Tien Kung humanoid robot wins Beijing’s half marathon, finishing 21km in 2h 40m with three battery swaps. Developed by state-backed X-Humanoid, a venture including Xiaomi and UBTech, aims to become “Android” of robotics and seeks to advance core platforms. Source:…
Adobe’s new Content Authenticity app helps tackle the rising issue of distinguishing real photos from AI-generated ones. Launched as a free public beta, it lets creators apply Content Credentials, securing cryptographic metadata for attribution. Users can batch-apply these credentials, ensuring…
Hugging Face engineer Julien Delavande developed a chatbot tool to estimate the energy consumption of individual AI prompts. The tool displays energy usage in relatable terms, like phone charge percentage and LED bulb time, highlighting the significant power demands of…
Perplexity’s CEO, Aravind Srinivas, revealed plans for its browser, Comet, to track all user activity online to deliver hyper-personalized ads. Launching in May, Comet aims to rival Google by understanding user behavior beyond its app, hoping users will appreciate more…
Ziff Davis, owner of CNET, IGN, and more, is suing OpenAI for copyright infringement, alleging it copied extensive content from its outlets without consent. Despite web crawler restrictions, OpenAI purportedly used this content for AI training. Ziff Davis seeks a…
Jericho Security has raised $15 million in Series A funding to expand its AI-driven cybersecurity platform, tackling the growing threat of deepfake fraud, which cost businesses $200 million in 2025. Their innovative approach, using adaptive AI, prepares organizations for evolving…
Anthropic CEO Dario Amodei aims to demystify AI models by 2027, highlighting the need for deeper interpretability to understand their decision-making processes. Despite early breakthroughs, Amodei calls for industry collaboration and regulation to ensure AI models are transparent and safe,…
A companion isn’t just an app—it’s a new way to interact with computing, like a friend.
Except for foundational model companies, the top 30 AI unicorns are less than 500 people businesses!
DeepSeek has launched FlashMLA, an open-source MLA decoding kernel optimized for Hopper GPUs. It supports BF16, offers high performance with 3000 GB/s memory bandwidth and 580 TFLOPS on the H800 SXM5 GPU. Designed for efficiency, it manages variable-length sequences and utilizes CUDA 12.6, enhancing AI calculations through advanced optimizations.
You must be logged in to post a comment.