Chen Tianqiao: Mainstream AI are 'Liberal Arts Large Models', MiroMind to Build 'Science Large Models' — Losing Access to AI is Scarier Than Water/Electricity Cuts

Chen Tianqiao classifies current large models into two paradigms: mainstream 'liberal arts large models' focus on language generation and text consistency, excel at simulation, and will become infrastructure in education and communication. MiroMind aims to build 'science large models' that focus on 'discovery', track causal relationships, verify hypotheses, and produce new knowledge.

MiroMind's MiroThinker 1.5 with 30B parameters set a new record in the BrowseComp benchmark, topping the FutureX global rankings. The team's goal is to maintain nearly 99% accuracy in complex reasoning chains of over 300 steps, with each step verifiable and traceable.

On AI ethics, Chen Tianqiao believes the core issue is 'AI access rights' rather than privacy or fairness. Stratification of AI capabilities will lead to cognitive fragmentation in humans. If a few giants control AI, cutting off access will be scarier than water or electricity cuts. He calls on governments to encourage open-source models to protect ordinary people's right to choose AI access.