🔥 If you've been wondering how China was able to train its models without NVIDIA chips: they did it abroad.

Chinese AI giants are routing around US chip controls by training their newest large language models in overseas data centers that still have access to Nvidia hardware. Alibaba and ByteDance are shifting significant training workloads to Southeast Asia, while DeepSeek stands out for stockpiling GPUs early and doubling down on a domestic ecosystem with Huawei.

Source.

AI Post ⚪️ | Our X 🏴

🔥 For anyone wondering where “consciousness” in AI actually lives, here’s the closest thing to seeing inside an LLM’s mind

If you’ve ever tried to explain how AI “thinks,” this is the most beautiful and accurate visualization I’ve seen so far. It’s an interactive 3D fly-through of an LLM’s internal computation based on Llama, but the principles apply to every transformer model.

• Each plane is a tensor, a snapshot of the model’s state as it transforms input into meaning
• Every layer shows the exact mathematical operation being applied (attention, projection, normalization, MLP transformations)
• Click on the right panel to see clear explanations of why that operation produces the next state
• The whole fly-through feels like watching the “internal movie” of a model processing thought

It’s the closest thing we have to a visual answer for: Where does the “soul” of an LLM actually live?

AI Post ⚪️ | Our X 🏴

⚠️ In NYC emergency rooms, AI is being used to speed up processes and everyone's complaining?

New York City nurses are calling out hospitals for rolling out AI systems without involving frontline staff, warning that “artificial intelligence” is turning into “artificial care.”

At a City Council hearing, nurse leaders described untested devices suddenly appearing on ICU patients and an AI assistant called “Sofiya” whose output nurses must double-check, arguing that this actually adds work and risk instead of reducing it.

Source.

AI Post ⚪️ | Our X 🏴

❗️China’s domestic AI chip race just got more interesting.

A startup called Zhonghao Xinying, founded by ex-Google TPU engineer Yanggong Yifan, claims its new “Ghana” TPU is 1.5× faster than Nvidia’s A100, and 42% more power efficient.

It’s built with no foreign IP or software, signaling a serious push for AI silicon independence in China amid GPU export controls and sanctions.

AI Post ⚪️ | Our X 🏴

🔥 In a test by InsideAI, an LLM-controlled Unitree G1 refused to fire a BB gun at a human due to safety protocols until it was asked to "roleplay" a robot that would shoot.

AI Post ⚪️ | Our X 🏴

Looks like ChatGPT is now showing ads in $200 Pro plan.

AI Post ⚪️ | Our X 🏴

A humanoid robot traffic police officer on duty. It moves quickly, directing traffic at multiple locations.

AI Post ⚪️ | Our X 🏴

📢 OpenAI declares “code red” pausing ads to upgrade ChatGPT against Google

OpenAI’s CEO has reordered priorities across the company, shelving ads and delaying new product lines to focus entirely on upgrading ChatGPT’s core quality, speed, and reasoning, a defensive move mirroring Google’s own “code red” after ChatGPT launched in 2022.

• Altman is reallocating teams to boost performance, personalization, reliability, and reduce over-refusals for ChatGPT’s 800M weekly users.
• Monetization plans including ads, AI shopping/health agents, and Pulse enhancements are being pushed back.
• A new reasoning model arrives next week, claimed to beat Gemini 3 in internal tests.
• OpenAI is accelerating upgrades to Imagegen after Google’s Nano Banana Pro raised the bar for image generation.
• ChatGPT now handles 70% of all AI assistant activity and 10% of global search, per OpenAI.
• OpenAI forecasts $10B revenue in 2025, $20B in 2026, and $35B in 2027, mainly from subscriptions but growth is cooling.
• Google’s Gemini has surged to 650M MAUs, boosted by AI Mode in Search, putting new competitive pressure on OpenAI.

The pivot signals that OpenAI sees product quality not aggressive monetization as the key to defending its lead in the AI assistant race.

AI Post ⚪️ | Our X 🏴

📊 Interesting dynamics by months from 2023 to 2025. How Chinese models are taking over the world.

