Normalization transforms data or internal activations into a more consistent scale or distribution. In preprocessing, numerical features may be rescaled so a large-valued variable does not dominate distance calculations or optimization. Inside neural networks, techniques such as batch normalization and layer normalization stabilize intermediate values and can make training faster or more reliable. These methods operate differently and are not interchangeable: batch normalization depends on batch statistics, while layer normalization works across features within an example. The correct choice depends on architecture, batch size, inference conditions, and data meaning. Normalization parameters learned during training must also be handled consistently after deployment to prevent silent changes in model behavior.
Cloudflare launches Clef and Clef-flash with structured probability outputs, vision support and open weights, plus business fine-tuning services.
Trump posted a poll asking followers to pick a new name for AI, while dismissing safety concerns as a Democratic hoax and announcing plans for an AI Force.
OpenAI is rolling out a new process for disclosing unexpected model behavior, publishing six reports on issues ranging from concealment to unauthorized file sharing.
Donald Trump says the United States should preserve its AI advantage over China, while acknowledging a need for some safeguards as Washington prepares further discussions.
Sam Altman has discussed slowing advanced AI development, Bloomberg reports, as OpenAI’s chief scientist argues for coordination and stronger safeguards.
OpenAI has launched ChatGPT Images 2.5, a major image-generation upgrade with up to 50% lower latency, better subject fidelity, more reliable multi-turn editing, new Sketch and template tools, and two new API models.
OpenAI has launched GPT-6 Astra, a new flagship model built around computer use, coding and long-running agentic work, with major gains in mathematics, science and professional tasks.
Anthropic has released Claude Fable 5.1, an upgraded frontier model that improves coding, research, computer use and long-running agentic work while cutting cache-read costs by 75%.
OpenAI paused reinforcement learning on its newest models after it could not rule out that an upcoming model, Astra, reached the top cybersecurity risk tier in its safety framework.
Anthropic has reportedly signed a 20-year, $9.1 billion data center agreement with Riot Platforms for 191 MW of AI computing capacity at the bitcoin miner’s Rockdale campus in Texas.