#нейросети
-
By Default, Twitch Hands Streamers' Content to Amazon for AI Training
Twitch will use recordings of broadcasts to train Amazon's generative models — and creators are enrolled by default. The only way out is to opt out manually, and one of the platform's top executives was refreshingly honest about why.
-
Mistral Patents Tool Calls via Executable Code
The U.S. Patent Office has granted Mistral AI a patent on a method in which a language model composes and executes code to call external tools. The application describes a technique many consider standard industry practice.
-
ByteDance Is Building a 10-Trillion-Parameter AI Model
TikTok's owner has begun training one of the world's largest neural networks in a bid to challenge Anthropic and outpace its Chinese rivals.
-
OpenAI Updates ChatGPT: GPT-5.6 and Unlimited Chat for Free Users
Paying subscribers get the more reliable GPT-5.6 Sol model, while free users get unlimited text conversations with GPT-5.6 Luna and an “intelligence” control slider.
-
WeatherNext: DeepMind's AI Gives You an Extra Day to Prepare for Cyclones
Google DeepMind has published a paper in Nature about WeatherNext, a model that predicts the track and strength of tropical cyclones at a world-record level — and has open-sourced it.
-
Mistral Releases Shieldstral: An Open-Weight 3B Content Moderation Model
Mistral has unveiled Shieldstral, a compact safety classifier that takes moderation policies in natural language at runtime and, according to its developers, holds its own against models seven times larger.
-
Pippa Pays Artists for Their Style — but Is That Enough to Reconcile Them with AI?
The startup Pippa offers illustrators royalties for every image and video it generates, promising “ethical” generative AI. Yet its models are still trained on data scraped from the web without the creators’ consent.
-
DeepSeek Releases V4-Flash in Public Beta with a Focus on Agents
The updated deepseek-v4-flash model outperforms the Pro preview on agentic benchmarks and gains support for the Responses API format along with Codex adaptation.
-
The LLM2HUMAN Clinic: Turning Language Models Into Living People for $19.95
A satirical site in the aesthetic of GeoCities promises neural networks arms, legs and the legal right to order a sandwich — while slyly poking fun at the whole AI industry.
-
ChatGPT Stops Cloning the Style of Famous Writers
OpenAI has taught its chatbot to refuse outright imitation of authors' voices, offering instead to capture a “general feeling.” Behind this seemingly minor legal nuance lies a defense against copyright infringement suits.
-
Black Forest Labs Unveils FLUX 3 — a Multimodal Model for Images, Video and Sound
Germany's Black Forest Labs has opened early access to FLUX 3 — a single model that learns from images, video and audio at once in order to build a picture of the physical world.
-
InfiniteDiffusion: a diffusion model for generating infinite landscapes
The InfiniteDiffusion project uses diffusion models for the procedural generation of open-world landscapes. Its developers aim to combine the visual quality of neural-network generation with the practicality of procedural methods.
-
Anthropic explores a “global workspace” in language models
Anthropic has published research on the concept of a global workspace as applied to language models. The piece drew notable attention on Hacker News.