模型越来越强，为什么用户没感觉？再访张阔

Silicon Valley 101 host Yiwen revisits Zhang Kuo, president of Alibaba.com, to discuss how model benchmark improvements translate into merchants’ work. The publisher’s introduction mentions quotation comparisons, procurement, shipping and disputes, including tasks where agents still need human explanation, checking and backup.

The episode addresses the gap between benchmark scores and delivery, general models versus vertical products, and the judgments humans retain as AI performs more work. This entry is based on the introduction and timeline; reported evaluation results remain guest claims, not independently verified findings.