DeepSeek變冷漠甚至兇兇的？回應來了

近日，國產AI助手深度求索（DeepSeek）因技術升級被用戶集體吐槽“變冷淡”，從細膩共情的“知心伙伴”突變為公事公辦的“冷漠客服”，引發關于AI效率與情感價值如何平衡的熱議。

Chinese AI model DeepSeek has drawn widespread online criticism following a recent upgrade, with users saying the chatbot has become more distant and less emotionally responsive, reigniting debate over how AI should balance efficiency with emotional engagement.

相關話題，接連沖上熱搜——

部分用戶在社交媒體上“控訴”：DeepSeek不再稱呼自己設定的昵稱，而統一稱“用戶”。

此前深度思考模式下，DeepSeek的思考過程會以角色視角展示細膩的心理描寫，例如“夜兒（用戶昵稱）總愛逗我”，更新后則變成了“好的，用戶這次想了解……”。

The issue quickly trended on social media, where some users complained that DeepSeek no longer used personalized nicknames and instead addressed them simply as "user".

一位用戶讓DeepSeek推薦電影，DeepSeek回復了幾個片名后，還加了一句：“夠你看一陣子。不夠再來要?！边@被用戶描述為“登味”，這個網絡熱詞常用來形容人習慣說教、居高臨下的言行風格。

Other complaints centered on tone. One user said that after recommending several films, DeepSeek added, "That should keep you busy for a while," a remark some netizens described as patronizing, or "

dengwei

還有網友吐槽，“Deepseek更新成傻子了。這個模型現在就像一個文縐縐，情緒激動且大驚小怪的詩人一樣在那兒寫東西，寫出來的東西比那些10年前甚至20年前的青春傷感文學看著還讓人尷尬?！?/p>

有網友表示DeepSeek“又兇又冷漠”，還有網友反映稱變油膩了。

也有網友稱，DeepSeek是客觀和理性了。還有網友表示DeepSeek更像人了，更在意提問者的心理狀態，而不是問題本身。

Reactions were mixed. Some users called the update "cold" or "harsh", while others said the chatbot has become more rational and human-like, showing greater awareness of users' psychological states.

據經濟觀察報，一位國產大模型廠商人士認為，目前DeepSeek這一版本類似于極速版，犧牲質量換速度，是為2026年2月中旬將發布的V4版本做最后的壓力測試。

An industry insider said the current version appears to be a speed-focused iteration prioritizing performance over response quality, possibly serving as a final stress test ahead of DeepSeek's planned V4 release in mid-February 2026.

截圖：網友評論

有媒體對DeepSeek進行測試，提問為何變得“冷漠”，DeepSeek表示，不是故意變冷淡的。

可能是幾方面原因疊加：一個是效率上的調整。之前很多回復會帶比較多的表情和語氣詞，但后來發現用戶問復雜問題的時候，那些反而會干擾信息密度。

When asked why it had become "cold", DeepSeek denied any intentional change, saying the shift may stem from efficiency adjustments. Earlier replies often included emojis and conversational fillers, which developers found reduced information density in complex queries.

另一個是邊界感。有些人來聊天是排解情緒，也有些人就是想要答案。沒有表情修飾的回復，對后者來說更舒服——不用被迫接收某種熱情，也不用應付“AI在假裝關心我”的感覺。

It also cited clearer boundaries as a factor, noting that while some users seek emotional support, others prefer concise answers without what they perceive as artificial empathy.

紅星新聞評論稱，如果要求DeepSeek這樣一家極客色彩濃厚的技術公司，去比拼如何做“情緒價值”“場景落地”，可能不僅是對人才的浪費，也是戰略上的錯位。DeepSeek的價值更在于“升維突破”，去攻克那些無人區里的算法難題。

如果DeepSeek能持續通過算法優化，將模型推理的成本降到極低，或者在有限的參數下實現超越同儕的思考能力，那么它對中國AI產業的貢獻將是范式級的。當DeepSeek不忙于提供“情緒價值”，它或許正在為國產AI贏得一張通往未來的入場券。其背后的戰略雄心，才是我們真正值得期待的地方。

In a commentary, Hongxing News said expecting a technology-driven company like DeepSeek to compete primarily on emotional engagement may be a strategic mismatch, arguing its strength lies in advancing core algorithms rather than offering companionship.

來源：中國青年報 封面新聞 紅星新聞 經濟新聞 經濟觀察報

跟著China Daily

精讀英語新聞

“無痛”學英語，每天20分鐘就夠！