AI tagged posts

Unified memristor-ferroelectric memory developed for energy-efficient training of AI systems

A unified memristor-ferroelectric memory for the energy-efficient training of AI systems
Credit: Dupouy/CEA

Over the past decades, electronics engineers have developed a wide range of memory devices that can safely and efficiently store increasing amounts of data. However, the different types of devices developed to date come with their own trade-offs, which pose limits on their overall performance and restrict their possible applications.

Researchers at Université Grenoble Alpes (CEA-Leti, CEA List), Université de Bordeaux (CNRS) and Université Paris-Saclay (CNRS) recently developed a new memory device that combines two complementary components typically used individually, known as memristors and ferroelectric capacitors (FeCAPs)...

Read More

Ultracompact semiconductor could power next-gen AI and 6G chips

chip
Credit: CC0 Public Domain

A research team, led by Professor Heein Yoon in the Department of Electrical Engineering at UNIST has unveiled an ultra-small hybrid low-dropout regulator (LDO) that promises to advance power management in advanced semiconductor devices. This innovative chip not only stabilizes voltage more effectively, but also filters out noise—all while taking up less space—opening new doors for high-performance system-on-chips (SoCs) used in AI, 6G communications, and beyond.

The new LDO combines analog and digital circuit strengths in a hybrid design, ensuring stable power delivery even during sudden changes in current demand—like when launching a game on your smartphone—and effectively blocking unwanted noise from the power supply.

What sets this developmen...

Read More

Size doesn’t matter: Just a small number of malicious files can corrupt LLMs of any size

Size doesn't matter: just a small number of malicious files can corrupt LLMs of any size
Overview of our experiments, including examples of clean and poisoned samples, as well as benign and malicious behavior at inference time. (a)DoS pretraining backdoor experiments. Credit: arXiv (2025). DOI: 10.48550/arxiv.2510.07192

Large language models (LLMs), which power sophisticated AI chatbots, are more vulnerable than previously thought. According to research by Anthropic, the UK AI Security Institute and the Alan Turing Institute, it only takes 250 malicious documents to compromise even the largest models.

The vast majority of data used to train LLMs is scraped from the public internet. While this helps them to build knowledge and generate natural responses, it also puts them at risk from data poisoning attacks...

Read More

Novel antibiotic targets IBD—and AI predicted how it would work before scientists could prove it

Two researchers pose in a university laboratory.
McMaster graduate student Denise Catacutan (left) and assistant professor Jon Stokes (right) have discovered a new antibiotic — and they leveraged cutting-edge AI to determine how it works.

Researchers at McMaster University and the Massachusetts Institute of Technology (MIT) have made two scientific breakthroughs at once: they not only discovered a brand-new antibiotic that targets inflammatory bowel diseases (IBD), but also successfully used a new type of AI to predict exactly how the drug works. To their knowledge, this is a global first for the AI.

Detailed in the journal Nature Microbiology, the discovery unveils a promising new treatment option for millions of people affected by Crohn’s disease and other related conditions, while also showcasing important new applications fo...

Read More