Governing Seamless Human-AI Interactions
As Thinking Machines' new interaction model shows, enhanced human-AI interaction will pose novel risks and require additional safety and governance layers to ensure humans remain protected
TL;DR
- Thinking Machines, founded by former OpenAI CTO Mira Murati, launched a research preview of a new interaction AI model.
- The model focuses on "human-like" interaction, moving away from traditional turn-based AI conversations.
- It employs a "multi-stream, micro-turn design" allowing simultaneous messaging, talking, listening, seeing, showing, and interjecting between humans and AI.
- Key capabilities include seamless dialog management, verbal/visual interjections, simultaneous speech, time-awareness, and concurrent tool calls/search/generative UI.
- This new paradigm is expected to be disruptive but also introduces new AI governance risks that require serious consideration.