LIVE PULSE
4.0 Anthropic CEO Amodei calls for slower AI development and shared safety rules — 11 src2.2 Agility Robotics unveils Digit 5 humanoid for warehouses and factories — 2 src2.0 Apple ships rebuilt Siri with Google Gemini, but not in the EU — 2 src1.8 Siri AI in macOS 27 Golden Gate: FAQ, Germany availability, privacy questions — 2 src1.4 Sam Altman says OpenAI will not go public in 2026, citing AI safety concerns — 5 src1.1 OpenAI contractors review real ChatGPT conversations to rate responses, report says — 2 src1.1 Anthropic data retention policy prompts firms to limit Claude use for sensitive work — 1 src1.1 Study examines issue bias in LLMs used as writing assistants before Swedish 2026 election — 1 src1.1 Study Audits Misalignment in Multi-Modal World Models — 1 src1.1 Retrieval-Grounded Reasoning Approach Proposed for Universal Multimodal Embeddings — 1 src4.0 Anthropic CEO Amodei calls for slower AI development and shared safety rules — 11 src2.2 Agility Robotics unveils Digit 5 humanoid for warehouses and factories — 2 src2.0 Apple ships rebuilt Siri with Google Gemini, but not in the EU — 2 src1.8 Siri AI in macOS 27 Golden Gate: FAQ, Germany availability, privacy questions — 2 src1.4 Sam Altman says OpenAI will not go public in 2026, citing AI safety concerns — 5 src1.1 OpenAI contractors review real ChatGPT conversations to rate responses, report says — 2 src1.1 Anthropic data retention policy prompts firms to limit Claude use for sensitive work — 1 src1.1 Study examines issue bias in LLMs used as writing assistants before Swedish 2026 election — 1 src1.1 Study Audits Misalignment in Multi-Modal World Models — 1 src1.1 Retrieval-Grounded Reasoning Approach Proposed for Universal Multimodal Embeddings — 1 src
HEATPULSEAI MAGAZINES
FLIP · FOLLOW · SAVE
Log inSign up

FP8

topic2 events
papersSEP 10 04:00 UTC

EFQ-Softmax Method Aims to Remove Exponentials from Attention Quantization

A new arXiv paper proposes EFQ-Softmax, a technique for quantizing the softmax step in Transformer attention without computing exponentials. Low-bit attention already speeds up inference by shifting matrix multiplications to FP8 or FP4 engines, but the softmax path often still needs higher precision for shifted-score exponentials and temporary probability storage. The work targets that remaining bottleneck in the attention pipeline.

quantization/transformers/attention/inference
papersSEP 10 04:00 UTC

Paper argues FP8 with Ozaki Scheme II can substitute FP64 on next-gen NVIDIA GPUs

An updated arXiv preprint contends that low-precision FP8 matrix operations, when combined with the CRT-based Ozaki Scheme II error-compensation technique, can handle numerical workloads traditionally reserved for double-precision (FP64) hardware. The authors focus on NVIDIA's B300-class AI accelerators, claiming their tensor cores make this approach viable for a range of matrix-dominated scientific computing fields. The paper is the first part of a series challenging the assumption that dedicated FP64 units are essential for high-performance computing.

hpc/fp8/gpus/nvidia