Jack Dorsey 旗下 Block 发布语音模型 Dessa,对话像笨拙的新朋友
摘要
Jack Dorsey 创立的 Block 发布了一款名为 Dessa 的语音 AI 模型,它完全以人类对话数据从零训练,官方通过公开真实聊天记录来展示其接近自然但偶有尴尬的对话能力。
核心要点
- Block 称 Dessa 是“从零开始、只用人类对话”训练出来的模型,而非在文本大模型基础上改造
- 官方对 Dessa 的定位是“首个能与真人进行真实对话、且完全用语音训练”的模型,属于全双工语音到语音(full-duplex speech-to-speech)架构
- 为展示效果,Block 公开了模型与人类的真实聊天记录,并承认对话偶尔会显得尴尬、诡异(uncanny)
- 由 Dessa 项目负责人 Ragavan Thurairatnam 分享的一段对话中,模型起初把说话人误认为一个叫“Ronaldo”的人
- 尽管存在误认,文章指出该段对话整体仍保持了较为自然的来回交流节奏
原文佐证
- Jack Dorsey's Block announced an AI model called Dessa that's trained "from scratch on nothing but human conversation," and to show off the model, Block shared chats it has had with humans that can occasionally be a little awkward and uncanny.
- The conversation between the model and the speaker does have a mostly natural back and forth (though for some reason, the model initially mistakes the speaker for somebody named "Ronaldo").
- We made the first AI model you can have a realistic conversation with that was trained entirely on speech! Dessa is a full-duplex speech-to-sp …
AI 洞察
Block 选择“只用人类对话、从零训练”而非拼接文本大模型加语音接口,是在押注语音交互的原生范式,全双工能力恰恰是当前语音助手最难自然实现的部分。主动晒出“笨拙”“认错人”的对话,是一种降低预期、以真实感换取可信度的策略,也暗示其现阶段的短板仍在语义理解与上下文一致性。若这条路走通,语音交互的评价标准可能从“识别准确率”转向“对话节奏与打断处理”,对现有语音助手与实时翻译类产品构成差异化压力。