Perceptron debuts Isaac 0.5 enabling robots to perceive, reason, act
Perceptron, a startup by former Meta researchers Armen Aghajanyan and Akshat Shrivastava, just dropped Isaac 0.5, a new AI model built to help robots handle all sorts of tasks in complex environments like warehouses or factory floors.
Unlike older AIs that stick to one job, Isaac 0.5 can actually "perceive, reason and act" in different situations, making it way more adaptable.
Model trained on a million hours
Isaac 0.5 learned from a massive set of videos (a million hours of what is known as general video) to teach robots things like spatial awareness and decision-making.
Perceptron raised $21 million to roll this out across industries like manufacturing and security, and they're even making the model open for anyone to check out.
Co-founder Aghajanyan called it, "Nothing like this really exists out there," and "We're really excited about it."