微软企业 AI 的 Lewis Liu 将 AI 智能体记忆拆成三种目标：个性化让智能体适配特定用户，自我改进让过往经验改变任务执行方式，组织记忆让公司积累的知识触达需要的人。他指出“记忆”背后隐藏着关键设计决策：什么能跨会话保留、何时重新浮现、如何影响下一次决策。Turing Post 据此发布了一份智能体记忆完整指南，从 7 个视角探讨如何把经历转化为更好的工作成果。
What is AI agent memory?
Memory lets an agent use information from earlier work in a later decision - a preference, a project fact, a working procedure, etc.
But "memory" also hides the design decisions: what survives a session, when it resurfaces, and how it shapes the next decision.
Lewis Liu, who works on enterprise AI at @Microsoft, separates 3 ambitions - each asks the system to carry something different into the future:
- Personalization - helps an agent adapt to a particular person
- Self-improvement - is responsible for experience that can change how the agent performs a task
- Organizational memory - can a company’s accumulated knowledge reach whoever needs it?
Once we make these ambitions explicit, we can ask a more useful question: what should be different the next time the agent does its work. How do we turn what happened into something an agent can learn from?
We just published a full guide on agent memory explaining 7 perspectives on turning experience into better work.
Worth a read to explore the harder questions behind agent memory → https://www.turingpost.com/p/agent-memory-experience-organizational-belief
来源：Turing Post (Ksenia Se) · x.com