Ling Guang 2.0 test: the permission and liability gap
Ant Group launched Ling Guang 2.0 on October 27, repositioning it from a natural-language app generator to a personal AI assistant for managing everyday tasks through services such as Alipay, DingTalk, Feishu and Mi Home. GeekPark tested the product for two days against Qwen, Muse and Claude Duet, finding that it could retrieve bills, read office data and create documents, but was blocked from sending messages, loading Taobao carts, controlling Mi Home devices and completing payments. Its analysis concludes that cross-platform permissions, liability allocation and commercial incentives are now more consequential barriers than model reasoning alone.
The test shows that personal AI assistants still face major operational limits when they must act across platforms, even when they can understand and monitor user information.
作者|汤一涛
编辑|靖宇
10 月 21 日的 Meta 财报会上,扎克伯格披露了一组关于个人智能体 Muse 的关键数据:超过 1000 万人设置了 Muse,购物是 Muse 最常见的用途,超过四分之一的用户在用它买东西,平均客单价超过200 美元;上线第一周,Muse 交易量便超出了预期,Meta 不得不与支付合作伙伴 Stripe 一起临时上调了信用额度。
六天后的 10 月 27 日,蚂蚁集团发布了灵光 2.0,将产品定位从最初的「30 秒用自然语言生成小应用」,彻底转向「能管事的个人 AI 助理」,官方口号概括为九个字:「记得住、盯得牢、办得了」。它打通了支付宝账单、钉钉、飞书、米家等服务,让 AI 替你管日常事务,试图成为中国版 Muse。
但 Muse 和灵光面对的问题截然不同。美国电子商务分散在大量独立网站中,Muse 可以通过云端浏览器直接访问店铺;而中国的日常消费与沟通高度集中在少数几个超级 App 内部,数据与交易链路彼此独立。
#Ant Group#Ling Guang 2.0#Alipay#DingTalk#Feishu#Mi Home#Qwen#Muse#Claude Duet#Taobao