Google Research shared its newest announcements. These cover understanding our planet, advancements in genomics, and progress in quantum computing. Our teams work with others in Google. They create new research and speed up solutions for products, businesses, and science. This cycle of research and real-world impact is powerful. We call it the magic cycle of research. It is getting faster. This is thanks to better AI models, new tools for discovery, and open platforms.
We see this progress in many areas. Our research is driving innovation faster than ever.
Watch the film
Our latest research breakthroughs
We recently highlighted three major breakthroughs. These are Google Earth AI, DeepSomatic, and Quantum Echoes. They show how our research is making a difference.
Google Earth AI: Understanding our planet like never before
Earth AI uses AI models for geography. It helps solve big world problems. It gives people a new way to understand what is happening on Earth. We have built many AI models for geography. These include models for floods, wildfires, cyclones, air quality, pollen, and weather. We also have models for agriculture, population changes, and city travel.
These models help millions of people. We keep improving them. Our river flood models now help over 2 billion people in 150 countries. Earth AI is a big program at Google. It uses AI to understand geographic data. It can find links between different data and models. This helps people ask complex questions. They can get answers in simple words. This makes Earth AI easy to use, even for people who are not experts.
We found that our AI system gives better answers than older models. We shared these results in our research blog and technical report. Google Earth will soon use our Earth AI models. This will let people search for things in satellite pictures. Our models are also available to testers on Google Cloud. Many partners are finding important uses for this technology.
DeepSomatic & Cell2Sentence: Fighting cancer with precision medicine
DeepSomatic is a new AI tool. It helps scientists and doctors understand genetic changes in cancer cells. This work builds on 10 years of genomics research at Google. We have created models like DeepConsensus and DeepVariant. These help us understand the genome better. They helped us map human and non-human genomes. This helps us understand diseases better.
Some cancers have complex genetic markers. These can be treated with special medicines. We wanted to study the genes of cancer cells more closely. DeepSomatic is our new AI tool. It is open-source. It helps scientists and doctors understand cancer cell genes. The model turns genetic data into images. Then, it uses a special network to tell the difference between normal genes and cancer genes.
Finding cancer genes could lead to new treatments. It could help doctors choose the best treatments for patients. Our partners are using DeepSomatic to find new ways to treat cancer. DeepSomatic follows other AI breakthroughs that fight cancer. We also released a large model for single-cell analysis called C2S-Scale. This work has helped find new ideas for cancer therapies.
Quantum Echoes: A step toward real-world quantum use
We are investing in quantum computing. This will help speed up scientific discovery. We have made progress with our Willow chip. Our work is supported by top scientists. Two of them won the Nobel Prize in Physics for their research.
We announced a new verifiable quantum advantage. This is called the “Quantum Echoes” algorithm. It runs on our Willow chip. It is much faster than the best classical computer. This algorithm helps explain how atoms interact in molecules. This is the first algorithm to show a real quantum advantage. It shows that quantum computers can do things that classical computers cannot.
Quantum computing can help design drugs. It can also help with fusion energy. We think we will see real-world uses for quantum computing in five years.
Watch the film
We had a chat about Quantum AI.
From accelerating discovery to better algorithms
We shared work from different areas. Our teams are driving new research. They are also creating real-world solutions. The chances for innovation are growing.
Health & Science
Our AI co-scientist helps researchers. It suggests new ideas and research plans. It speeds up discoveries in science and medicine. Our AI-powered software helps scientists write code. This code is used to test and improve research ideas. This means scientists can test many solutions quickly.
AMIE is a medical AI agent. It can talk with patients. It can understand medical problems like a doctor. We are testing AMIE with doctors. This helps us see how it works with real patients. MedGemma is a powerful AI model for health. It is open to use. Many people are using it for health AI development.
Factuality & Efficiency
We are improving how LLMs handle facts. We study how they show uncertainty. We also study how they store factual knowledge. We are working on making AI understand videos better. This helps improve our image and video models.
Making LLMs more efficient is important. We are using new methods like speculative decoding. These methods make AI faster without losing quality. We are also working on new ways to save energy.
Algorithmic innovation
Our research helps create new models for ads. It helps connect advertisers with customers. We are improving Google Maps routing. We also made voice search better in India. Our privacy research helps protect user data. We are making progress on AI for predicting future events. This is used in BigQuery and other tools.
We are exploring new ways to improve learning. We created LearnLM and Learn Your Way. These tools help make learning better. We are also exploring new AI uses. This includes using AI for game engines. This can create immersive virtual worlds.
Watch the film
We discussed Google's research efforts.