AI Goes to Space: Latest Speech by Academician Wang Jian Puts Forward Global Collaboration Initiative of "Three-Body Computing Constellation"

On August 13th, it was reported that Wang Jian, an academician of the Chinese Academy of Engineering, the director of Zhijiang Laboratory, and the founder of Alibaba Cloud, recently attended the 2025 "AI for Good Global Summit" and delivered a keynote speech titled "Computing and AI: Endless Frontiers and Exploration".

During his speech, Academician Wang Jian reviewed Turing's early discussions on the relationship between computing and intelligence and emphasized that computing is not only a tool but also a fundamental science on par with physics and life sciences. He believes that AI is not a replacement for human intelligence but a better tool to extend human creativity. In the practice of Zhijiang Laboratory, AI is used to promote scientific research cooperation and data sharing in fields such as geology, build an open scientific AI architecture, and pay attention to governance mechanisms and open science principles.

He also proposed a large - scale scientific model R & D plan from "0 to 1" and the global initiative of the "Three - Body Computing Constellation", exploring the extension of AI and computing to space to achieve collaborative computing between satellites to address global challenges such as climate change and natural disasters.

The following is the full text of Academician Wang Jian's speech:

Hello, everyone! I am deeply honored to participate in this event. I cherish this occasion very much. "AI for Good" not only stimulates a lot of thinking but also provides a valuable opportunity for global people to communicate. Today, I would like to share Zhijiang Laboratory's research in the field of the relationship between computing and AI and how to use cutting - edge technologies to explore previously inaccessible unknown areas.

Wang Jian said that computers are, to some extent, similar to pens and paper.

When discussing computing and AI, I think they are like two sides of a coin. In the late 1940s, Turing proposed the relationship between computing and intelligence in his first paper. In a report, he expounded on the essence of humanity from the perspective of a "universal machine", and this view is well - known. He once pointed out that pens and paper are extremely powerful tools. Today, although we have computers, their functions are still, to some extent, similar to those of pens and paper. In my opinion, pens and paper have always been excellent tools, and at that time, their functions were comparable to those of today's computers.

The following year, Turing published his first paper on computing and intelligence. I would like to emphasize that he first proposed the concept of the "intelligent machine" and then discussed the relationship between computing and intelligence. It is worth noting that this paper was published in a psychology journal - by the way, my academic background is also psychology. More strikingly, when Turing conceived these ideas, the term "computer" did not refer to a machine but to a person engaged in computing. In an era when there were no computers, Turing had envisioned a "digital computer" - a system in which people or devices performed tasks that humans did at that time.

If we put aside the concept of "computer" and return to the essence of computing, its importance is astonishing. We can understand its significance from two aspects. First, computing is not only a tool but also a way of thinking that helps us think and solve problems. Second, computing is not just a branch of computer science; it is a very fundamental discipline on par with physics and life sciences. That's why AI can thrive based on computing.

The Atlantic called last year's Nobel Prize an important moment for AI.

Last year, the pioneers of AI won the Nobel Prize, which triggered many changes. The Atlantic called it the "penicillin moment" and the "X - ray moment" for AI (meaning an important and glorious moment for this industry). I am particularly pleased that this is highly consistent with the theme of this event, indicating that AI is contributing to human well - being, which makes me excited. At that time, the pioneers were exploring how AI could transform scientific research methods and promote the progress of science and technology.

More notably, when Fortune introduced Geoffrey Hinton, it successively referred to him as a cognitive psychologist, a computer scientist, and the "godfather of AI". Before engaging in AI research, Hinton was a psychologist. Looking back to the mid - 1980s, he co - authored several papers with psychologists. Psychology is a discipline that focuses on the essence of humanity. Therefore, the early pioneers of AI have always been exploring the relationship between AI and humans.

From this perspective, I think there is no direct connection between AI and human intelligence. It is a technology that expands human creativity, a tool, but an excellent tool far beyond pens and paper. With these technologies, humans can achieve many things that were unimaginable without them.

Wang Jian talked about the application of AI in geology.

My passion and the mission of Zhijiang Laboratory are to explore how technology can assist scientists. To this end, we need to deeply understand the needs of science. About ten years ago, a geologist described these needs in a paper, which were simple yet clear: First, share all research data, such as rock - related data; second, ensure the open sharing of research results; third, establish infrastructure to support collaboration. These seemingly simple needs are areas where AI can play a role.

Wang Jian introduced the GeoGPT project.

Therefore, three years ago, inspired by the vision of the "Deep - Time Digital Earth" proposed by the International Union of Geological Sciences (IUGS), we launched the GeoGPT project. In essence, it is an AI system in the field of earth science, meeting the needs described in that paper ten years ago.

