Huawei, Tencent, and Alibaba Cloud are all leveraging the seemingly difficult fields of medicine, astronomy, and meteorology to make weather forecasts | Time | Alibaba Cloud

In the new era of the integration of "big computing power", "big data", and "big knowledge", scientific intelligence, as one of the hot fields, is becoming a powerful tool to promote the transformation of scientific research paradigms and achieve the original innovation source of science and technology. At the 2023 World Artificial Intelligence Conference venue, reporters noticed that many companies are applying AI technology to scientific research, technological innovation, and achievement transformation in various disciplines. At the booth of Jingtai Technology, an automatic robotic arm is operating flexibly in a small workstation. The AI drug research and development automation solution represented by this has been named one of the treasures of this year's World Artificial Intelligence Conference. "This is a super muscle in the real world, connected to the AI platform's super brain, which can free up the hands of R&D personnel, reduce simple repetitive experiments, and focus more on creative work. For example, designing small molecule drugs for targets can be virtual predicted and screened through algorithms, reducing up to 90% of experimental requirements." The person in charge of the exhibition booth explained.

Compared to traditional manual methods, automation systems have higher accuracy and can record standardized and traceable process data and positive and negative result data throughout the entire process. This allows artificial intelligence algorithm models to continuously learn and iterate, better design potential new drug molecules, optimize reaction processes, and form a feedback loop between artificial intelligence, automated experiments, and data. The high threshold of precision and safety in the pharmaceutical industry gives Jingtai Technology an inherent advantage in expanding into other fields. Currently, AI+automation related solutions have been successfully applied in industries such as chemical, testing, functional materials, and bio based materials, and are actively exploring more applications in new materials, new energy, and other fields, such as ceramic materials, lithium-ion batteries, and perovskite batteries. In addition to AI pharmaceuticals, scientific intelligence can also help scientists "look up at the stars". From satellite exploration to meteorological prediction, AI is constantly expanding new possibilities. The Huawei Cloud Disk Ancient Meteorological Model showcased at this conference is the world's first AI model with accuracy exceeding traditional forecasting methods. It can achieve hourly and highest accuracy meteorological forecasting within a range of 20 kilometers, with a prediction accuracy of over 85% for typhoon generation time and movement. The on-site staff of Huawei booth introduced that the Pangu Meteorological Model has achieved meteorological forecasting accuracy that exceeds traditional numerical methods for the first time. With the help of innovative network structures and hierarchical time aggregation algorithms, the accuracy of key elements and commonly used time ranges in meteorological forecasting such as humidity, wind speed, and temperature has exceeded the current forecasting method, and the speed has increased by more than 1000 times compared to traditional methods. Tencent has also applied AI technology to the field of astronomy. In 2021, the United Nations Astronomical Observatory launched the "Star Exploration Program", which has helped the Chinese Sky Eye - China's 500 meter Aperture Spherical Radio Telescope - to quickly discover 30 pulsars. This year, with the assistance of AI technology, researchers discovered two rapid radio bursts for the first time from sky survey observation data. "Compared to pulsars, fast radio bursts are much more difficult to detect due to their late discovery time, limited AI training data, and low frequency of occurrence." The on-site staff explained that the team has significantly improved the accuracy and data processing speed of the algorithm model by introducing a newly designed end-to-end AI algorithm and attention mechanisms in multi instance learning and large model technology. Standing on the high ground of theoretical research, universities are more closely participating in the research of scientific intelligence, leveraging the foundation and advantages of talent cultivation, discipline construction, and scientific research in the field of artificial intelligence to accelerate important breakthroughs in key technologies. At the Alibaba Cloud booth, the focus was on showcasing the largest cloud based research and intelligent computing platform CFFF in domestic universities. This is a scientific research "supercomputer" built to discover and solve complex scientific problems, providing over a thousand card parallel intelligent computing with advanced public cloud models and supporting large-scale model training with billions of parameters.

It is reported that based on the 100G high-speed data transmission network and Alibaba Cloud's large-scale heterogeneous computing power fusion scheduling technology, hierarchical storage technology, AI and big data integration technology, the "Jinsi" No.1 deployed on Fudan campus and the "Qiewen" No.1 hosted at Alibaba Cloud's Ulanqab data center 1500 kilometers away have connected to form a true "supercomputer". All experimental equipment in Fudan Campus 4 can be accessed at high speed, achieving unified management of heterogeneous computing power and unified scheduling of computing tasks, meeting the scientific and intelligent research and application needs in different application scenarios.