來源Digital Trends•較早收集於 3m
Google Messages 測試 AI 驅動的「輕觸草稿」回覆功能
💡了解 Google 如何將生成式 AI 直接嵌入行動通訊中，以提升用戶生產力。
⚡ 30 秒速覽
有什麼變化
引入「輕觸草稿」功能以進行自動化文字生成
為什麼重要
此功能標誌著 Google 推動將生成式 AI 整合至日常通訊工具的策略，可能提升用戶參與度及對 AI 輔助寫作的依賴。
下一步行動
密切關注 Google AI Studio 的更新，查看是否有類似的生成式通訊 API 可供第三方整合使用。
誰應關注：Developers & AI Engineers
關鍵要點
- •引入「輕觸草稿」功能以進行自動化文字生成
- •專注於生成更長且具備上下文關聯的訊息回覆
- •將生成式 AI 直接整合進手機通訊體驗中
🧠 深度解析
背景與延伸：來自公開資料，非原文內容。引用 11 個來源。
🔑 增強重點摘要
- •The 'tap to draft' feature is an evolution of Google Messages' existing 'Smart Replies' and 'Tap to Send' functionalities, specifically designed to allow users to edit AI-generated responses before they are sent, addressing previous issues of accidental dispatches.
- •This new functionality places the AI-suggested reply directly into the message compose field, giving users a crucial intermediate step to review, modify, or confirm the text before sending it.
- •The message suggestions within 'tap to draft' are powered by Google's Gemini Nano, an on-device AI model that processes the 20 previous messages for context without sending them to Google's servers, enhancing user privacy.
- •Beyond 'tap to draft' for quick replies, the broader Gemini integration in Google Messages allows users to directly chat with the AI assistant for tasks like drafting longer messages, brainstorming ideas, planning events, and even generating images within a separate chat interface.
- •The 'tap to draft' feature is opt-in, meaning users must manually enable it in settings, as the default behavior remains 'Tap to Send' for instant replies.
🛠️ 技術深入
- The 'tap to draft' message suggestions are powered by Google's generative AI technology, specifically the Gemini Nano model.
- For generating message suggestions, the AI model utilizes the context from the 20 previous messages in the conversation.
- When rewriting a drafted message in different styles, the model uses only the single draft message for context.
- Crucially, for message suggestions, the data processing occurs entirely on the user's phone, meaning no messages are sent to Google's servers, ensuring on-device privacy.
- To use Gemini within Google Messages, devices must have 6GB RAM or higher, a personal Google Account (not Family Link or Workspace), the user must be 18 or older, RCS chats must be enabled, and the latest version of the Google Messages app is required.
- Direct chats with the Gemini assistant in Google Messages are not end-to-end encrypted; these interactions are sent to the cloud for processing, may be used for training, and can be reviewed by humans (anonymized), with data stored for 18 months.
🔮 前景展望基於引用來源的 AI 分析
Enhanced user privacy for quick replies will become a standard expectation.
The on-device processing of 'tap to draft' suggestions using Gemini Nano, which keeps message content off Google's servers, sets a new benchmark for privacy in AI-powered messaging features, likely influencing future user expectations and competitor offerings.
AI integration will deepen within core communication platforms, moving beyond simple suggestions.
The broader Gemini integration in Google Messages, offering capabilities like brainstorming, event planning, and image generation alongside drafting, indicates a trend towards more comprehensive AI assistants embedded directly into messaging apps.
On-device AI models will become more prevalent and sophisticated in mobile applications.
The reliance on Gemini Nano for on-device processing of 'tap to draft' suggestions highlights a growing capability and trend towards powerful AI models running locally on mobile devices, reducing latency and cloud dependency.
⏳ 時間線
2021-10
Google introduces Pathways, a new AI architecture designed for training single models on many tasks efficiently.
2024-01
Reports surface about Google testing AI (initially Bard, later Gemini) integration into Google Messages, raising early privacy discussions.
2024-02
Google officially announces the integration of its Gemini AI model into the Messages app, rolling out in beta for drafting, planning, and brainstorming.
2024-03
Further details emerge about Gemini's capabilities in Google Messages, including potential for image generation and integration with other Google apps, with beta rollout to specific devices.
2026-03
Google Messages begins beta testing 'Tap to Draft' (initially 'Tap to Edit'), allowing users to edit Smart Replies before sending, addressing previous 'Tap to Send' issues.
2026-06
'Tap to Draft' starts a wider public rollout in the stable version of Google Messages as an opt-in feature.
📎 來源 (11)
Factual claims are grounded in the sources below. Forward-looking analysis is AI-generated interpretation.
📰
AI 週報
閱讀本週精選 AI 大事摘要 →
👉相關動態
AI 策展新聞聚合。所有內容版權歸原始發布者所有。
原始來源： Digital Trends ↗
每週電子報
每週一封，可隨時退訂。