Tilde Research Introduces Aurora: A Leverage-Aware Optimizer That Fixes a Hidden Neuron Death Problem in Muon

Nous Research releases Contrastive Neuron Attribution (CNA), a method that identifies and ablates sparse MLP neuron circuits to steer LLM behavior — no sparse autoencoder training, no weight modification, and no degradation of general capability benchmarks. The post Nous Research Releases Contrastive Neuron Attribution (CNA): Sparse MLP Circuit Steering Without SAE Training or Weight Modification appeared first on MarkTechPost.

Let’s unpack what Demis Hassabis said at the end of yesterday’s Google I/O keynote. This is Optimizer, a weekly newsletter sent from Verge senior reviewer Victoria Song that dissects and discusses the latest gizmos and potions that swear they're going to change your life. This week's issue is a special early edition tied to The Verge's Google I/O coverage. You can expect our next issue at its usual time next Friday. Opt in for Optimizer here. Toward the end of this year's Google I/O keynote, Google DeepMind CEO Demis Hassabis declared, with a completely deadpan face, that the company hopes to "reimagine the drug discovery process with the goal of one day solving all disease." This is the sort of statement that the phras … Read the full story at The Verge.

This is Optimizer, a weekly newsletter sent every Friday from Verge senior reviewer Victoria Song that dissects and discusses the latest gizmos and potions that swear they're going to change your life. Opt in for Optimizer here. A few days ago, my esthetician was smearing hot wax on my face. The two caterpillars I call eyebrows were in desperate need of taming - as was my lady 'stache. I hate this monthly ritual, but facial hair is a sore spot. Hirsutism is perhaps one of the few visual indicators of a condition that's plagued me for the past decade. Until this week, I've always known it as polycystic ovary syndrome (PCOS). Normally, I sp … Read the full story at The Verge.

Self-driving has been “almost here” for over a decade. But somewhere between DARPA challenges and a handful of driverless trucks hauling freight between Dallas and Houston, Aurora co-founder and CEO Chris Urmson’s story changed. The self-driving truck company started commercial driverless operations last April and is now scaling from a handful of trucks to hundreds this year.  On this episode of TechCrunch’s Equity podcast, we’re bringing you a […]

A new Google paper argues that image generation pretraining is to computer vision what GPT-style pretraining is to NLP — and the benchmark numbers back that up. The post Google DeepMind Introduces Vision Banana: An Instruction-Tuned Image Generator That Beats SAM 3 on Segmentation and Depth Anything V3 on Metric Depth Estimation appeared first on MarkTechPost.

Machine learning models can be confident even when they shouldn't be. This article introduces Deep Evidential Regression (DER), a method that lets neural networks rapidly express what they don't know. The post Introduction to Deep Evidential Regression for Uncertainty Quantification appeared first on Towards Data Science.

A software developer claims to have reverse-engineered Google DeepMind's SynthID system, showing how AI watermarks can be stripped from generated images or manually inserted into other works. A claim that, according to Google, isn't true. The developer, going by the username Aloshdenny, has open-sourced their work on GitHub and documented his process, claiming all it required was 200 Gemini-generated images, signal processing, and "way too much free time." A little weed also seemed to help. "No neural networks. No proprietary access," Aloshdenny said on Medium. "Turns out if you're unemployed and average enough 'pure black' AI-generated im … Read the full story at The Verge.

AI is consuming staggering amounts of energy—already over 10% of U.S. electricity—and the demand is only accelerating. Now, researchers have unveiled a radically more efficient approach that could slash AI energy use by up to 100× while actually improving accuracy. By combining neural networks with human-like symbolic reasoning, their system helps robots think more logically instead of relying on brute-force trial and error.