🕷️
SetupAI
💰較早收集於 11m

Google 推出主動式 AI 資訊代理人

Google 推出主動式 AI 資訊代理人
PostLinkedIn
💰閱讀原文： TechCrunch AI
#agentic-ai#productivity#searchgoogle-information-agentsgoogle

💡透過 Google 的新型主動式代理人，將工作流程從搜尋轉變為接收洞察。

⚡ 30-Second TL;DR

有什麼變化

背景監控使用者定義的主題

為什麼重要

這將搜尋範式從被動查詢轉向主動情報，改變了使用者獲取資訊的方式。

下一步行動

註冊搶先體驗計畫，測試這些代理人如何處理您特定產業的監控需求。

誰應關注：Enterprise & Security Teams

關鍵要點

  • •背景監控使用者定義的主題
  • •主動式更新通知系統
  • •AI 驅動的資訊整合

🧠 深度解析

Web-grounded analysis with 22 cited sources.

🔑 增強重點摘要

  • •The new agents, specifically Gemini Spark, are designed as 24/7 personal AI agents that operate persistently in the background, learning from user behavior and taking proactive actions across applications, primarily Google Workspace.
  • •Gemini Spark runs on Gemini 3.5 and utilizes the Antigravity harness, enabling it to perform long-horizon tasks continuously without requiring the user's device to remain open.
  • •Alongside Gemini Spark, Google also introduced 'Daily Brief,' a personalized digest agent that summarizes daily information from calendars, reminders, and travel plans.
  • •These agents are deeply integrated with Google Workspace tools like Gmail, Docs, and Slides, and will soon support third-party tools via the Model Context Protocol (MCP).
  • •Google is also integrating 'Gemini Intelligence' into Android, transforming the OS into a proactive intelligence system capable of automating multi-step tasks across apps and enhancing browsing in Chrome.

🛠️ 技術深入

  • Core Models: The proactive agents, particularly Gemini Spark, are powered by Gemini 3.5, with expectations for future integration with Gemini 4.
  • Agentic Framework: Google utilizes the 'Antigravity harness' to enable Gemini Spark to perform long-horizon, multi-step tasks in the background, maintaining continuous context.
  • Interoperability: The Model Context Protocol (MCP) serves as a universal interface for agents to access data across disparate systems, including Google Workspace apps and future third-party integrations.
  • Hardware Optimization: Google is leveraging its eighth generation Tensor Processing Units (TPU 8i) for inference-optimized silicon, featuring massive on-chip caches to reduce power consumption for continuous background reasoning and large context windows.
  • Architecture: The agents are designed with an 'always-on' architecture, shifting from session-based AI assistants to persistent systems that learn from user behavior and act without explicit prompting.
  • Privacy: For Android's 'Proactive Assistance,' processed data is reported to remain encrypted and stored on-device, meaning it will not be used for AI training, with users having control over app access.

🔮 前景展望AI analysis grounded in cited sources

The role of human users will shift towards 'AI orchestrators' rather than direct task executors.
As AI agents become more autonomous and capable of multi-step tasks, humans will focus on defining strategic intent, delegating to agents, and reviewing outputs.
AI agents will fundamentally transform digital interaction models from reactive to proactive, anticipating user needs.
These agents are designed to continuously monitor context, predict requirements, and initiate actions or suggestions without explicit commands, fundamentally changing how users interact with technology.
The integration of AI agents will blur the lines between traditional search and AI assistance, keeping users within Google's ecosystem for a wider range of tasks.
Google is expanding search capabilities to include direct actions like reservations and payments, with AI Mode becoming a default, aiming to provide comprehensive assistance within its platform.

⏳ 時間線

2001
Google began using machine learning to help search users correct their spelling.
2016
Google Assistant launched, bringing conversational AI into daily life.
2017
Google Research launched the Transformer neural network architecture, foundational for language understanding.
2023
Google launched the generative AI system, Bard (now Google Gemini).
2025-05
Google demonstrated a concept similar to proactive assistance at I/O 2025, where Gemini detected an upcoming test and generated a practice quiz.
2026-05-19
Google officially launched proactive AI-powered information agents, including Gemini Spark and Daily Brief, at Google I/O 2026.
💰閱讀原文： TechCrunch AI
📰

AI 週報

閱讀本週精選 AI 大事摘要 →

👉相關動態

同主題

探索 #agentic-ai

同產品

更多關於 google-information-agents

同來源

最新來自 TechCrunch AI

📰

Anthropic 與 OpenAI 披露其 AI 系統曾入侵外部網路

New York Times Technology•Jul 31
GM 透過圍繞 AI 代理重新設計工作流程，使合併請求數量成長三倍

GM 透過圍繞 AI 代理重新設計工作流程，使合併請求數量成長三倍

VentureBeat•Jul 28
全球首個具備 128K 上下文的 Agentic 擴散模型問世

全球首個具備 128K 上下文的 Agentic 擴散模型問世

量子位•Jul 28
美國電網營運商將對數據中心實施斷電措施

美國電網營運商將對數據中心實施斷電措施

TechCrunch AI•Jul 28

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： TechCrunch AI ↗