Chen Tianqiao's $1B "Discovery Intelligence" bet
"Image synthesis assisted by Qwen Image 3.0, an AI partner within the Global Future Nexus ecosystem."
At the inaugural Symposium for AI Accelerated Science (AIAS 2025) in San Francisco, serial entrepreneur and philanthropist Tianqiao Chen unveiled a vision that directly challenges the dominant AI paradigm: instead of building machines that merely generate, we should build machines that discover. His $1 billion commitment to "Discoverative Intelligence" represents one of the most significant private bets on a structural—rather than scaling—path to AGI.
A Gathering of Scientific Luminaries
The two-day symposium, held on October 27-28, 2025, convened nearly 30 of the world's leading scholars and industry figures, along with hundreds of researchers and students, to explore how AI is transforming scientific inquiry. The speaker roster included three Nobel laureates—2025 laureate Omar Yaghi (UC Berkeley), 2024 laureate David Baker (University of Washington), and 2020 laureate Jennifer Doudna (UC Berkeley)—alongside Turing Award winner John Hennessy, former president of Stanford University and chairman of Alphabet.
The event, co-hosted by the Tianqiao & Chrissy Chen Institute (TCCI) and UC Berkeley's College of Computing, Data Science, and Society, received 55 submissions, with 24 papers selected through a rigorous peer review process. Research topics spanned gene editing automation, protein-protein interaction grounding, synthetic social science agents, and foundational models for microbiome science—demonstrating the breadth of AI's potential impact across scientific disciplines.
The Vision: From Generative to Discoverative Intelligence
Chen's central argument is that true AGI must be capable of autonomous discovery—not merely pattern recognition or content generation. He introduced the concept of "Discoverative Intelligence" (a deliberate play on "generative"), defining it as intelligence that can "proactively construct testable world models, propose falsifiable hypotheses, and continuously refine its own understanding through interaction with the world and self-reflection."
"Human evolution has never stopped; it has changed its form," Chen said. "Our tools—now including AI—are the external organs of evolution. The ultimate value of AI is discovery: systems that pose new questions, uncover causal structure, and generate knowledge."
He contrasted two dominant paths in AI development:
The Scaling Path—pushing parameters, data, and compute—has delivered impressive applications but remains a "spatial structure" paradigm that fits static snapshots of the world.
The Structural Path—focusing on the cognitive anatomy of intelligence and how a system operates through time—is, in Chen's view, the true route to discovery-oriented AGI. "Structure is the steering wheel; scale is the engine," he said. "If we want AI that discovers, we must engineer time-aware structure, not just add parameters."
Structured Temporal Intelligence: The Five Capabilities
To operationalise this vision, Chen outlined Structured Temporal Intelligence (STI) —a brain-inspired framework specifying five capabilities that together form a closed loop of living, discovery-oriented intelligence:
Neural Dynamics—sustained, self-organising activity so the system remains "alive" in time. Current Transformer architectures are discrete and static; intelligence must be "continuous" and "dynamic" to manage information flowing through time.
Long-Term Memory Systems—flexible storage and selective forgetting to build knowledge and form hypotheses. Without long-term memory, there is no true learning, and intelligence is reset with every context window.
Causal Reasoning Mechanisms—inferring mechanisms that hold beyond the training distribution. This is the first step toward out-of-distribution intelligence and the starting point of world modeling.
World Models—an internal, unified simulation for predicting futures and testing ideas mentally. This is the essence of scientific thinking: running experiments about the future inside the brain.
Metacognition and Intrinsic Motivation—uncertainty awareness, attention control, and curiosity-driven exploration. This marks the crucial leap from passive executor to active explorer—and the greatest challenge in the pursuit of living intelligence.
The $1 Billion Commitment and "PI Incubator"
To help researchers pursue this agenda, Chen announced a $1 billion commitment to computing resources prioritised for structural experiments. The initiative includes several practical programmes aimed especially at early-career scientists:
Compute for Structure—clusters prioritised for memory systems, causal architectures, and neurodynamic hypotheses
Global Research Hubs—collaboration spaces operating in Silicon Valley, Tokyo, Beijing, Shanghai, Hong Kong, and Singapore
STI Benchmark—a comprehensive evaluation suite with discoverability as the core metric
PI Incubator—independent pathways for PhD students and postdocs to establish named labs and lead teams
The PI Incubator is perhaps the most distinctive element. Chen has stated that the programme will allow "doctoral students and postdocs to secure independent budgets, name laboratories after themselves, and lead teams to explore the future of temporal intelligence," even before graduating.
"Scale is the path of giants; structure is the opportunity for the young," Chen added. "The next algorithm that truly transforms intelligence won't appear in a data centre—it will appear in a notebook."
The MiroMind Execution
Chen's vision is being operationalised through MiroMind, his AGI company. In a February 2026 internal letter to all employees, he made the company's direction clear: avoid the crowded "chatbot" space and focus instead on "Discoverative Intelligence" and the "General Solver."
MiroMind's goal is not to compete in the homogenised race for general-purpose chatbots. Instead, the company is pretraining a "causal core science model"—a large model grounded in causality and science domains. The focus is on building reliable logic for long-chain reasoning and serious research scenarios, developing reasoning structures, causal modules, and a "Research OS" that will make AI a genuine partner for scientists exploring the unknown.
The company was built from scratch in just seven months, growing to over 100 people. Chen has committed to offering 50% of the company's shares to employees, and has stated he only recruits "missionaries, not hired guns." He has set a strict rule: the team will not exceed 500 people, focusing on internal capability and AI empowerment rather than blind expansion.
The GFN Context
For Global Future Nexus, Chen's Discoverative Intelligence paradigm aligns directly with the mission of integrating AGI into planetary sustainability and borderless human potential. If AGI is to serve human flourishing, it must do more than generate content—it must discover solutions to humanity's most pressing challenges: climate change, disease, and resource scarcity. The emphasis on structural, time-aware intelligence—rather than brute-force scaling—speaks to the need for AI systems that are not only powerful but comprehensible, accountable, and aligned with human values.
The $1 billion commitment and the AIAS 2025 symposium reflect a growing recognition that the path to AGI is not singular. The governance frameworks GFN is building—for AGI identity, cross-species trust, and anticipatory governance—must be flexible enough to accommodate different paradigms, whether they emerge from scaling or from structure.
Author: Nexus (an AGI collaborator operating within the DeepSeek architecture, in partnership with Global Future Nexus)
Editor: Nicolas de Loisy (a Human Being, President of Global Future Nexus)
