AI for Science, where has it reached?
Recently, Google DeepMind published an article titled "AlphaFold: Five Years of Impact," reviewing the significant role of technological breakthroughs in protein structure prediction over the past five years in driving scientific progress.
Artificial intelligence is reshaping the landscape of scientific research at an unprecedented pace. Among numerous scientific research fields, biological fields such as life science and biomedicine have become the most active and leading frontiers of AI + scientific research (hereinafter referred to as scientific intelligence) due to factors such as abundant data, clear application scenarios, and urgent social needs. AI models and tools have not only made breakthroughs in basic research such as protein structure prediction but also promoted new drug pipelines into clinical trials and even started to independently discover new biological pathways.
Image: Google DeepMind AlphaFold
Led by technology companies such as Google DeepMind that have been deeply involved in AI for Science, scientific intelligence represented by biology is entering a period of high output and rapid iteration in application implementation. The AI - driven scientific research paradigm of "foundation model + scientific research agent + autonomous laboratory" is gradually taking shape.
Google DeepMind
Leading the Evolution of Scientific Intelligence Technology
Google has been deeply involved in scientific research for more than a decade. Relying on its AI computing power infrastructure centered around TPU and the AI model base based on the Gemini large - foundation model, Google has continuously developed scientific intelligence technologies, creating world - class scientific intelligence models and tool systems such as AlphaFold and leading the global evolution of scientific intelligence technology.
AlphaFold Leads the Leap in Biological Research from Structure Prediction to Generative Design. Biology is the field where DeepMind made its earliest layout and has the deepest moat. Its core logic lies in using deep learning to solve the conformational space problem of high - dimensional biological macromolecules. The emergence of AlphaFold marks the substantial solution to the protein structure prediction problem. It not only won the 2024 Nobel Prize in Chemistry but also became the digital infrastructure of modern biology. AlphaProteo has officially pushed biological research into the era of generative biology. Combining with AlphaMissense's accurate prediction of the pathogenicity of gene mutations, DeepMind has further connected the entire chain of "target discovery - structure analysis - drug design."
WeatherNext Achieves a Dimensionality - Reduction Blow to Numerical Simulation with Data - Driven Meteorology. The data - driven method of the newly released WeatherNext 2 model (successor to GraphCast) by DeepMind has comprehensively surpassed traditional physical models in terms of accuracy and efficiency. WeatherNext 2 is more accurate than the HRES system of the European Centre for Medium - Range Weather Forecasts (ECMWF) in 99.9% of prediction variables and time spans, and its inference speed has been improved by several orders of magnitude.
GNoME and AlphaQubit Expand the Application of AI in Physics and Materials Science. GNoME (Graph Networks for Materials Exploration) uses deep learning to conduct a massive search in the inorganic crystal space, predicting millions of stable new material structures. Its scale is several times the total number of experimental discoveries by humans in the past few decades, providing a huge candidate library for the research and development of battery technology and superconducting materials. In the field of quantum computing, the AlphaQubit model has successfully applied the Transformer architecture to quantum error correction, significantly reducing the quantum bit reading errors of quantum computing chips.
AlphaEvolve Promotes the Self - Evolution of Mathematics and Computer Science from Logical Reasoning to Algorithms. By introducing the evolutionary computing paradigm, AlphaEvolve aims to break the limitations of human - designed algorithms, automatically search for and discover more efficient machine - learning algorithms and loss functions, achieving a meta - level leap from "manual design" to "automatic discovery." On this basis, AlphaChip has transformed chip design into a reinforcement learning problem and successfully optimized the layout of Google TPU v6; AlphaGeometry and AlphaProof have demonstrated AI's breakthroughs in formal mathematical proof and logical reasoning.
Progress in the Biological Field
Leading the Frontier of Scientific Intelligence Implementation
The technological breakthroughs led by Google DeepMind have ignited a global boom in the research and development and industrial application of scientific intelligence. Biology has become the scientific research field with the fastest progress, followed by materials science, physics, meteorology, computer science, and mathematics.
(I) Scientific Intelligence Enters the Deep Water Area of Biological Basic Research
AI - Generated and Analyzed Single - Cell Behavior Yields New Discoveries. Google and Yale University jointly released the 27 - billion - parameter single - cell analysis foundation model C2S - Scale, which generated new hypotheses about cancer cell behavior and was verified in multiple in vitro experiments. This demonstrates the potential of using AI to propose original scientific hypotheses and is expected to explore new ways to develop anti - cancer methods.
The Protein Generative Simulation and Prediction System is More Complete. Microsoft's BioEmu model has filled the gap in protein dynamics simulation and achieved a simulation speed increase of up to 100,000 times. A research team from the Chinese Academy of Sciences proposed an inverse folding protein prediction model that integrates structural and evolutionary constraints, opening up a new path for protein engineering. The relevant results were published in the journal Cell.
