RCreddit.com
16
·开发者社区 · RSS
AI Often gets Very basic facts wrong. The models hone in on Averages - Not precise answers
Claude 相关模型动态已经出现，适合跟踪能力变化、生态影响和后续可用性。
一位用户在欧洲旅行时，尝试使用 Gemini 和 Claude 查询公交时刻表，但这两个付费订阅的AI模型都给出了完全错误的答案。当用户追问原因时，AI承认它们会采取捷径，通过“插值”来提供近似答案。对于带着孩子和行李的旅行者来说，这种不准确性导致的时间误差（一个多小时）是完全无法接受的，用户因此质疑AI处理简单请求的能力。
We as a family are traveling in Europe and have asked both Gemini and Claude simple queries about bus schedules — and they Both were totally wrong. By the way we Pay for subscriptions to both.
When pushing them on why they couldn’t even read a simple schedule they both admitted they take short cuts and “interpolate” approximate answers.
This is not at all helpful when traveling with kids and lugging many heavy bags - to have the AI be an hour plus off.
If they can’t even help with simple requests …
主题标签ClaudeGemini模型发布订阅权益
原始关键词#averages#answers#precise#models#basic#facts