AI News Today

Stanford geneticist uses Co-Scientist to help find new treatments for chronic liver disease and liver fibrosis.

Learn how our WeatherNext AI model help forecasters give communities unprecedented time to prepare ahead of the historic Hurricane Melissa.

Our recent paper, “LLMs Corrupt Your Documents When You Delegate”, has generated discussion about the reliability of AI systems in delegated workflows. We appreciate the interest in this work and want to clarify several important points about what the paper does—and does not—claim. The research aims to develop robust evaluation methods for long-horizon delegated and […] The post Further Notes on Our Recent Research on AI Delegation and Long-Horizon Reliability appeared first on Microsoft R

At least we probably won’t have another pandemic.

This is what a lull looks like at this point.

The real recent story of AI has been the background work being done on Cybersecurity, as we process the Mythos Moment along with GPT-5.5, and figure out both how to patch the internet and what our new regulatory regime is going to look like.

mimalloc is an open-source, modern, scalable memory allocator that is a drop-in replacement for malloc and free. It is relatively small (~12K lines), with clear internal data structures, and is easy to build and integrate into other projects. It provides bounded worst-case allocation times (up to OS primitives), bounded space overhead, low internal fragmentation, and minimal contention by relying almost exclusively on atomic operations. The post mimalloc: A new, high-performance, scalable memory

Introducing GridSFM, a small foundation model that can predict AC optimal power flow in milliseconds, boosting efficiency and unlocking cost savings. Learn how GridSFM gives grid operators direct visibility into congestion, stability, and system health. The post GridSFM: A new, small foundation model for the electric grid appeared first on Microsoft Research.

We did reading yesterday.

On therapy sessions, Slack archives, and citations that didn't exist. Source

MatterSim is expanding what AI can do for materials science—from faster large-scale simulations to MatterSim-MT, a new multi-task model for simulating properties beyond potential energy surfaces alone. The post Advancing AI for materials with MatterSim: experimental synthesis, faster simulation, and multi-task models appeared first on Microsoft Research.

Reading is the most fundamental thing in education.

Using SocialReasoning Bench, we observed a stable pattern across models—agents execute competently, but fail to consistently improve the user’s position, even with explicit instructions to optimize for user interest. The post SocialReasoning-Bench: Measuring whether AI agents act in users’ best interests appeared first on Microsoft Research.

<img src="https://storage.googleapis.com/gweb-uniblog-publish-prod/images/UK_1920x1080.max-600x600.format-webp.webp">This week, the new, AI-powered Google Finance is launching across Europe, with full local language support. This reimagined experience offers a suite of powerful capabil…

Image captioning is one of the most fundamental tasks in computer vision. Owing to its open-ended nature, it has received significant attention in the era of multimodal large language models (MLLMs). In pursuit of ever more detailed and accurate captions, recent work has increasingly turned to reinforcement learning (RL). However, existing captioning-RL methods and evaluation metrics often emphasize a narrow notion of caption quality, inducing trade-offs across core dimensions of captioning. For

Tailwind CSS v4.3 is here with first-party scrollbar styling, even more logical property utilities, new zoom and tab-size utilities, better @variant support, and all the v4.2 stuff we shipped while forgetting blogs existed.

When I started this series, everyone was going crazy for coding agents.

Current critic-less RLHF methods aggregate multi-objective rewards via an arithmetic mean, leaving them vulnerable to constraint neglect: high-magnitude success in one objective can numerically offset critical failures in others (e.g., safety or formatting), masking low-performing “bottleneck” rewards vital for reliable multi-objective alignment. We propose Reward-Variance Policy Optimization (RVPO), a risk-sensitive framework that penalizes inter-reward variance during advantage aggregation, sh

At Apple, we believe privacy is a fundamental human right. As AI capabilities increase and become more integrated into people’s daily lives, advancing research in privacy-preserving techniques is increasingly important to ensure privacy is protected while users enjoy innovative AI experiences. Apple’s fundamental research has consistently pushed the state-of-the-art in this domain, and earlier this year, we hosted the Workshop on Privacy-Preserving Machine Learning & AI. This two-day event

We propose HeadsUp, a scalable feed-forward method for reconstructing high-quality 3D Gaussian heads from large-scale multi-camera setups. Our method employs an efficient encoder-decoder architecture that compresses input views into a compact latent representation. This latent representation is then decoded into a set of UV-parameterized 3D Gaussians anchored to a neutral head template. This UV representation decouples the number of 3D Gaussians from the number and resolution of input images, en

We introduce a framework for learning latent representations of 4D objects which are descriptive, faithfully capturing object geometry and appearance; compressive, aiding in downstream efficiency; and accessible, requiring minimal input, i.e., an unstructured dynamic point cloud, to construct. Specifically, Velox trains an encoder to compress spatiotemporal color point clouds into a set of dynamic shape tokens. These tokens are supervised using two complementary decoders: a 4D surface decoder, w

One of the major differentiators unlocked by learned codecs relative to their hard-coded traditional counterparts is their ability to be optimized directly to appeal to the human visual system. Despite this potential, a perceptual yet practical image codec is yet to be proposed. In this work, we aim to close this gap. We conduct a comprehensive study of the key modeling choices that govern the design of a practical learned image codec, jointly optimized for perceptual quality and runtime — inclu

Data Mining & Modeling

Data Mining & Modeling

<span class="byline-author">Posted by Thomas Brunner, Yu-Han Liu, Moni Pande</span><div><br /></div><div><span id="docs-internal-guid-49e83394-7fff-3df2-a974-c7291d12beb8"><p dir="ltr" style="line-height: 1.38; margin-bottom: 10pt; margin-top: 0pt;"><span face="Roboto, sans-serif" style="font-size: 12pt; font-variant: normal; vertical-align: baseline; white-space: pre-wrap;">At Google, our Threat Intelligence teams are dedicated to staying ah

Generative AI

Generative AI

Generative AI

General Science

Education Innovation

<span class="byline-author">Posted by Jiacheng Lu, Software Engineer, Google Pixel Team</span> <p>Google is continuously advancing the security of Pixel devices. We have been focusing on hardening the cellular baseband modem against exploitation. <a href="https://security.googleblog.com/2023/12/hardening-cellular-basebands-in-android.html">Recognizing the risks</a> associated within the complex modem firmware, Pixel 9 shipped with <a href="https://security.googl

<span class="byline-author">Posted by Ben Ackerman, Chrome team, Daniel Rubery, Chrome team and Guillaume Ehinger, Google Account Security team</span> <p> Following our April 2024 <a href="https://blog.chromium.org/2024/04/fighting-cookie-theft-using-device.html">announcement</a>, Device Bound Session Credentials (DBSC) is now entering public availability for Windows users on Chrome 146, and expanding to macOS in an upcoming Chrome release. This project represents

Generative AI