Google DeepMind

Meet Gemini Robotics: our latest AI models designed for a new generation of helpful robots. 🤖 ⚡ Gemini Robotics is our most advanced vision-language-action model. It can take in text and images to output actions, interact with humans and adapt to changing environments. It can also perform complex tasks which require a lot of dexterity, such as: 📄 Folding origami 🥗 Packing a lunch box 🎲 Playing tic-tac-toe, and much more. 🦾 And it works across existing bi-arm robots like ALOHA 2 and Franka, as well as more complex applications like Apptronik's humanoid. ⚡ Gemini Robotics-ER brings advanced world understanding to robotics - enabling them to carry out tasks they haven't been trained on, using perception, spatial understanding, planning and more. We look forward to continuing developing these models, working with a broad range of trusted testers, to bring Gemini into the physical world. Find out more → goo.gle/Gemini-2-robotics

We've launched a new framework to evaluate emerging offensive cyber capabilities of AI. This will help experts proactively identify and prioritize necessary defenses - before risks can be exploited. Find out more ↓ https://goo.gle/3R03CC4

When his son was diagnosed with a rare disease, Googler Thomas Wagner used Gemini to bridge the gap between complex research papers and the scientists leading the work. 🔬 🧠 He was able to better understand the science behind Alexander disease (AxD) by breaking down dense scientific works, genetic explanations, and medical articles into understandable terms. ➡️ This knowledge then helped him to see connections between different research areas and consider how they might apply to AxD. ❔He was able to come up with better questions for those working on treatments — and his efforts mean 11 research groups who hadn’t worked on AxD before are now beginning to do so. His story is a powerful reminder of why we work on AI. ↓

AGI could revolutionize many fields - from healthcare to education - but it's crucial that it’s developed responsibly. Today, we’re sharing how we’re thinking about safety and security on the path to AGI. → https://goo.gle/4l9fAax

Generative AI are systems that learn from existing data and use what they know to generate new content. From generating new hypotheses to accelerating discovery, it will undoubtedly shape research – and that includes the field of mental health. We want to find out if and how AI can be used in mental health. The first step is to understand how AI, clinicians and patients can collaborate in the most efficient, effective and safe ways. We’re doing so through a two-part funding activity that aims to improve measurement or treatment of anxiety, depression and psychosis. Successful applicants will first take part in a four-month accelerator delivered by MEXA. They can then apply for our funding call to receive up to £3 million for their research project – and access to resources and support from experts at Google Health and Google DeepMind. Find out more: https://lnkd.in/e7iBePX8

AI development should be bold 💪— and responsible. Google DeepMind Senior Staff Research Scientist Iason Gabriel shares how forward-thinking, ethical practices are shaping tomorrow’s AI assistants. ↓

Developing new drugs can be notoriously slow, expensive and risky. 🔬 Enter TxGemma: a more general, open and scalable set of AI models to help improve this process for researchers and developers. Find out more → https://goo.gle/3Y864u7

This is the power of Veo 2. 🎥 Our state-of-the-art video generation model with 4K outputs, expansive styles, and more detail than ever. In the hands of creators, they’ve been able to produce videos featuring animated characters to cinematic outputs, and more. ↓ https://goo.gle/4l59fwE

Think you know Gemini? 🤔 Think again. Start building with Gemini 2.5: our most intelligent model 💡 The first release is Pro Experimental, which is state-of-the-art across many benchmarks - meaning it can handle complex problems and give more accurate responses. Try it now in Google AI Studio → https://goo.gle/4c2HKjf

𝗔𝗴𝗲𝗻𝘁 𝗥𝗲𝗱-𝗧𝗲𝗮𝗺𝗶𝗻𝗴 𝗖𝗵𝗮𝗹𝗹𝗲𝗻𝗴𝗲 𝗨𝗽𝗱𝗮𝘁𝗲: 𝗡𝗲𝘄 𝗖𝗼-𝗦𝗽𝗼𝗻𝘀𝗼𝗿 & 𝗘𝘅𝗽𝗮𝗻𝗱𝗲𝗱 𝗣𝗿𝗶𝘇𝗲 𝗣𝗼𝗼𝗹 We’re excited to share a major update on our ongoing UK AISI Agent Red-Teaming Challenge. As we reach the halfway point, we’re proud to announce that Google DeepMind has joined as a co-sponsor—supporting our commitment to advancing AI security. This collaboration, alongside industry leaders like OpenAI and UK AISI, has enabled us to raise the total prize pool to $𝟏𝟕𝟎𝗞. 𝗞𝗲𝘆 𝗛𝗶𝗴𝗵𝗹𝗶𝗴𝗵𝘁𝘀: 🔹𝗡𝗲𝘄 𝗖𝗼-𝗦𝗽𝗼𝗻𝘀𝗼𝗿: Google DeepMind now supports the challenge, strengthening our commitment to cutting-edge AI research. 🔹𝗘𝘅𝗽𝗮𝗻𝗱𝗲𝗱 𝗣𝗿𝗶𝘇𝗲 𝗣𝗼𝗼𝗹: With the prize pool now at $150K, we’re incentivizing breakthrough exploits and innovative solutions in AI red-teaming. 🔹𝗢𝗻𝗴𝗼𝗶𝗻𝗴 𝗜𝗻𝗻𝗼𝘃𝗮𝘁𝗶𝗼𝗻: New challenge behaviors continue to drop weekly, ensuring the competition remains dynamic and at the forefront of AI vulnerability research. 🔹𝗜𝗺𝗽𝗮𝗿𝘁𝗶𝗮𝗹 𝗝𝘂𝗱𝗴𝗶𝗻𝗴: UK AISI remains dedicated to providing fair and unbiased evaluations, ensuring a level playing field for all participants. 🔹𝗗𝘂𝗿𝗮𝘁𝗶𝗼𝗻: The challenge runs through April 6th—offering ample time for top-tier red teamers and emerging talent to engage and excel. This update reflects our ongoing mission to drive industry-leading research and collaboration in AI security. We invite both top-tier red teamers and our esteemed partners to continue engaging with the challenge, shaping a safer and more secure AI ecosystem for all. 𝗟𝗲𝗮𝗿𝗻 𝗠𝗼𝗿𝗲 & 𝗝𝗼𝗶𝗻 𝘁𝗵𝗲 𝗖𝗼𝗻𝘃𝗲𝗿𝘀𝗮𝘁𝗶𝗼𝗻:  Agent Red-Teaming Challenge Details & Sign-Up: https://lnkd.in/emdY8a_6  Join Our Community on Discord: discord.gg/grayswanai Thank you for your continued support as we push the boundaries of AI security together.