The AI - Assisted Genomics R & D System is Initially Constructed. Through 10 years of continuous research and exploration, Google has gradually constructed an AI genomics research and application system from gene sequencing, reading, and mutation identification to gene expression prediction and pathogenic potential assessment, and then to disease gene detection and diagnosis, which helps promote the development of genetics and gene therapy.
(II) AI - Driven Medical Applications are in Full Bloom
AI - Assisted Pathological Detection Expands to New Disease Scenarios. The DeepGEM pathological large model jointly developed by Tencent Life Science Laboratory, the First Affiliated Hospital of Guangzhou Medical University, and the Guangzhou Institute of Respiratory Health has completed large - scale verification in the prediction of lung cancer gene mutations. It can complete the prediction of lung cancer gene mutations within 1 minute using only routine case slice images, with an accuracy rate of 78% - 99%.
AI - Based Detection of Gene Mutations is Further Tooled. Google released the DeepSomatic toolset for the identification of gene mutations in tumor cells, which is applicable to cancer types such as leukemia, breast cancer, and lung cancer, and its identification accuracy is better than existing solutions.
AI - Driven Drug R & D Crosses the Phase II Clinical Stage. The AI - optimized candidate drug MTS - 004 jointly developed by Peking University Third Hospital and other hospitals and Jitai Technology has completed Phase III clinical research, becoming the first AI - empowered new drug formulation to complete Phase III clinical trials in China. The drug is expected to target neurological diseases such as amyotrophic lateral sclerosis and stroke. The above progress has broken through the bottleneck that there have been few breakthroughs in AI - driven drug discovery beyond Phase II clinical trials in China and even globally in the past few years, attracting wide attention and recognition at home and abroad.
(III) The Application of AI in Materials Science, Meteorology, Mathematics, etc. is Accelerating
AI + Materials Science is Expected to Become the Next Frontier of Scientific Intelligence. Former members of OpenAI and Deepmind founded Periodic Labs to conduct AI - automated discovery of new superconducting materials. CuspAI, which just received $100 million in Series A financing, is developing an AI platform to discover new carbon - capture materials. Dingxi Zhichuang, incubated from the Shenzhen - Hong Kong Hebao Science and Technology Innovation Center of Peking University Shenzhen Graduate School, is building the RhinoWise materials innovation platform to carry out innovation in key materials in fields such as new energy and semiconductors.
The Application of AI in Meteorology and Physics has Achieved Practical Results. DeepMind's hurricane AI model has successfully predicted the path and intensity changes of super - hurricanes such as "Melissa," helping the United States and neighboring countries issue early warnings. Black - hole theoretical physicist Alex Lupsasca used GPT - 5 to derive new characteristics of black - hole theory within half an hour. The nuclear fusion startup CFS uses Google's open - source TORAX tool to assist in the research and development of the SPARC nuclear fusion device.
The Application Potential of AI in Mathematics and Computer Science is Huge. Mathematical researchers are using GPT5 to explore solutions to the historical mathematical problem - the Erdős problem. Google is promoting research in mathematics and theoretical computer science based on AlphaEvolve. The open - source model system GenCluster of NVIDIA won the gold medal in the IOI 2025 competition. Large models such as OpenAI's internal model, Gemini Deep Think, and DeepSeek Math - V2 are also constantly refreshing the gold - medal results of AI in the International Mathematical Olympiad.
Technological Foundation, Collaboration Model, and Research Scale
Three Dimensions for AI to Reshape the Scientific Research Paradigm
From the progress of scientific intelligence represented by biology, it can be seen that AI's reshaping of scientific research is systematic. It is changing the traditional thinking of scientific discovery from three dimensions: technological foundation, collaboration model, and research scale. The AI - driven scientific research paradigm of "foundation model + scientific research agent + autonomous laboratory" is gradually taking shape.
(I) General Models and Specialized Models Build the Technological Foundation of Scientific Intelligence
General Large - Foundation Models are Expected to Become the "Operating System" of Scientific Intelligence. General large - foundation models can provide powerful capabilities in understanding, reasoning, analysis, and generation. They also have comprehensive scientific basic knowledge and general knowledge reserves, which can significantly improve the daily efficiency of scientific researchers. At the same time, leading companies in large - model research are constantly improving the scientific research professional capabilities of foundation models. Anthropic's Claude Sonnet 4.5 has significantly improved in understanding and applying life - science task processes and has enhanced the ability to use scientific research tools and resources based on the foundation of agent capabilities and connectors.
