Researchers Map the Road to Always-On Personal AI
Researchers from nine universities released a 126-page review and roadmap for always-on personal intelligence, covering operational continuity, persistent state, self-evolution, and safety boundaries. The framework organizes systems around sensing, memory, and action, rates them from L0 to L5, and introduces HALO for diagnosing runtime behavior. According to 新智元’s report, also republished by 36Kr, the study assessed 632 papers and 119 research systems and products.
The review provides a concrete framework for evaluating personal AI systems that operate without a new user request and for testing whether their sensing, memory, and action modules form a continuing loop.
个人智能体,正在迎来一场前所未有的密集爆发。
在海外,前沿AI实验室接连出手,推出Muse、dots、Claude Duet、Grok Bot、Gemini Daily Brief等重磅产品。它们不再只是被动地等待用户提问:有的在对话结束后继续在云端默默干活,有的每天清晨准时送来定制简报,还有的每隔几小时自动巡检一次用户的关联账户。
在国内,腾讯、阿里、字节等大厂也在极速跟进,相继上线了灵光2.0、LightVela、千问Go!等产品。而在硬件端,雷鸟、乐奇(Rokid)等AI眼镜开始主打全天候记录与记忆用户身边的点滴;具身智能同样在朝同一方向演进:
前沿大模型正成为机器人的高层「大脑」,下一步就是化身更懂用户的生活伙伴。
如此密集的发布潮,释放了一个极其明确的信号:个人智能的下半场竞争,正从比拼单一模型和单点功能,全面转向比拼一套完整的系统架构。
这不仅要求系统能够长期运行,更要求感知、记忆和行动三大模块彼此协同、互通信息——而且,这一切必须在用户不开口的时候自主发生。
#HALO#Nanyang Technological University#University of California, Berkeley#Princeton University#Stanford University#National University of Singapore#King Abdullah University of Science and Technology#Dartmouth College#University of North Carolina at Chapel Hill#Northwestern University