AI for Science: Searching for the Next Nuclear Explosive Point Intelligence | Artificial Intelligence | AI for Science

Can AI solve the three body problem—— In the field of scientific intelligence, many people use the example of the Three Body Problem.

Liu Cixin's novels have made the "Three Body Problem" well-known. Simply put, the Three Body Problem refers to the problem of the motion laws of three celestial bodies with arbitrary masses, initial positions, and initial velocities under the influence of universal gravity.

In the novel, there is no solution to the three body problem, but the emergence of artificial intelligence has made it possible to solve this ancient problem that has troubled scientists for hundreds of years. The underlying principle is actually to use the powerful data induction and analysis capabilities of artificial intelligence to learn scientific laws and principles, derive models to solve practical scientific research problems - AI driven scientific research, also known as AI for Science.

However, AI for Science is not complete in a sense. Just like reading ten thousand books, one still needs to travel ten thousand miles. Someone added the latter half of the sentence: AI for Science, then for industry. It has gradually become a consensus in the industry to first use AI to solve the underlying scientific principles of things, and then further solve industry problems.

Looking up at the starry sky

On the opening day of the 2023 World Artificial Intelligence Conference, engineers from Huawei's booth forwarded a message on their social media: the top international academic journal Nature has published the research result of Huawei's Cloud Pan Gu Model R&D Team - "Three dimensional neural network for accurate mid-term global weather forecasting".

This is the first paper in recent years to be published in the main issue of Nature by a Chinese technology company as the sole signatory. The reviewer of "Nature" commented on the achievement: "The Huawei Cloud Disk Ancient Meteorological Model allows people to re-examine the future of meteorological forecasting models, and the opening of the model will promote the development of this field."

The research results of Huawei Cloud Disk's ancient meteorological model have been published in the official issue of Nature

Weather prediction is one of the research fields driven by artificial intelligence.

The Huawei Cloud Pangu Big Model R&D team has found that there are two main reasons for the insufficient accuracy of AI meteorological forecasting models: firstly, the original AI meteorological forecasting models are based on 2D neural networks and cannot handle uneven 3D meteorological data well; Secondly, AI methods lack mathematical and physical constraints, thus accumulating errors during the iteration process.

To this end, the team proposed a three-dimensional neural network adapted to the Earth coordinate system to process complex and uneven 3D meteorological data, and through innovative means, reduced the number of forecast iterations, thereby reducing errors. By training deep neural networks on 43 years of global weather data, the Pangu meteorological model surpasses traditional numerical prediction methods in accuracy and speed, with a speed increase of more than 10000 times compared to traditional numerical forecasting.

In addition to predicting weather, AI also helps scientists observe celestial phenomena. Also at the World Artificial Intelligence Conference, Tencent announced for the first time that it had used AI to assist the world's largest radio astronomical telescope, the China Tianyan FAST, in locating two fast radio bursts.

"Based on Tencent Cloud+AI, combined with Tencent Cloud's industry model and powerful computing power, the speed of searching for fast radio bursts has increased by 1800 times." According to Wang Yabiao, an expert researcher at Tencent Youtu Laboratory, fast radio bursts are a mysterious astronomical phenomenon that emits the energy released by the sun for a whole year in just one millisecond, "flickering" the universe.

AI helps FAST capture the "flicker" of the universe

The study of rapid radio bursts helps humanity better understand the origin of the universe. This research work is one of the key tasks of China Tianyan and has also become a key direction for Tencent Cloud AI for Science. In 2023, the Star Exploration Program captured two flashes of the universe through AI technology for the first time.

"By driving scientific research through AI and breaking through the bottleneck of traditional scientific research capabilities, we have made breakthrough progress in fields such as astronomy, meteorology, and life sciences," said Wang Yabiao.

Aiming at the forefront

AI can not only help scientists look up at the vastness of the universe, but also drive humanity forward in the forefront of science. After Einstein's death, scientists conducted decades long research on his brain. Exploring the workings of the brain has fascinated scientists, and neuroscience has become an important arena in global technology competitions.

In recent years, countries such as Switzerland, the United States, and Japan have successively launched research projects targeting the brain. China officially launched the "China Brain Plan" in 2021, elevating "brain like computing and brain machine intelligence" to a higher priority for development. China's "14th Five Year Plan" includes "technology development of brain like computing and brain machine integration" as one of the five key areas of the brain science plan.

Not long ago, the Graph Computing Laboratory of Ant Technology Research Institute and the Neuropharmacology Laboratory of Fudan University Brain Science Research Institute jointly launched a brain like research project - a brain simulation architecture based on graph computing. Through the close integration of graph computing, artificial intelligence, and experimental neuroscience, a new generation of large-scale high-precision brain simulation systems will be developed, providing more efficient technical means for understanding biological intelligence and treating brain diseases.

Chen Wenguang, Dean of Ant Technology Research Institute

What is graph calculation? What is brain simulation? Chen Wenguang, Dean of Ant Technology Research Institute and Head of Ant Graph Computing Technology, introduced that graph computing is a computing model that constructs correlation relationships based on points and edges. Its biggest advantage is that it can perform dynamic and real-time analysis of point edge relationships on a scale of billions.

