765 новостей • Страница 7 из 39
AI-RAN от Samsung: почему встраивание ИИ в сотовые сети — это капитуляция
Успех Anthropic создаёт дефицит серверов — или маскирует кризис эффективности ИИ?
Anthropic finds AI isn’t replacing jobs yet, but early data shows growing inequality as experienced users gain an edge, raising concerns about future displaceme
Few TensoRF доказал: проблема 3D AI не в данных, а в архитектуре
Google Research : Google Research details TurboQuant, a quantization algorithm to enable massive compression of LLMs and vector search engines without sacri
Super Micro said Yih-Shyan 'Wally' Liaw, a co-founder, has resigned from the server maker's board after he was indicted in the U.S.
Learn how OpenAI’s Model Spec serves as a public framework for model behavior, balancing safety, user freedom, and accountability as AI systems advance.
Jennifer Elias / CNBC : A Jeffrey Epstein victim files a class action against the Trump administration and Google, claiming Google's search and AI Mode pub
Ivan Mehta / TechCrunch : Google launches Lyria 3 Pro music generation model, with better creative control and allowing users to create three-minute tracks
Google is launching Lyria 3 Pro, an upgraded music model that generates longer, more customizable tracks, as it expands AI music tools across Gemini, enterprise
Учёные научились читать мысли через ЭЭГ — но забыли спросить, зачем это нужно
Reuters : Norwegian startup Lace, which is building a form of lithography that uses a helium atom beam instead of light to create chip designs, raised a $40
Компьютерное зрение научилось видеть то, что камеры скрывают
Niko Gallogly / New York Times : Wearable maker Whoop raised $575M led by Collaborative Fund at a $10.1B valuation and says it hit $1B in ARR by 2025's end
Financial Times : European VC firm Hummingbird raised $800M to find “misfit” founders, taking its total assets to ~$2B, after backing Kraken and
Harry McCracken / Fast Company : An oral history of Apple's earliest days told by people who lived it, including Steve Wozniak, Ronald Wayne, Mike Markkula
Исследователи нашли способ починить RL для языковых моделей — и это плохая новость
Бенчмарки AI-моделей сломаны: исследователи нашли способ выявить испорченные тесты
Malicious litellm_init.pth in litellm 1.82.8 — credential stealer The LiteLLM v1.82.8 package published to PyPI was compromised with a particularly nasty cre
Auto-evolve: Enhancing large language model’s performance via self-reasoning framework - Amazon Science Research Research areas - Sustainability - Security,