Although the GeoGPT project is just a tool, the feedback from global scientists shows that it is extremely valuable for scientific research. We participated in many international conferences, such as the EGU conference held in April this year and United Nations activities, which attracted scientists from all over the world. This is both due to the needs of scientists and provides them with substantial support.

Conducted research on the classification of fossil sponges.

We closely cooperated with Marcus Stevenson to conduct research on the classification of fossil sponges, which is an interesting paleontological work.

With simple AI - driven technology, we expanded the known types of fossil sponges from double - digit numbers to more than 3,000. This breakthrough discovery is astonishing. Without such technology, scientists might not achieve such results in their entire lives. I am particularly pleased that we brought this classification standard to Africa and held a seminar in Nigeria to assist local scientists in their research.

Open AI architecture.

Although focused on earth science, this work has also brought another progress, which is that we built an open AI architecture to promote scientific discovery.

First, we ensure that users are aware of various large - language models, that is, "foundation models", and can freely choose the open - source models they need. This brings two benefits: First, users have the right to make independent choices; second, different models have their own advantages and disadvantages, and users can understand the unique advantages of each model accordingly.

Second, we recognize that in addition to the well - known foundation models, the scientific field also needs "domain - specific foundation models". This type of model is designed for specific applications and needs to process complex scientific data far beyond text. Scientific challenges require innovation beyond language.

On this basis, it is crucial to develop tools that are easy for scientists to use to help them focus on research. I am glad that this architecture is working well.

At the same time, governance is particularly important, especially when developing new technologies. We need to pay attention to issues such as security, privacy, and intellectual property rights. We have established an excellent governance committee for GeoGPT. I'm not sure if it is the only application in the world with such a committee, but we attach great importance to it to ensure that it serves scientists and benefits humanity.

This has become a successful practice of open science. Last year, Michael and others wrote an article about GeoGPT for earth scientists, and we regard it as an example of open science. Facing new technologies, we need to think about their development direction, how the architecture can benefit others, and how to build a good governance mechanism to ensure their positive role.

AI is becoming a fundamental discipline like mathematics.

Based on the experience of GeoGPT, we are extending these explorations to other research fields. In a broader framework, this concerns the integration of "AI + science". More importantly, AI is becoming a fundamental discipline like mathematics. In my opinion, AI is another form of mathematics that will assist all fields of science and technology.

We focus on three tasks. First, build a "large - scale scientific model" codenamed "0 to one". It is different from large - language models because it needs to integrate non - text scientific data and rely on infrastructure to share the results. We have established the zero2x.org website to ensure that global users can share these capabilities via the Internet.

Second, we are exploring the infinite possibilities beyond the Earth and initiating a global initiative called the "Three - Body Computing Constellation". The background is that space has inspired endless imagination. In the late 1940s, when Turing proposed that "a human with a pen and paper is a universal machine", Sir Fred in the UK envisioned taking pictures of the Earth from space. At that time, there were no satellites or space stations, and no one knew what the whole Earth looked like, but this vision greatly promoted our understanding of the Earth.

Astronaut Bill Anders of Apollo 17 took a famous photo. I particularly appreciate his famous quote: "We came to discover the Moon but we actually discovered the Earth." This photo greatly deepened our understanding of the Earth. All current challenges require a deep understanding of the Earth, and computing and AI are important ways to achieve this goal.

Apollo 11 was the first to introduce computing technology into space.

Historically, Apollo 11 was the first to introduce computing technology into space, carrying more than 16,600 transistors. Three years later, Intel's first CPU only contained about 2,000 transistors. Integrated circuit technology was first applied in space rather than on Earth. Space technology not only applies existing technologies but also provides a broad space for exploring new technologies.

Today, we can both explore the technological frontiers and need to address global challenges such as climate change, natural disasters, and land degradation. Earth observation helps us deeply understand the Earth and solve problems. I highly agree with the vision of "Earth intelligence". In the field of AI, this is a very valuable application.

Global challenges urgently require global cooperation and the aggregation and processing of multi - source data in orbit. We don't need to transmit satellite data back to Earth but can process it in space. This has given rise to the idea of introducing AI into space, and AI in space depends on what I call "Space Computing". I have long studied cloud computing, but previous cloud computing was limited to the ground. This is the first time we have introduced computing technology into space.

Computing satellites are expected to become the fourth type of satellite.

This redefines many fields. Currently, we have three types of satellites: communication, navigation, and observation. Now, we are expected to have a fourth type: computing satellites. We hope to build this system in a collaborative way, so we named it the "Three - Body Computing Constellation". Many people may know this term from science - fiction novels, but it comes from Newton's "three - body problem": it is easy to analyze two objects, but adding a third one makes it extremely complex. This means that collaboration is full of challenges, but we are determined to achieve this goal.