We live in a golden age of research.
Never before have we had the opportunity to advance computer science so rapidly and apply it so quickly to global challenges in ways that help people in their daily lives. Since the beginning of my career, I have been excited about the “magic cycle” of research. Real-world issues motivate new basic and applied research, which has a positive impact on the real world. Today, with the right infrastructure, people, and approaches, we are not only making rapid advances in everything from AI to algorithms to computing infrastructure, but also leveraging those technologies to improve people’s daily lives. and impact society faster than ever before. Previously, sometimes within a few months.
I recently took over the role of head of Google Research and have seen this firsthand. That’s why we wanted to share our perspective on the amazing advances we’re seeing and how important research is in driving useful innovation.
Our approach: shock-based curiosity
In fact, Google itself started with research. In our 1998 book Anatomy of a Large-Scale Hypertext Web Search Engine, we looked at how PageRank can provide a fundamentally better way to find information on the Web. But it wasn’t just a research paper, it was an application of that research. That gave rise to Google. Over the past 26 years, this approach to conducting research rapidly has transformed not only our company, but also the way people interact with the world’s information. In fact, many of the rapid advances in AI we see around us today stem from the invention of Transformer by Google Research.
In all of our research, we ask ourselves how we can create incremental change rather than just incremental change. What can we do that is impossible today? And what is the biggest impact we can have? How can this bring about real change in the world?
Google Research today includes fundamental and applied research in fundamental machine learning and algorithms, computing systems and quantum AI, and impact on science, AI, and society. And across all of these areas, we’re implementing work on machine learning efficiency, facts and evidence for AI systems, and new data technologies.
Breakthroughs for the benefit of people and planet
We motivate our research by pursuing the biggest questions that matter to advance science and make a difference for people and communities around the world. What is the most effective way to mitigate climate change? How can we make billions of people healthier? How can we enable new experiences? And all of this? Can we push the limits of computing and AI systems to achieve this? Each of these poses an exciting research question, and in many cases we are already turning our research into solutions.
For example, to address climate change, a trial with American Airlines used AI to reduce contrails by 54%, ensuring airlines avoid contrails and thereby reduce climate impact. has been demonstrated. To address the growing wildfire crisis, we are partnering with leading wildfire authorities to develop a future global AI-powered technology that can detect and track classroom-sized wildfires in under 20 minutes. We developed FireSat, a constellation of satellites. And let’s think about flood forecasting. When the Google Research team started the project in 2018, experts I met said it was probably impossible to predict river flooding. But by developing AI that can build global hydrology models, we have not only proven that it is possible, but also used it to provide accurate early warnings to local communities and save lives. I was able to apply it for this purpose.
Meanwhile, to support health and well-being, we develop AI that helps screen for breast and colorectal cancer, prevent mass blindness, spot potential skin diseases, and detect diseases based on cough sounds. did. Although we are in the early stages of breakthroughs in AI and genomics research, we are still making advances in large-scale language models for healthcare and are already poised to improve healthcare for hundreds of millions of people. Masu.
But perhaps one of the biggest efforts has to do with advances in computing itself and how it fundamentally changes the scope of problem-solving. We developed a state-of-the-art attention model and used graph mining to improve search quality. We are also working on approaches to ground language models at scale, such as training models to rely on source documents for summarization and factual consistency assessment. This research led to improvements such as the double-check feature in the Gemini app. We’ve made AI more efficient with research in speculative decoding, efficient inference techniques, and cascading, and helped improve privacy with federated learning and differential privacy. And our quantum computing team has produced new results that, by reducing the error rate of quantum processors, reach computational levels unmatched by even the world’s most powerful classical computers, putting them on track to crack. We just published it in Nature. Opening up a whole new computing power to the world.
These are just a few examples of the work done across Google Research.
Why partnerships are important for turning research into results
Of course, as we move our research toward impact, we are keenly aware that technology does not automatically become beneficial. You can’t just “invent it and forget it” by releasing powerful technology based on the simple assumption that it will be useful. It requires careful management, partnership, and a fundamentally human-centered perspective on how to design and guide emerging technologies. This is one of the reasons why we work in collaboration with academia and many other researchers, supporting a truly global research ecosystem with the aim of bringing new technologies to fruitful outcomes. It creates tools and best practices to help you. We are actively working to advance socio-technical research to strengthen AI principles such as fair datasets, interpretability, and algorithmic fairness, making AI models more efficient and fact-based. There is still significant work ahead to make it robust and secure.
Our greatest impact is when we collaborate with our research partners. Since 2005, Google has collaborated with more than 1,000 research institutions and invested more than $400 million in academic research around the world, much of it led by Google Research. We find experts across disciplines, roll up our sleeves, get to work, and work together to advance science. Our connectomics research partnered with Harvard University to use AI to create the most detailed mapping of neurons in the human brain to date, revealing newly discovered structures. All of these are aimed at helping scientists understand fundamental processes such as thinking, learning, and memory. Google Research also collaborated for the first time with Howard University and other HBCUs to build a high-quality African American English (AAE) speech dataset that Google and others use to improve their voice products. This relates to our overall commitment to reducing barriers and better serving our communities by making technology available in more languages.
Together with our partners and through Google’s unique products, we harness research advances to benefit billions of people. For example, as populations grow and move in the Global South, millions of people’s buildings are not visible on any map and they risk missing out on necessities such as electricity, healthcare, and postal delivery. So Google Research in Africa used AI to significantly improve the Open Buildings dataset. Converting blurry, low-resolution satellite imagery into useful high-resolution building outlines has enabled partners such as the World Bank, World Resources Institute, UN-Habitat, and WorldPop to collaborate. And Sunbird AI can use this to enable everyone to participate in global development. This, along with SKAI efforts, helped strengthen damage assessments in partnership with the United Nations.
In another area, our operations research team recently showed how cargo shippers can double their profits, deliver 13% more containers, and use 15% fewer vessels. This is beneficial not only for companies but also for supply chains around the world.
Finally, of course, we partner extensively with our product teams to drive innovation across Google. And our responsibility includes looking to the horizon, exploring the art of the possible, and how we can apply breakthrough technologies to deliver maximum benefit for years to come. It also includes imagining. come.
towards the future
We feel great urgency, given the scope of the problems facing humanity, but also great optimism, given what we have already achieved. Our decades of experience shows that Google Research is second to none when it comes to delivering valuable advances. We’ve delivered breakthroughs that shaped Google’s identity as a company, helped launch new fields of computer science, and pushed the frontiers of innovation and technology forward with thousands of publications. The progress we have shared, from everyday tasks to the most ambitious and imaginative initiatives, is already helping people and addressing society’s most pressing challenges, from health care to education to climate change and climate science. I’m working on it.
We will continue to share our breakthroughs on the Google Research blog, conferences, and other events. We are passionate about exploring and creating the future with all the partners and communities we work with.
