MarkTechPost (2 hours ago)
FlashMap is a speculative decoding drafter from Polymathic AI, Princeton and USC ISI researchers. It drafts 16 tokens jointly in one pass and reaches up to 6.14x lossless speedup on Qwen3-8B.
Read the full story on MarkTechPost.
Malaysian customs officers have seized 30 servers containing advanced AI chips at Kuala Lumpur International Airport (KLIA), the authorities said on Wednesday. The shipment was worth about RM49.5m ($12.1m), and it was moving through Malaysia without the required permit.
Workers tell WIRED that NOA—or “No Ordinary Assistant”—was unhelpful. Bestselling authors also say AI is ruining their relationships with Big Five publishers.
How AI is leading to weak measurement practices that impact the bottom line.
The base model—without thrusters or the flight system—will likely cost several times that amount to buy. “I’m going to do everything in my power to get the flying one,” says one customer.
The FIA, which governs motor sport worldwide, has named Amazon Web Services (AWS) its official partner for AI, cloud computing, machine learning and cloud infrastructure. The deal was announced today, and it covers every level of the sport, from club racing up to Formula 1.
See how Legora uses OpenAI models, ChatGPT, and Codex to accelerate legal workflows, expand contract analysis, and free lawyers to focus on higher-value work.
Google DeepMind released Genie 4 as a research preview. The real-time world model adds native audio, live dialogue and text steering, with Gemini 3.8 Flash planning events inside the simulation loop.
UC Berkeley’s LEANN vector index stores a pruned graph and recomputes embeddings at query time. It matches HNSW accuracy at 90% recall using 4 GB instead of 188 GB, at the cost of seconds-level search.
Liquid AI released LFM3, 6 open-weight 2B-class models for text, vision, audio, omni and decisions. LFM3-1.7B uses a tri-hybrid backbone, a 256K context window and fits in about 1 GB at 4-bit.
Mirendil released Fern 0.1, its first model, pretrained from scratch with a 10M-token context window. It targets agentic AI R&D work, scores 100% on RULER single needle at 10M, but trails frontier models on agentic coding.
The Guardian Technology
71 active sources
26 Turkish · 45 English
All sources
