Google DeepMind

Privacy-enhancing technologies (PETs) can provide organisations with enhanced security and confidentiality of data and code, but several challenges and barriers stand in the way of their adoption. One such barrier is a lack of trust: sceptical, privacy-aware individuals find it difficult to trust that PETs will effectively keep their sensitive data and code protected and safe. In this research, we set out to answer a set of research questions regarding trust in Google’s Project Oak. We aimed to analyse different transparency measures that PET providers can use to increase trust, and the extent to which they do so in practice. Find out more. #Privacy #Trust https://lnkd.in/dFuyxvWE

You can now use our state-of-the-art video generation model Veo 2 in YouTube Shorts. 🎥 With the Dream Screen feature, creators can: ✨ Generate standalone clips which fit seamlessly into your storytelling with a quick text prompt ✨ Make unique backgrounds for videos 💡 With a better understanding of real-world physics and movement, the model can help render specific styles or cinematic effects. 🔍 Any content generated using this tool is labelled with SynthID, which accurately identifies AI-generated images, audio, text and video. 🌐 These features are launching to everyone in the US, Canada, Australia, and New Zealand - with plans to expand in the future. Find out more → https://goo.gle/4jTa7nt

We're empowering future generations to use AI responsibly. Our new Experience AI resources - co-created with Raspberry Pi Foundation - are designed to help young people learn about AI and data privacy, navigate misinformation, and harness generative tools safely. 📚 Teachers, extracurricular clubs, and parents can use them with confidence - even without deep AI knowledge. → https://lnkd.in/eXGv2nAQ

Come and join us for our workshop, Natural Interactions with Foundation Models, at AMLD - Applied Machine Learning Days 2025 in Lausanne 🇨🇭! Together with my colleagues Diego Antognini, PhD, Leandro Kieliger, Philippe Schlattner, Vincent Coriou, and Anastasia Fadeeva from Google DeepMind, we are excited to show you how to use stroke interactions to interact with large language models. We hope to see many of you at AMLD - Applied Machine Learning Days - Applied Machine Learning Days, especially on Friday, February 14th, from 14:00 to 17:30! If you plan to attend, please fill out this form in advance: https://lnkd.in/eX2B3vS6. Learn more about AMLD: https://lnkd.in/egqVyeQ9

Today, we’re making Gemini 2.0 available to everyone - and expanding the model family. ✨ Say hello to: 🔵  An updated 2.0 Flash, which is optimized for high-volume, high-frequency tasks at scale - enabling developers to start building production applications. You can now use it via the Gemini API in Google AI Studio, Google Cloud’s #VertexAI platform and the Gemini app. 🔵  2.0 Pro Experimental: our best model yet for coding tasks and complex prompts. With a 2 million token context window, it’s able to analyze and understand large amounts of information. 🔵 2.0 Flash-Lite, our most cost-efficient AI model yet - now available across Google products. It has better quality than 1.5 Flash, at similar cost and speed, and comes with a 1 million token context window, multimodal input and text output. 🔵 And 2.0 Flash Thinking Experimental now available in the Gemini app. Find out more → https://goo.gle/3CMXg5p

As we make progress towards AGI, developing AI needs to be both innovative and safe. ⚖️ To help ensure this, we’ve made updates to our Frontier Safety Framework - our set of protocols to help us stay ahead of possible severe risks. These include: 🟢 Outlining an industry leading approach to deceptive alignment risk 🟢 Implementing a more consistent procedure for how we apply deployment mitigations 🟢 Recommending specific security levels for different critical capabilities Find out more → https://goo.gle/4hhKDOS

We’re putting our most advanced AI for weather forecasting into more people’s hands. ☁️🌐 Scientists can now access Google’s WeatherNext models as well as real-time and historical data to accelerate research, improve how we respond to disasters and much more. → https://goo.gle/40BBEkp

Our CEO Demis Hassabis joined the Financial Times's Roula Khalaf to talk about #AlphaFold, how AI can revolutionize scientific discovery, and why we could be 5-10 years away from AGI. 🌐 Watch in full → https://goo.gle/4azvtlC

Thanks Alex Kantrowitz for the great chat. Really enjoyed covering some of my favourite topics... scaling, AGI, and some of our latest breakthroughs.