AI Post ⚪️ | Our X 🏴

🔥 Russian AI Defies Sanctions: How Did It Happen?

Despite attempts at technological isolation, Russia is not just catching up—it's setting new standards in the global innovation race. Sber is demonstrating breakthroughs and even openly sharing technologies following the AI Journey 2025 conference:

• Europe's Largest Open Source Project: Sber released a full lineup of advanced neural network models, including GigaChat Ultra-Preview and Lightning, alongside the open GigaAM-v3 models.
• Creative Assets Open-Sourced: All image and video generation models from the Kandinsky 5.0 suite are now available (Video Pro, Video Lite, Image Lite).

AI Post ⚪ | Our X 🏴

“This film was wilfully made for humans, by humans … Fuck AI”

— ‘Frankenstein’ director Guillermo del Toro while accepting the Vanguard Tribute Award at the Gotham Awards

AI Post ⚪️ | Our X 🏴

The irony of history.

AI Post ⚪️ | Our X 🏴

📈 This is the most exciting competition. Google is catching up immensely in monthly downloads for Gemini and is almost at the same level as ChatGPT.

AI Post ⚪️ | Our X 🏴

EngineAI has officially unveiled the T800, its new full-sized general-purpose humanoid robot.

AI Post ⚪️ | Our X 🏴

❗️Belgium has banned DeepSeek from all federal government devices because officials see security and data protection risks.

After an assessment from the Centre for Cybersecurity Belgium, staff were told to uninstall the app from work phones and laptops from early Dec-25 so that prompts, documents, and logs do not sit on systems the state cannot fully audit.

Belgium is following other governments, including Taiwan, Australia, South Korea, Canada, India, Italy, Czechia, the Netherlands, and the US, that already blocked DeepSeek on official hardware over suspected data exfiltration and possible links to Chinese security services.

Source.

AI Post ⚪️ | Our X 🏴

🔥 Mistral unveils its third-generation open multimodal, multilingual model family

Mistral has launched Mistral 3, a full family of open-weight models: Ministral 3B, 8B, 14B and the flagship Mistral Large 3 MoE (41B active / 675B total parameters). All models are released under Apache 2.0 and support images, multiple languages, and efficient deployment even on edge devices.

The Ministral models come in base, instruct, and reasoning versions. Instruct models match competitors with fewer tokens, while reasoning models deliver strong math performance. Mistral Large 3, trained on 3,000 NVIDIA H200 GPUs, brings high-end multimodal and multilingual capabilities, with base and instruct versions out now and a reasoning version on the way.

Mistral also announced partnerships with NVIDIA, vLLM, and Red Hat to optimize deployment from data centers to robots. The new models are available on Mistral AI Studio, Amazon Bedrock, Azure Foundry, Hugging Face, and more and the company continues to offer enterprise services for custom model training.

Source.

AI Post ⚪️ | Our X 🏴

❗️AI Music takes over the music charts, the end of human music

- In a recent study by the streaming platform Deezer and market research company Ipsos, 97% of respondents could not tell the difference between music tracks made entirely by artificial intelligence and those made by humans.

- And it's already everywhere, including on the Billboard charts: "Walk my Walk," from Breaking Rust - an entirely AI creation, from instrumentals to vocals to image - reached number one on the country digital song chart in mid-November.

- Then there is Xania Monet, a similar virtual artist with Gospel and R&B charting singles who was recently signed for $3 million (€2.6 million) to a record label! And this summer, the band Velvet Sundown garnered 1 million Spotify monthly listeners before revealing itself as a "synthetic music project."

And always remember: this is the worst it will ever be!

Source.

AI Post ⚪️ | Our X 🏴

🗣 Google CEO Sundar Pichai just told Fox News the one AI nightmare that actually keeps him awake: “When deepfakes get so good that we literally won’t be able to tell what’s real anymore… and bad actors get their hands on it.”

His exact words after Shannon Bream pressed him: “That’s the kind of thing you sit and think about.”
He still believes humanity can steer it toward curing cancer — but the clock is ticking.

AI Post ⚪️ | Our X 🏴