颜水成团队发布 VoiceMem，提出流式双脑架构实现实时交互记忆
VoiceMem 用流式双脑架构给实时语音助手装上记忆，解决多轮交互中模型不认人的痛点，是迈向具身智能的关键一步。
实时语音交互系统往往没有记忆，每次对话都从零开始，难以建立信任。VoiceMem 是一种类脑记忆系统，将检索延迟压缩到 134 ms，并支持语音、说话人、声音事件等多模态信息，让 AI 助手真正记住你。
正文摘录
- title: 颜水成团队重磅发布 VoiceMem,提出「流式双脑架构」打造下一代实时交互基础设施 - sourcecompany: 机器之心 - bodymarkdown:    Science fiction foreshadowed this day long ago, and the real progress in AI over the past two years is pushing it toward a discussable proposition: AI may not stop at being a "tool" but could become an entity coexisting with humans —…