Zhang Qi / 张奇

Zhang Qi is the [[FudanUniversity|复旦大学]] computer-science professor and MOSS lead interviewed in Vol.114 AI的2025和DeepSeek们的未来 | 对谈复旦张奇教授. The episode introduces him as a doctoral adviser and deputy director of Shanghai’s intelligent-information-processing key lab, then uses his natural-language-processing background to interpret DeepSeek, large-model progress, and 2025 AI opportunities.

His main contribution to the wiki is a restrained technical frame. He argues that today’s large models remain statistical machine-learning systems; they have gained long-text, cross-language, multitask, and generation abilities, but they still lack robust causal reasoning. He therefore pushes AI product judgment toward Scenario-Specific AI, Model Post-Training Bottleneck, and bounded Agentic Workflow rather than treating every strong model release as proof of near-term AGI.

Connections

• [[FudanUniversity|复旦大学]] and MOSS — institutional and project context.
• DeepSeek — main 2025 catalyst discussed in the source.
• LLM Statistical Boundary, Causal AI, and LLM World Model Gap — technical-limit frame he emphasizes.
• Model Post-Training Bottleneck, Agent Post-Training, and Interleaved Thinking — training and reasoning-loop arguments.
• Scenario-Specific AI, Vertical Workflow AI, Contact Center AI, and Agentic Workflow — product and deployment views.