"Using graph computing for brain simulation can restore the structure of neurons with high precision, simulate biological experiments to get data, and further combine biological experiments to verify each other, create a 'ring' for mutual verification of biological experiments and computer experiments, and finally achieve high-precision brain simulation," Chen Wenguang said.

"Science and technology is a spiral upward process that requires adjusting the research pace according to the maturity of technology at different stages." Wang Xiaofei, a member of the project team and an engineer at the Institute of Brain Science at Fudan University, said that launching graph based brain like research is not only a breakthrough in the application boundaries of technology development maturity, but also a window for cutting-edge technology to assist cutting-edge scientific research.

Driving industries

If the starry sky is too distant and the human brain is too mysterious, looking back, scientific intelligence is actually changing thousands of industries, and biomedicine is one of the industries that AI is accelerating to rewrite.

In the field of drug development, the "Double Ten Law" is often mentioned. It takes 10 years and an investment of 1 billion US dollars to successfully develop a new drug. The long cycle, high cost, and low success rate of new drug research and development have become a challenge in the industry's development. Until 2020, in the international protein structure prediction competition, AlphaGo - AlphaFold in the biological community accurately predicted the 3D structure of proteins, which means that the speed of new drug discovery will be greatly accelerated.

Blue represents AI prediction results, green represents standard answers

According to the top international academic journal Nature, investing heavily in artificial intelligence can increase the profits of the pharmaceutical industry by over 45%.

At this year's World Artificial Intelligence Conference, Jingtai Technology was selected as the "Treasure of the Town" and the highest award SAIL Award for its "AI Drug Research and Development Automation Solution" and "Intelligent Automated Drug Research and Development Platform", winning two major awards at this year's Artificial Intelligence Conference.

Ma Jian, co-founder and CEO of Jingtai Technology, introduced that Jingtai Technology is a technology company that has taken the lead in applying "GPT like technology" to the research and development of macromolecular drugs. The ProteinGPT model developed by Jingtai Technology has made great progress in solving the problem of one bond drug for macromolecules and has been applied in the development of industrial grade macromolecular drugs with good landing effects. It comprehensively covers scenarios such as antibody discovery, antibody modification, antigen design, and protein de novo design.

He further explained, "Jingtai Technology's ProteinGPT technology roadmap shares similarities with ChatGPT, and there are many innovative improvements that can truly be implemented in drug development, known as the biological version of ChatGPT."

AI pharmaceutical company Yingsi Intelligence is also making rapid progress. A candidate drug designed by AI for idiopathic pulmonary fibrosis has been administered to the first patient in a phase II clinical trial. "This drug has the potential for anti fibrosis and anti-inflammatory effects, and is also the first AI designed drug to enter the clinical stage," said Dr. Ren Feng, CEO and Chief Scientific Officer of Yingsi Intelligence. "The completion of the first phase II clinical trial is an important progress in the AI pharmaceutical field."

Layout for the future

On July 6th, the "China AI for Science Innovation Map Research Report" jointly written by the China Institute of Science and Technology Information, the New Generation Artificial Intelligence Development Research Center of the Ministry of Science and Technology, and relevant research institutions was released at the World Artificial Intelligence Conference. The report shows that China has the highest number of published papers on scientific intelligence and is actively promoting the development of global scientific intelligence.

"AI for Science is increasingly showing great potential to break through the bottleneck of traditional scientific research capabilities and is becoming a new paradigm for global scientific research." Zhao Zhiyun, Secretary of the Party Committee and Director of the China Institute of Science and Technology Information, said that in recent years, China's scientific intelligence field has closely followed the world's leading edge, with high research enthusiasm and rapid capacity improvement, forming a good development trend.

Development achievements of AI for Science in China, the United States, and Europe

Huawei's rotating chairman Hu Houkun also believes that the key to the development of artificial intelligence is to "go deep and practical" and empower industrial upgrading. "We hope that AI for Science can bring more new ideas, methods, and tools to scientists and scientists, and also provide new impetus for our industry."

For an innovative city, there need to be several "nuclear explosion points", which are the source of innovative ideas and the turning point of the new track. The emergence of scientific intelligence is undoubtedly nurturing more "nuclear explosion points". At present, Shanghai has identified 16 sub sectors of five future industries, including brain computer interfaces, synthetic biology, quantum technology, new energy storage, and aerospace utilization, all of which are closely related to scientific intelligence.

Industry insiders point out that from the perspective of international competition, scientific intelligence can quickly improve a country's technological strength. AI and Science will promote each other, providing greater competitive advantages for scientific research in various countries and accelerating the formation of high-tech barriers.

In Zhao Zhiyun's view, the significance of scientific intelligence lies not only in finding the next "nuclear explosion point", but also in the hope of "leading the birth of the next world science center in China." This is consistent with Shanghai's goal of building a technology innovation center with global influence.