Chen Tianqiao's Plan to Build AI That Discovers New Knowledge

Chen Tianqiao, the entrepreneur who built Shanda into a major Chinese gaming company, is now pursuing a different ambition: artificial intelligence designed to discover knowledge rather than only reproduce patterns found in existing data.

Chen calls the concept “discoverative AI.” He says Shanda Group plans to commit more than $2 billion to AI startups, proprietary technology, and the computing infrastructure needed to support the work. His stated goal is not simply to automate tasks people already perform, but to build systems that can tackle problems beyond current human capabilities.

From online games to AI research

Chen became prominent during China's early online-gaming boom. Shanda Interactive Entertainment listed on Nasdaq in 2004 after the success of The Legend of Mir II. He later took the company private in 2012 and sold it in 2014.

After stepping away from the public spotlight, Chen turned Shanda into an investment platform. Its interests have included public markets, venture capital, neuroscience, and brain-computer interfaces.

Chen describes discoverative AI as a system that could combine long-term memory, causal reasoning, and predictive modeling to identify new relationships or anticipate complex events. That is a more demanding objective than generating text or images from patterns learned during training.

The distinction matters. A language model can produce a plausible explanation without proving that it has discovered a new fact. A system intended for scientific discovery would need evidence, repeatable methods, and validation by domain experts.

How the project relates to AGI

Chen's ambition overlaps with the idea of artificial general intelligence, or AGI: a still-theoretical system able to perform a broad range of intellectual tasks rather than specializing in one narrow area.

He has suggested possible applications such as drug discovery, disaster planning, investment analysis, and forecasting geopolitical events. These examples illustrate the goal, but they should not be treated as demonstrated capabilities. High-stakes predictions require reliable data, transparent methods, and careful human oversight.

Researchers disagree about whether AGI is close, how it should be measured, and whether present machine-learning approaches can achieve it. Chen's project is therefore best understood as a long-term research program, not proof that human-level or superhuman general intelligence has already been built.

Brain research shaped Chen's direction

Chen has spoken about experiencing panic attacks and later becoming deeply interested in the human brain. In the United States, he committed substantial funding to neuroscience and work associated with the California Institute of Technology.

Shanda has also invested in brain-computer interface research. BCI systems translate neural activity into digital commands and are being investigated for medical and assistive uses. Research in this area may inform ideas about memory, learning, and cognition, although a better model of the brain does not automatically produce general-purpose AI.

Building dedicated computing infrastructure

Alongside research and startup investment, Chen has outlined plans to spend an additional $1 billion on data-processing and computing infrastructure. He says the objective is to explore new cognitive architectures rather than compete only by making existing models larger.

His infrastructure proposal includes a pilot U.S. computing center with capacity of several megawatts. Chen has discussed geothermal power and the use of land held in Oregon and Ontario as part of a longer-term clean-energy strategy for data centers. The project remains a plan, and its final location, capacity, cost, and energy mix may change as engineering and regulatory work progresses.

An ambitious project with open questions

Chen rejects the idea that the current AI boom is merely a short-lived technology bubble. His optimism is reflected in the scale of the proposed investment and his belief that machine intelligence could become a major historical turning point.

Large funding commitments, however, do not resolve the hardest scientific questions. The project will ultimately be judged by whether it produces testable discoveries, independent validation, and systems that can operate safely in real-world settings. Until then, “discoverative AI” is a research direction and a statement of intent rather than an established category of technology.