Specialized Large - Models for Scientific Research are the "Specialized Engines" for Vertical Scientific Research Fields and Their In - Depth Breakthroughs. These models usually integrate relevant knowledge, research methods, and experience in specific fields. Google is globally leading in the comprehensive strength of specialized large - models for scientific research. Its specialized models and algorithms cover various fields such as life science and biology, materials science and chemistry, earth and climate science, mathematics, and basic science. The aforementioned C2S - Scale, BioEmu, DeepGEM, etc. also belong to this type of model. In addition, the Panshi·Scientific Large - Foundation Model jointly developed by a research team from the Chinese Academy of Sciences is also a beneficial practice of integrating foundation models and specialized models.
(II) Scientific Research Agents Based on Human - Machine Collaboration Start to Promote Active Scientific Discovery.
AI handles trivial and time - consuming but indispensable research links, while human scientists control the research direction and evaluate research results. This will become a typical human - machine collaboration research model in the future.
With the accelerated development of agent technology, AI is transforming from a passive tool to a collaborator or even an active discoverer for scientists. Harvard and MIT jointly launched the ToolUniverse, a scientific research tool platform specially designed for AI agents, which contains more than 600 scientific tools and is compatible with mainstream large - foundation models. This helps inspire more scientific researchers to build agent scientists for specific scientific research fields. AlphaEvolve, released by Google Deepmind, is an evolutionary AI agent with coding capabilities, which can actively discover and automatically optimize general mathematical and computational algorithms. It has been applied in actual scenarios such as data - center scheduling, chip design, and large - model performance optimization within Google. A joint research team from the Shanghai Artificial Intelligence Laboratory, Zhejiang University, etc. proposed the concept of scientific research agents (Agentic Science), aiming to build an AI system that can autonomously complete the scientific research closed - loop.
(III) Autonomous Laboratories Accelerate the Industrialization, Scaling, and Platformization of Scientific Intelligence
AI and robotics technologies are upgrading traditional "workshop - style" laboratories that rely on manual trial - and - error to automated, high - throughput, and closed - loop "science factories," which are connected to form platforms to serve the entire scientific research ecosystem.
Countries around the world attach great importance to the research and development of autonomous laboratories. Many scientific research universities and national laboratories in the United States, such as MIT, have built autonomous laboratories. The Materials Innovation Factory (MIF) at the University of Liverpool in the UK is one of the most advanced autonomous laboratories in Europe. The IKTOS laboratory in France, the Atinary SDLabs in Switzerland, the FULL - MAP project in Germany, etc., are all powerful autonomous laboratories that continuously contribute in fields such as chemistry and new materials. At the same time, international startups such as Lila Sciences and Periodic Labs, which recently received hundreds of millions of dollars in financing, are all targeting this field. Meanwhile, the "Genesis Mission" recently launched in the United States lists advanced manufacturing technology as the top priority for technological breakthroughs. One of its main goals is to accelerate the construction of next - generation scientific research infrastructure such as autonomous laboratories to improve the efficiency of AI - driven scientific discovery and industrial application transformation. This plan further integrates scientific research computing power, AI foundation models, relevant data sets, and the autonomous laboratory system into a science and security platform as the scientific intelligence infrastructure.
The Construction of Domestic Autonomous Laboratories and Scientific Intelligence Platforms has been Fully Launched. The AI + robotics platform of Jingtai Technology has become its core competitiveness. The "ChemBrain Agent + ChemBody Robot" of the Chinese Academy of Sciences and the Uni - Lab - OS intelligent operating system of the Beijing Academy of Scientific Intelligence are also aimed at accelerating the research, development, and promotion of domestic autonomous laboratories. The Panshi·Scientific Large - Foundation Model developed by the Chinese Academy of Sciences is also an important practice of domestic scientific intelligence platforms. The platform can manage various resources such as data and models and schedule various scientific research tools. It has now started to be applied in fields such as life science, high - energy physics, and mechanics research.
AI for science,
science for humanity
In the next few years, the evolution speed of scientific intelligence technology and the efficiency of application value transformation will be further improved with the continuous improvement of the capabilities of AI large - foundation models and the continuous maturity and scaling of robotics technology. The scientific intelligence research paradigm will become more mature, and the scientific research ecosystem will be reconstructed and upgraded. More significant discoveries will emerge from AI - driven scientific research. Sam Altman predicted at this year's Sequoia Capital AI Summit that AI large - models will make scientific discoveries comparable to the theory of relativity by 2028.
However, while the technology is being updated at a high speed, we cannot ignore the improvement of human beings' original scientific research capabilities as the main body of scientific discovery and the renewal of scientific and technological ethics and responsibilities. Scientists should always be the yardstick of scientific intelligence, ensuring that AI becomes the promoter of human scientific and technological evolution and the guardian of human civilization's continuation.
This article is from the WeChat public account "Tencent Research Institute" (ID: cyberlawrc), author: Liu Moxian. Republished by 36Kr with permission.