How do we make AI-powered trip ideas in Search both creative & practical? Our new research details the method: using LLMs to understand your travel goals, paired with optimization algorithms to handle real-world logistics like opening hours. Learn more →https://goo.gle/4jH4BmF
About us
From conducting fundamental research to influencing product development, our research teams have the opportunity to impact technology used by billions of people every day. We aspire to make discoveries that impact everyone, and sharing our research and tools to fuel progress in the field is fundamental to our approach.
- Website
-
https://research.google/
External link for Google Research
- Industry
- Technology, Information and Internet
- Company size
- 1,001-5,000 employees
Updates
-
Today on the Google Research Blog we're announcing a novel generative AI model, called dynamical-generative downscaling, that efficiently bridges the gap in resolution between the outputs of standard Earth system models and what's needed by climate & environmental risk assessment tasks. Learn more at https://goo.gle/3FIsw7e
-
-
Google Research is now also on X! →https://goo.gle/3ZM838i Follow along to stay in the loop with the real-world impact of our work — across health, climate, weather, quantum, algorithms, privacy, hardware, AI, and more.
-
-
Today on the blog we propose Action-Based Contrastive Self-Training, a data-efficient #ReinforcementLearning tuning approach for improving multi-turn conversation modeling in mixed-initiative LLM interaction. Read all about it →https://goo.gle/3Sxas2T
-
-
Understanding genuine progress in quantum computing is not always straightforward. Dive into this with Julian Kelly and Sinead Bovell in the #GoogleIO session "Quantum computing: Reading signals from the noise," to address how to assess breakthroughs, applications, and more → https://goo.gle/3ZmogB9
-
User-level differential privacy is a method that is often employed in federated learning applications to train models across distributed devices while maintaining strong privacy guarantees. Here we investigate and describe improved algorithms for fine-tuning large language models with user-level differential privacy in the more flexible datacenter training environment. Learn more at https://goo.gle/3Sj6GKg
-
-
Each year at Google I/O, we share some of Google’s most advanced technologies. We show how they can be helpful and provide new experiences, and how developers and other communities can use them to create. In today’s blog post, Yossi Matias, Vice President and Head of Google Research, presents many of the Google Research contributions to the innovations presented at Google I/O 2025. Many of these new technologies emerged from years of work within Google Research, often in collaboration with other teams and building on multiple breakthroughs in AI and other areas of computer science. Read more at https://goo.gle/3SOoWeB
-
-
Today at Google I/O, we announced MedGemma, our most capable open model for multimodal medical text and image comprehension. MedGemma is based on Gemma 3, can be fine-tuned for specific needs and is designed to be a starting point for developers building health applications, such as analyzing radiology images or summarizing clinical data. The models are now available as part of our Health AI Developer Foundations (https://goo.gle/3YTIp16). This is the latest of our long-standing efforts at Google Research to advance AI to make healthcare more accurate and accessible for billions of people around the world. Read more here → https://goo.gle/4mqXfWP
-
-
Everyone deserves to be understood. This Global Accessibility Awareness Day, we’re spotlighting Project Euphonia, a Google Research initiative committed to improving speech recognition for people with non-standard speech. Through open-source tools, developers can create personalized audio tools and fine-tune open-weight models for diverse speech patterns, paving the way for more inclusive technology. #GAAD Learn more → https://goo.gle/4dq8Cui
-
-
Today on the blog we introduce a notion of sufficient context to examine retrieval augmented generation (RAG) systems, developing a method to classify instances, analyzing failures of RAG systems & proposing a way to reduce hallucinations. Read more →https://goo.gle/43gp3Vk
-