Meta rolls out Muse, a new AI image generator

Toward More Controllable AI Video Editing: An Early Research Exploration at Netflix

Netflix giới thiệu hai mô hình chỉnh sửa video AI giai đoạn đầu là Vera và VOID. Vera sử dụng mô hình diffusion phân lớp, chỉ tái tạo vùng chỉnh sửa (kèm alpha matte) thay vì toàn bộ clip, bảo toàn nội dung chưa chỉnh sửa. VOID chuyên xóa vật thể trong video với kỹ thuật inpainting hợp lý vật lý, tái tạo cảnh thực tế khi vật thể bị loại bỏ. Cả hai mô hình đều vượt trội so với các phương pháp hiện có trong nghiên cứu.

Lập trình viên muốn phát triển các giải pháp AI tiên tiến trong xử lý video nên tham khảo để hiểu cách thiết kế mô hình hiệu quả như Vera và VOID, từ kiến trúc đặc biệt đến kỹ thuật điều khiển chi tiết để nâng cao chất lượng và tính khả thi của các ứng dụng AI video trong tương lai.

South Korea’s ‘fake news’ law takes effect

South Korea's new law targeting false information online is now in force, allowing courts to award punitive damages up to five times proven losses against news outlets and large social media accounts — including YouTube creators — that spread false or manipulated content. Platforms with over a million daily users must remove flagged content upon reports, with fines up to 1 billion won for repeat offenders. Critics, including the Journalists Association of Korea and the US State Department, warn the law could chill press freedom and turn platforms into censors, since platforms — not courts — make the initial call on what counts as false. The law was passed following a political crisis involving YouTube election-fraud claims, and its enforcement raises broader questions about who gets to define misinformation, especially as generative AI accelerates the spread of fabricated content.

Meta debuts Muse Image, its first AI image model built under Alexandr Wang’s lab

Meta has launched Muse Image, its first AI image generation model developed under the Superintelligence Labs division led by Alexandr Wang. The model is rolling out inside Meta AI, Instagram, and WhatsApp, allowing users to generate images from text prompts, edit existing photos, and create images featuring friends based on their public Instagram posts. An opt-out option is available in settings, and all generated images include an invisible watermark. Advertisers will gain access soon for marketing material creation. This is the second major product from Wang's lab, following the Muse Spark LLM released in April. Meta also plans to monetize its AI infrastructure by selling model access and compute to outside developers via a cloud offering.

Over 100 US-built autonomous ATVs have been fighting in Ukraine for nine months

Forterra has deployed over 100 autonomous Lancer ATVs in Ukraine since October, completing 1,100 missions, carrying 777,440 pounds of cargo, and performing 52 casualty evacuations across 2,500 miles. The vehicles are based on Polaris ATVs with custom sensor and compute stacks, and are gas-powered with 750kg payload capacity. Despite the scale, soldiers still primarily teleoperate them in combat because the autonomy cannot yet identify and react to unexpected enemy threats. Forterra is working on combining classical robotics with generative AI for more generalised reactions. Cost remains a key barrier — the vehicles are too expensive to treat as disposable like drones. Competitors including Scout AI, Field AI, and Overland AI are also trialling UGVs with the US military.

Introducing Muse Image and Muse Video

Meta Superintelligence Labs has launched Muse Image and previewed Muse Video, its first media generation models. Muse Image operates as an agentic model that uses coding and search tools, performs self-refinement during generation, and scales with test-time compute. It supports multi-reference image composition, iterative editing, and integrates with the broader Meta ecosystem including Instagram and WhatsApp. Muse Video, built on the same pretraining base, offers competitive text-to-video generation with native audio support. Both models include Content Seal, an invisible watermarking system for AI provenance. Muse Image currently ranks No. 2 on Arena for text-to-image and editing benchmarks, while Muse Video ranks No. 3 for text-to-video.

The AI Big Crunch Is Starting

A critical look at the current AI hype cycle, arguing that organizations have confused adopting AI with understanding it. Drawing on examples like Ford rehiring engineers after AI fell short, Forrester data on regretted AI layoffs, and the book 'AI Snake Oil,' the piece contends that AI is a powerful tool but not a strategy. The 'AI Big Crunch' metaphor describes the inevitable correction from inflated expectations to practical, grounded use — a phase the author believes is now beginning.

Inside the Emerging AI Art Market

A growing but fragmented market for AI-generated art is taking shape, spanning NFTs, physical museum installations, and stock image platforms. Artist SHL0MS exposed anti-AI bias by selling a real Monet cropped of its signature as an NFT for $40,000. Refik Anadol's Dataland, the world's first generative AI museum in Los Angeles, sold 1,000 AI data sculptures at $5,000 each in 34 minutes. Stanford research shows AI images boosted stock platform sales 80% while crowding out traditional contributors. Curators like the Whitney's Christiane Paul draw a sharp line between prompt-generated images and serious AI art, which involves training custom models and building control systems — work they describe as far harder than traditional media.