Mira Murati Wants Her AI to ‘Keep Humans in the Loop’

Mira Murati Wants Her AI to ‘Keep Humans in the Loop’

Mira Murati, former CTO of OpenAI, has unveiled a preview of Thinking Machines Lab’s new “interaction models” that natively understand continuous human communication through camera and microphone inputs. Unlike typical voice interfaces that transcribe speech into text for a language model, these models process pauses, interruptions, and tone changes in real time, adapting on the fly. The company has raised billions but released only one product so far—Tinker, an API for fine-tuning open-source models. The interaction models have not yet been publicly released.

This launch positions Thinking Machines Lab against the dominant industry trajectory where frontier labs like OpenAI, Anthropic, and Google build models that automate entire workflows from text prompts with minimal human involvement. Murati’s alternative vision—superintelligence that collaborates with rather than replaces humans—resonates with a recurring pattern in the substrate: the human-in-the-loop counter-movement. It also updates the open debate about whether AI’s optimal path is full automation or augmented collaboration, with Thinking Machines staking out a clear position on the collaboration side.

The significance lies in its potential to reorient product design priorities across the foundation-model segment. If interaction models prove effective, they could pressure other labs to invest in native multimodal interfaces that preserve human agency. This is not yet a market inflection, but it introduces a high-signal alternative to the prevailing autonomy-first approach, especially relevant for enterprise and creative applications where human judgment remains critical.

#ThinkingMachinesLab #MiraMurati #HumanInTheLoop #InteractionModels #FoundationModels #AICollaboration