Expert Explores Rise of 'Physical AI'
The next wave of AI will focus on interpreting data from the physical world, moving beyond text- and image-based models. Ivan Poupyrev, CEO of Archetype AI, explained on the DataFramed podcast that foundation models are now enabling IoT and sensor data to be converted into actionable insights. He notes these models rely on different data, with roughly 50% from sensor measurements, which will create new analytics opportunities in areas like smart retail and event marketing.
- Archetype AI's foundation model, named Newton, was built to interpret and fuse real-time data from a wide array of sensors—including accelerometers, gyroscopes, radars, cameras, and microphones—with natural language. - The company, co-founded by CEO Ivan Poupyrev, emerged from stealth with a $13 million seed funding round led by Venrock and with participation from the Amazon Industrial Innovation Fund and Hitachi Ventures. It has since raised a $35 million Series A round with investors including Bezos Expeditions and Samsung Ventures. - Before co-founding Archetype, Ivan Poupyrev led a team at Google's Advanced Technology and Projects (ATAP) division that developed Project Soli, which used a miniature radar chip for gesture recognition in devices. - In smart retail, physical AI uses cameras and sensors to analyze foot traffic heat maps, which helps in optimizing store layouts and product placement for better customer engagement. - Beyond retail, early enterprise users are already applying the technology; NTT DATA and Kajima have implemented "Physical Agents" to improve efficiency and safety in warehouses and on construction sites. - Amazon's "Just Walk Out" technology is a prime example of physical AI in action, using computer vision and sensors to track items shoppers take, eliminating traditional checkout lines. - The broader trend is toward a "trillion sensor economy," where intelligent sensor networks can make autonomous decisions and optimize processes in real-time without human intervention, impacting areas from industrial automation to smart cities.
