As soon as we started the interview with Yossi Matias, vice president of engineering and research at Google and leader of Google Research, this scientist demonstrated that he knows, in detail, the recent meteorological challenges in Portugal. Essential information for those who lead crisis response at one of the largest technology companies in the world. Google was already making Flood Hub available, a platform that predicts flooding in riverbeds using artificial intelligence, crossing satellite images and terrain maps. However, Groundsource, announced this Thursday, represents a qualitative leap in addressing what until now was invisible to machines: flash floods in urban areas.
The heart of this new approach lies in what Matias calls the “magic cycle of research”. According to the engineer, this process consists of “having a curiosity driven by impact, asking the questions that matter”, solving them scientifically and applying the solution to reality. Groundsource uses Gemini to analyze decades of public reports, “turning millions of documents into a high-quality data archive.” More than 2.6 million historical flood events have been identified in more than 150 countries, allowing artificial intelligence models to be trained to predict disasters in urban areas up to 24 hours before they happen.
Personal trauma transformed into global mission
This methodology solves, Matias assures, a critical data gap. The scientist recalls that, in 2018, most experts told him that predicting floods was “totally impossible” due to the complexity of the variables. However, the team persisted, starting with a pilot project in India that protected one million people, evolving into a global hydrological model that now covers two billion citizens.
Yossi Matias’ obsession with alert systems was born from a personal experience. About fifteen years ago he experienced the threat of a forest fire up close. Amidst the tension, he tried to search Google for useful information about what to do or where to run — and found practically nothing. “I was surprised by the lack of information,” he recalls. “I feel distressed every time I read about a disaster because I remember my feelings when I saw that huge fire.” Out of this feeling of helplessness was born a small team within Google dedicated to improving the information available during crises. The objective became a kind of strategic compass. “My North Star”, says Matias, is simple: ensuring that “no one is surprised by a natural disaster”.
Predict to act: from Alentejo to Nigeria
According to the researcher, the usefulness of these systems depends mainly on confidence in the predictions. And, therefore, he cites a recent case in Nigeria, where Flood Hub made it possible to organize an evacuation before a flood reached remote rural regions.
Expanding predictions to the urban environment through Groundsource is “crucial because it is in these areas where population density and rapidity of events make human response more difficult.”
When we asked if Google Earth AI could help countries like Portugal deal with forest fires, the scientist responded affirmatively, explaining the FireSat project. It is a satellite system developed in partnership with several international organizations. The idea is to put around 50 satellites into orbit — one has already been launched — capable of observing every point on Earth every 20 minutes. This resolution “will make it possible to detect fires as small as a classroom”. For the Google VP, this could completely change the way fires are fought. The goal is simple: “stop many forest fires before they even start.”
The digital polymath and the future of science
Google’s vision goes even further. The scientist sees artificial intelligence as an accelerator of scientific discovery. The technology company developed AI Co-scientist, an artificial intelligence system designed to accelerate academic research. This system is capable of carrying out bibliographical research, generating hypotheses and validating theories in multiple disciplines simultaneously. For Matias, this means that each student will have access to a “virtual laboratory” that is more powerful than many research centers. The AI Co-scientist acts as a “polymath in your pocket”, connecting dots between biochemistry, physics and materials science.
“The most exciting trajectory I currently see is the acceleration of scientific discovery itself,” says the Google Research leader. Matias rejects the idea that technology will reduce the number of scientists. On the contrary, he believes there will be more people doing science and asking more ambitious questions about disease, energy and new materials. “AI Co-scientist is already being used to solve complex engineering and mathematical problems,” but the greatest value lies in its ability to democratize access to high-level research.
However, this technological acceleration does not eliminate the human factor. On the contrary, the researcher argues that the scientific method and ethics are now more important than ever. In other words, artificial intelligence should be seen as an “amplifier of human ingenuity”, with researchers having the task of defining what the common good is and ensuring that machines operate in accordance with society’s values. The role of man is to define what is “good” and what the ultimate goal of technology should be. “If it sounds good to us, it’s because it’s good”, he concludes, in reference to Duke Ellington’s phrase.
If Yossi Matias’s optimism is confirmed, the time when populations were helpless victims of the whims of nature may be coming to an end. It is an ambitious commitment that involves partnerships with the academic community and international organizations. For this vice president of Google, the progress made in recent years in areas considered impossible gives us reason to believe that the Earth simulator is closer than we imagine.
