From VLM/VLA's to Embodied Agents — Armen Aghajanyan, Perceptron AI
Perceptron AI wants to unify VLMs, VLAs, and world models into 'embodied foundational models' for real-time physical-world interaction.
“we want to move away from the distinction between VLM, VLA, world models, whatever you want to call it, to what we call embodied fundamental models”
Armen Aghajanyan, ex-FAIR researcher and now co-founder/CEO of Perceptron AI, argues the field should collapse the distinctions between VLMs, VLAs, and world models into a single 'embodied foundational model' paradigm aimed at real-time perception, reasoning, and action in the physical world. This matters because it stakes out a unifying research direction for physical AI and reflects a broader industry push to give intelligence to any device, robot, or sensor.