NVIDIA AI News - Page 9 of 16

Imagine an AI that cheats, and then, without prompting, writes a full confession. OpenAI's new “Confessions” tool does exactly that.

The artificial intelligence arms race just got a little more personal. OpenAI drops GPT-5.2 into the agentic arena, and Runway finally unveils its...

NVIDIA dumped a massive pile of training data onto the internet. It’s a set of files meant to teach machines how to move through the physical world,...

The promise of a new AI model is worthless without a quick way to connect to it. Google just delivered one for Gemini 3 Pro.

The appetite for long-context language models is insatiable, yet the memory they consume grows faster than our hardware can keep up.

Nvidia has just given every other model developer a headache. The company’s new Nemotron 3 Super is not merely faster.

Eight thousand one hundred and ninety two graphics cards. Nvidia marshaled them to train a single model, DeepSeek-V3 671B, for the latest MLPerf...

Banks are packed with useful data and crippled by it. Transaction logs, customer notes, and fraud alerts are all locked in separate systems, talking...

The sci-fi fantasy of shouting at floating holograms? Forget it. NVIDIA’s real play for augmented reality is silent. It’s almost mundane.

Three different companies decided what AI should be this week, and their answers have nothing in common.

Jensen Huang sat down with Lex Fridman for nearly two hours, and somewhere in that sprawling conversation, he had to defend the future of gaming...

AWS announced its Trainium3 UltraServers today. More importantly, the company has now deployed over one million of its custom Trainium accelerators.

NVIDIA just gave away a piece of its AI factory. At NeurIPS, they open-sourced the library used to manufacture the synthetic data that trains their...

Paytm is tired of waiting for Nvidia. The Indian payments giant is now sending its AI work to a hardware shop called Groq.

Nvidia won everything. The MLPerf Training 6.0 benchmark results are in, and the company's Blackwell platform took first place in every single test.

Nvidia has built a robot that reads academic papers and rewrites its own training code. It corrects itself.

Agentic AI is finally leaving the lab with a security badge pinned to its chest, but governance remains the loose thread.

Most algorithms branded as "fast" are just recompiled versions of old code; they hit a computational wall and fail.

What if a medical language model could deliver the reasoning power of 40 billion parameters while only waking up a fraction of its brain?

The countdown begins: Samsung has finally locked a date for its next Galaxy Unpacked event.