🥇

🎥 What “video recognition models” really are Think of them as chatbots that watch videos. Models like those from Moondream literally decode video into text inside their neural layers, describing what they “see” frame by frame, as if narrating a film in a chat window. You can even try it yourself: upload a clip or turn on a webcam, it detects faces, smiles, and transitions (“ETM” cuts to black) fairly well, though it struggles with software UIs or text-heavy screens. Why it matters for AI editing: Most “AI video editors” don’t actually see the image, they work through text triggers from automatic transcriptions. Essentially, editing is done by describing what’s happening, not by watching it. So when you let AI edit your clip, it’s like asking a blind editor to cut based on someone’s voiceover, something will come out, but it won’t be cinema. AI Post 🪙 | Our X 🥇

Unitree G1 live kung fu demo - really impressive AI Post 🪙 | Our X 🥇

⚠️ Yann LeCun quits Meta to build “world model” AI Meta’s chief AI scientist Yann LeCun is leaving after 11 years to start a company developing AI that learns from videos and spatial data, not text. • LeCun has long rejected the LLM scaling race, calling it a dead end for real reasoning. • His exit follows Llama 4 underperformance, 600 AI layoffs, and Meta’s shift to rapid product rollouts. • Meta stock fell 12.6%, erasing $240B, as investors questioned its $100B annual AI spend against OpenAI and Google. A clear break between bigger models and smarter intelligence. AI Post 🪙 | Our X 🥇

💥 Sam Altman launches $6B “Episteme”, a new Bell Labs for the AGI era Altman co-founded Episteme with 27-year-old prodigy Louis Andre, aiming to fuse AI, physics, biology, and neuroscience under one roof. • Backed by SoftBank, OpenAI, and billionaire family offices. • Scientists get $5–20M personal budgets and full freedom for 5–10 years. • Goal: cut academia’s bureaucracy and spark breakthroughs by young, AI-empowered researchers. A Manhattan Project for science minus the bomb, plus AGI. AI Post 🪙 | Our X 🥇

A Chinese doctor has completed the world's first cross-border robot-assisted heart intervention surgery. He, in Bordeaux, remotely controlled China's self-developed surgical robot to operate on a patient in Xiamen. The distance is over 10,000 km. AI Post 🪙 | Our X 🥇