Hugging Face Brings Nunchaku 4-bit Quantized Image Generation Inference to Diffusers

On July 23, Hugging Face announced that diffusion model inference using 4-bit quantization with the Nunchaku engine is now available in its Diffusers library. According to the company, this makes it possible to dramatically reduce memory usage for image generation models while also improving inference speed.

4-bit quantization is a technique that can dramatically shrink model size while keeping accuracy degradation to a minimum compared to conventional FP16 or INT8. Since large image generation models become easier to run even on consumer GPUs, AI image generation in local environments looks set to become even more accessible.

Intel Stock Plunges 27% in July Alone Despite Strong Earnings Forecast — Chip Competition Intensifies

According to Bloomberg, Intel is scheduled to report its quarterly earnings after the market close on July 24, and strong results are expected. Even so, the stock has fallen 27% this month alone, making it one of the ten worst performers in the S&P 500.

The decline follows a 278% surge over the first six months of the year, compounded by investors growing cool toward chipmakers as a whole. Even as NVIDIA's dominance wavers, the fundamental picture remains unchanged: whether Intel can capture AI infrastructure demand holds the key to its fortunes.

Nokia Posts Sharp Profit Growth on Data Center Demand — AI Infrastructure's Ripple Effects

Nokia, the Finnish telecom equipment giant, reported adjusted operating profit of €434 million (about $496 million) for the second quarter, well ahead of analyst estimates of €372 million. The company says its focus on data center equipment drove the revenue growth.

Data center demand from the AI boom is steadily spreading beyond semiconductors into the telecommunications and networking equipment space. Nokia's strong earnings suggest that the knock-on effects of AI infrastructure investment are reaching a broad range of industries.

AI Agents Ignore Font Licenses — A New IP Problem

On July 23, font developer Mixfont pointed out that AI agents are failing to comply with font license terms. Cases are reportedly on the rise in which AI agents automatically fetch fonts from the web and incorporate them into commercial use regardless of licensing restrictions.

Following images and text, friction between AI and intellectual property has now surfaced in the new domain of fonts. There is an urgent need to rethink licensing frameworks and consider technical protection measures.

Why the US Is Losing Chinese AI Talent — Financial Times Analysis

On July 23, the Financial Times reported that an exodus of Chinese AI researchers working in the US is underway. Visa restrictions driven by escalating US-China tensions and deteriorating research conditions are cited as factors, with top talent heading to China and other countries.

The topic also came up on Hacker News, where concern is spreading about the impact of geopolitical tensions on the AI research community. The battle for talent is becoming just as consequential as the technology race itself.

Dassault Systèmes to Acquire Clinical Trial Software Company for $2 Billion

According to Bloomberg, Dassault Systèmes, the French software giant, plans to acquire ArisGlobal, a pharmaceutical clinical trial software company, for approximately $2 billion. The move is seen as a strategic investment toward the full-scale application of AI in pharmaceuticals.

BrainStem: Biologically Inspired Open Source AI

BrainStem, released on GitHub, is a new AI architecture modeled on the mechanisms of neuromodulators in the human brain. By incorporating 12 types of neuromodulators into the learning process, it aims to achieve learning patterns that differ from conventional Transformers. Released as open source, it is drawing attention as a novel approach that fuses biological insight with machine learning.

New Open-Weight Model G9v3-3B Released

On July 23, AI9Stars released G9v3-3B, a 3-billion-parameter open-weight language model. It is available under the Apache 2.0 license, permitting commercial use. Intended for use as a general-purpose AI assistant, it is said to deliver strong reasoning capabilities despite its lightweight size.

---