Ex-Meta Scientists Unveil Perceptron Visual AI to Guide Factory Robots
Quick Read
- Startup Perceptron, founded by two former Meta AI researchers, has released a new model called Isaac 0.5
- The model designed to help robots “perceive, reason and act” in real-world settings like warehouses and factories
- Isaac 0.5 is open-weight, meaning its parameters are publicly inspectable
- Perceptron recently raised $21 million in a round led by Bessemer Venture Partners. The company plans to sell into manufacturing, logistics, security, and media industries
Two scientists who used to work at Meta have started a company that’s trying to bring smart, “seeing” AI out of your phone and onto the factory floor. Their startup, Perceptron, just launched a new Perceptron visual AI model called Isaac 0.5. The model built to help robots understand the world around them like a warehouse or a busy factory and figure out what to do next.
Founded in November 2024 by ex-Meta FAIR scientists Armen Aghajanyan and Akshat Shrivastava, Perceptron says most existing robotics software forces a tradeoff: bulky, generalist models that need heavy cloud computing, or narrow tools that handle only perception or only control never both. Isaac 0.5 aims to do both at once, letting robots read labels, judge spatial layouts, and plan multi-step actions like sorting packages, all in one flexible system.
The model trained on roughly a million hours of video, blending general footage with “ego video” recordings from a human’s point of view and “UMI video,” which captures repetitive human motions robots can learn to copy. Perceptron hasn’t disclosed exact data sources but says it built petabyte-scale datasets spanning images, text, video, and robotic movement.
Isaac 0.5 Released
Notably, Isaac 0.5 is being released as an open-weight model, so outside researchers and companies can inspect exactly how it works a move that could accelerate adoption among developers building automation tools. Perceptron is targeting manufacturing, logistics, warehousing, security, mobility, and media companies as potential customers.
The launch comes on the heels of a $21 million funding round led by Bessemer Venture Partners, giving the young startup fresh capital to compete in the fast-growing “physical AI” space, where firms are racing to move AI capabilities out of chatbots and into machines that act in the real world.
