1. Overcoming the 50-year “grand challenge” of protein structure prediction
For decades, experts have described unraveling protein folding as a “major challenge.” In 2022, Google DeepMind shared the predicted structures of 200 million proteins from its AlphaFold 2 model. Previously, it typically took more than a year to determine the 3D structure of a single protein, but AlphaFold can predict these shapes in minutes with remarkable accuracy. By publishing protein structure predictions in a free database, scientists around the world can now accelerate progress in areas such as developing new drugs, fighting antibiotic resistance and tackling plastic pollution. As a next step, the AlphaFold 3 model will be built on AlphaFold 2 to predict the structure and interactions of all molecules of life.
2. Showing the human brain in unprecedented detail to support health research
Few things have remained as mysterious for as long as the human brain. Building on more than a decade of connectomics research, Google has partnered with Harvard University’s Lichtman Institute and others to map tiny parts of the human brain at a level of detail previously unattainable. The project, announced in 2024, revealed structures inside the human brain that had never been seen before. Additionally, a complete dataset with AI-generated annotations for each cell is publicly available to accelerate research.
3. Save lives with accurate flood predictions
When Google’s Flood Prediction Project started in 2018, many thought it would be impossible to provide accurate large-scale flood predictions due to a lack of data. However, the researchers achieved similar or better reliability than nowcasts (with a lead time of 0 days) in predicting extreme river events in ungauged watersheds with lead times of up to 5 days. We were able to develop an AI model. In 2024, Google Research will expand this reach to 700 million people in 100 countries around the world and improve its AI model to deliver the same accuracy in 7 days as the previous model had a lead time of 5 days. We have made it available.
4. Detect wildfires earlier so firefighters can stop them faster
As the climate gets hotter and drier, wildfires are increasingly disrupting communities around the world. In 2024, Google Research will partner with the U.S. Forest Service to launch an AI model and new global satellite constellation designed to detect and track classroom-sized wildfires by providing high-resolution imagery in under 20 minutes. I developed a FireSat. This allows fire authorities to respond more quickly, potentially saving lives, property and natural resources.
5. Predict the weather faster and more accurately
In 2023, Google DeepMind announced and opened the model code for GraphCast, a machine learning research model that predicts weather conditions up to 10 days in advance, more accurately and much faster than the industry’s gold standard Weather Simulation System (HRES). I made it into a source. GraphCast was also able to more accurately predict the path of a cyclone (and related risks such as flooding), accurately predicting that Hurricane Lee would hit Nova Scotia three days earlier than traditional models.
6. Advancing the frontiers of mathematical reasoning
AI has always struggled with complex mathematics due to a lack of data and reasoning skills. And in 2024, Google DeepMind launched AlphaGeometry, an AI system that solves complex geometry problems at a level approaching that of a human Olympic gold medalist. This is a breakthrough in AI performance and a pursuit for more advanced general AI systems. The Gemini-trained model AlphaGeometry 2 was then combined with the new model AlphaProof, and together they solved 83% of the historic International Mathematics Olympiad (IMO) geometry problems over the past 25 years. Demonstrating that AI’s increasing reasoning power has the potential to solve problems beyond current human capabilities brings us closer to systems that can discover and validate new knowledge. .
7. Accurately predict chemical reactivity and reaction rates using quantum computing
Google researchers, in collaboration with the University of California, Berkeley, and Columbia University, have run the largest chemical simulation ever on a quantum computer. The results, published in 2022, were not only competitive with traditional methods, but also didn’t require the cumbersome error mitigation typically associated with quantum computing. The ability to perform these simulations allows for more accurate predictions of chemical reactivity and kinetics. This is a precursor to applying chemistry in new ways to help solve real-world challenges.
8. Accelerating materials science and the potential for more sustainable solar cells, batteries, and superconductors
In 2023, Google DeepMind announced Graph Networks for Materials Exploration (GNoME), a new AI tool that has already discovered 380,000 materials that are stable at low temperatures, according to simulations. At a time when our world is seeking new approaches to energy, processing power, and materials science, this research could pave the way to better solar cells, batteries, and potentially superconductors. there is. Additionally, to ensure that this technology benefits everyone, Google DeepMind has made GNoME’s most stable predictions available through the Materials Project on an open database.
9. Taking meaningful steps towards nuclear fusion and abundant clean energy
As the old joke goes, “Nuclear fusion is the energy of the future, and it always will be.” Controlling and using the energy that fuels stars, including our sun, is an area of science. exceeds. Then, in 2022, Google DeepMind announced that it had developed an AI that can autonomously control the plasma in a fusion reactor. By collaborating with EPFL’s Swiss Plasma Center, Google DeepMind has built the first system that can autonomously stabilize and shape plasma inside an operating fusion reactor, delivering stable fusion and abundant clean energy. We have taken an important step towards providing it for everyone.
