Google DeepMind

Last week, Hurricane Isaias made landfall in the Florida Panhandle. Our thoughts are with everyone affected and with the recovery efforts still underway. Earlier and more accurate forecasting can help protect communities by giving them more time to prepare, and we are grateful that Google’s WeatherNext model was able to assist the National Hurricane Center's expert forecasters. Based on initial evaluations, here's how our model supported the response to Isaias: - Tracking the Hurricane’s Path: WeatherNext was the most accurate model on the storm’s track. While traditional physics models predicted landfall closer to New Orleans, WeatherNext correctly predicted a path further east. This allowed NHC forecasters to highlight the potential for wind and storm surge impacts in the Florida Panhandle from their very first forecast. - More Time to Prepare: Up to three days before landfall, our model consistently predicted its eventual location in the Florida Panhandle to within 60 km or less, supporting the NHC as they honed in on a track toward the Florida Panhandle and issued hurricane watches and warnings to give officials in Florida, Alabama, and Mississippi more time to prepare. - Early Indications about Intensity: Early on, WeatherNext accurately predicted the intensification to a major hurricane. Alongside the regional hurricane models HWRF and HMON, WeatherNext’s intensity forecasts helped NHC forecasters to predict that the Isaias would approach the coast as a category 2 hurricane from the very first advisory. Giving communities more time and more accurate warnings is critical for protecting lives and livelihoods. While model performance varies from storm to storm, Isaias reflects our broader findings. In our recent Nature paper co-authored with NHC forecasters, evaluating 200+ global tropical cyclones, we found our model provides ~1 day of additional lead time for both track and intensity. It’s been an honor for the Google DeepMind and Google Research teams to work alongside NHC forecasters. By combining AI, physics models, satellite data, and the crucial expertise of meteorologists, we're excited to keep pushing the boundaries of forecasting to help protect communities.

SynthID has been named one of TIME’s Best Inventions of 2026. With over 180 billion images and videos already watermarked, it’s making it easier for people to distinguish AI-generated content from other media → https://goo.gle/47V1BPV

Here’s how researchers are using Gemini 4 Argon to accelerate scientific discovery. ⚛️ Quantum computing: The model optimized our real-time AlphaQubit error-correction decoder to run with nearly twice the throughput. It also designed a routine to compute arcsin, a core building block for quantum simulations, that cuts computing costs by more than 100x ☀️ Nuclear fusion: Advancing limitless clean fusion energy requires optimizing the complex 3D magnetic fields inside fusion machines called stellarators. Alongside University of Maryland professor Matt Landreman, our team found a broader class of solutions to long-standing math problem Grad’s conjecture - once thought impossible to solve - giving reliable benchmarks for stellarator simulation codes. 🧬 Enzyme design: Argon reduced months of screening to a matter of hours - curating data, tuning models, and applying biochemistry filters to generate five novel enzymes - all confirmed active in wet-lab testing. To further the pace of research, we’re offering 10,000 scientists free access via Google Antigravity for 6 months. Find out more → https://goo.gle/4yFTUce

Over half of reported crimes in Singapore use sophisticated social engineering. With GovTech Singapore, we’re building on-device fraud detection using our open model Gemma. By fine-tuning it on local scam tactics, it identifies threats in real time without compromising personal privacy → https://goo.gle/4yu2WIE

Gemini 4 Argon taught a robot to play cornhole just by watching a video. It mapped the physics, coded its own simulation, and taught Apptronik’s Apollo humanoid to sink 7 throws in a row. By turning video into a training ground for physical AI, Argon compresses weeks of manual modeling into just a few hours. Find out more → https://goo.gle/4etktJu

Gemini 4 Argon is now globally available. Built to solve your hardest problems from start to finish, the model brings frontier capabilities across coding, enterprise workflows, and cyber defense. Here’s how you can try it: 🔵 Everyday users: In the Gemini app and AI Mode in Search for Google AI Pro and Ultra subscribers. 🔵 Developers: Gemini API paid tier, plus Antigravity and Google AI Studio for Google AI Ultra subscribers. 🔵 Scientists: Apply for the top usage tier in Antigravity free of charge. 🔵 Enterprises: Gemini for business, the API, and in Antigravity. Find out more → https://goo.gle/4yFTUce

In the latest essay as part of our DeepMind Institute work, my colleague Alex Imas and I argue that AI has the potential to bend the curve of scientific discovery in a way that drives not only scientific progress, but also economic growth and societal progress, but only if the global scientific enterprise adapts systems and institutions and clears bottlenecks and enables scientists everywhere.  This comes at a particularly helpful time, when the type of scientific progress that drives economic growth and societal progress is facing headwinds; sustaining historical rates of innovation now requires exponentially larger research teams and capital expenditure.  AI enables and assists the work of scientists in ways that can substantially accelerate scientific productivity (i.e. more science and faster). We see a strong case that AI is not only a general purpose technology, but also an invention of a method of invention (IMI) like a microscope or telescope. What’s exciting about IMIs is that they permanently shift the production function of discovery itself.  However to fully realize AI’s potential as an IMI and driver of economic growth, scientific progress cannot be simply increasing the volume of research output (and without validation). Instead it must be leveraged by scientists to expand the boundaries of what is scientifically possible and extend the knowledge frontier. We offer ideas on how this might be accomplished, drawing in part on recent research on how scientists are using AI in their work. As with all DeepMind Institute essays, we hope sharing our thinking serves as a springboard for further discussion and solutions. You can read our full essay on the DeepMind Institute’s site: https://lnkd.in/ejUryJ8r

SynthID Detector is now available to everyone. 🌐 You can check whether digital content was generated using Google AI, or with tools from our industry partners – including OpenAI, NVIDIA, and Kakao Corp and coming soon, Apple. SynthID Detector can quickly identify our SynthID watermarks in video, audio and images – even after edits, like adding filters. ✂️ You can also use existing AI identification features found in Google Chrome, Google Search, and the Gemini app. Try it out → synthid.com

Meet EmbeddingGemma 2, our first natively multimodal open model for on-device embeddings. It expands beyond text to unify code, images, audio, and video in a shared space. Developers can use it to add multimodal search to their apps – like finding moments in a video using a voice memo – or pair it with Gemma 4 to answer questions across local files, completely offline. At 740M parameters, EmbeddingGemma 2 is competitive across benchmarks – even outperforming some specialist models more than twice its size. Check out the weights on Hugging Face and Kaggle, released under an Apache 2.0 license, or read the full details → https://goo.gle/4hPuHoD

AI could transform how we tackle antimicrobial resistance (AMR) – from discovering new treatments to tracking drug-resistant infections. But how do we know when an AI system is accurate, reliable and ready to use? Supported by Google DeepMind, we are launching a new three-year programme to develop the methods and standards needed to evaluate AI for AMR. Working with the global AMR community, we’ll build shared frameworks and benchmarks to help researchers compare approaches, understand their limitations and develop AI that can be used with greater confidence in research and healthcare. Read more: https://lnkd.in/eTCrYUGJ #AntimicrobialResistance #AMR #AntibioticResistance #ArtificialIntelligence #WorldAIWeek Imperial College London Imperial College Healthcare NHS Trust Imperial College Computing