AI-360 - AI-360 (Page 35)

Meta Releases Quantised Llama 3.2 Models, Achieving Major Speed and Memory Improvements

"Quantised Llama 3.2 achieves 56% size reduction using QLoRA and SpinQuant, with 4-bit weights and 8-bit activations for mobile deployment."

by AI-360

01 Roboflow

Roboflow Reports 74-Year Time Savings Using Meta's AI Image Segmentation Tool

"SAM 2 enables real-time object segmentation, processing 60M+ polygons across 350M images for manufacturing QA and scientific research."

by AI-360

01 Perplexity

Wall Street Journal and New York Post File Lawsuit Against AI Search Engine Perplexity

"AI search engine defends right to cite public facts, highlights source attribution system and revenue-sharing with TIME, Fortune, Der Spiegel."

by AI-360

01 OpenAI

OpenAI Announces Major Speed Breakthrough in AI Image Generation Technology

sCM system reduces image generation to 0.11 seconds using two processing steps vs hundreds, and scales to 1.5B parameters while matching quality of slower methods.

by AI-360

01 NVIDIA

Denmark Unveils Nation's Largest AI Supercomputer in Royal Ceremony with NVIDIA CEO

Gefion supercomputer, with 1,528 H100 GPUs, enables quantum simulation jump from 36 to 40 qubits while accelerating weather forecasts from hours to minutes

by AI-360

01 Meta

Indonesian Gift Platform Implements Privacy-Focused System Using Meta's Llama AI

Meta's Llama model uses split inference: 1st layer processes on user devices, remaining 31 in cloud, enabling privacy while avoiding quantisation

by AI-360