Machine learning news — deep learning, reinforcement learning, neural architecture search, diffusion models, and new ML frameworks and libraries.
A developer's guide explains why edge AI benchmarking on Android requires measuring thermal throttling, memory pressure, and accuracy loss—not just execution time—and recommends using Kotlin 2.x with AICore for productio…
PrismML released Bonsai 27B, a 27-billion-parameter AI model compressed to 3.9 GB that runs on an iPhone 17 Pro Max at 11 tokens per second, the first model at that capability tier to fit on a smartphone. The ternary var…
Brown, a key figure in launching one of Amazon Web Services' earliest offerings and overseeing its compute and machine learning divisions, has left AWS, creating a leadership void in the cloud giant's AI strategy. His te…
A new mechanistic interpretability study of the MAIA 3 chess transformer finds that knight-fork detection is primarily assembled compositionally from check and queen-attack subcomponents, with causal ablation evidence ty…
Researchers at Meta, Stanford University, the University of Tokyo, and France's École Normale Supérieure developed the EgoBabyVLM Challenge, a test that reveals cutting-edge vision language models fail to match the learn…
Tim Dettmers et al. in their 2022 LLM.int8() paper found that large outlier features emerge across all transformer layers as models scale, causing naive INT8 quantization to break down. They introduced vector-wise quanti…
Built Technologies, a real estate finance software provider processing over $500B in projects, deployed an AI-powered document processing engine on Amazon Bedrock and the AWS Intelligent Document Processing Accelerator t…
PyTorch 2.13, released this week, brings FlexAttention support for Apple Silicon with up to 12x speedup on sparse patterns and a new fused LinearCrossEntropyLoss operator that cuts peak GPU memory by up to 4x during larg…
Amazon Web Services (AWS) introduced a solution to monitor Amazon SageMaker Pipelines across multiple AWS accounts and Regions using custom Amazon CloudWatch dashboards. The serverless, event-driven architecture uses a h…
A diffusion model's creativity is a mathematical consequence of neural networks learning a 'smoothed' version of the score function, driving the model to interpolate between training data points along the hidden data man…
MarkTechPost published a tutorial on building a Gin Config controlled PyTorch pipeline that separates training code from experiment configuration using .gin files. The pipeline implements a configurable MLP with scoped a…
A 26-billion-parameter Gemma 4 mixture-of-experts model can run at reading speed on a dual-socket 2013 Ivy Bridge Xeon server with no GPU, achieving ~5.2 tokens per second decode on a Q8_0 quant, according to community r…
A new automated pipeline using SSH reverse tunnels and self-healing ADB scripts eliminates the manual three-tier network bottleneck in remote-to-edge AI benchmarking, enabling developers to trigger on-device benchmarks f…
Livestream shopping app Whatnot acquired Shaped, a machine learning company specializing in real-time recommendation and search systems, to strengthen its discovery and personalization capabilities. The deal aims to solv…
Dave Treadwell, current Senior Vice President of eCommerce Foundation, will move to AWS effective August 1 to lead the AWS Compute and Machine Learning Services team, replacing Dave Brown, who is leaving after nearly 19 …
Phoronix benchmarked Ubuntu 26.04 LTS, Windows 11, and CachyOS on the $5399 Razer Blade 18 with Intel Core Ultra 9 290HX Plus and NVIDIA GeForce RTX 5090, finding Ubuntu competitive but CachyOS with newer kernel and driv…
Arm architecture powers the world's most advanced and energy-efficient supercomputers, with the TOP500's new #1 system LineShine becoming the first to exceed two exaflops of sustained double-precision performance using A…
A developer outlines a population health risk stratification pipeline for Medicare Advantage plans, emphasizing a layered signal approach combining RAF scores, chronic conditions, and utilization data. The pipeline ranks…
A poll by the Computer Entertainment Supplier's Association (CESA) found that 100% of Japanese online game developers are using AI, primarily for user preference analysis and user behavior prediction. The survey, conduct…
The PyTorch-Triton 3.7 release introduces the Triton Plugin Extensions system, a framework for dynamically loading custom compiler passes, dialects, and DSL extensions into upstream Triton at runtime without forking or r…