A physical AI spinout from Imperial College London has secured record-breaking seed funding to develop a new kind of robot foundation model building on the founders\u2019 research at Imperial.<\/p>\n
Machenta has raised $40 million in the largest ever seed raise by a UK robotics startup. The funding round led by Creandum and Union Square Ventures will help the company scale an AI model that enables robots to learn new tasks instantly from a single demonstration.<\/p>\n
“Machenta\u2019s early success signals an opportunity for the UK to become a global leader in AI-enabled robotics, drawing on the specialist capabilities of universities like Imperial \u2013 and you can expect to see a growing stream of game-changing technologies from Imperial and our collaborators in this space.” Professor Mary Ryan Vice Provost (Research and Enterprise) at Imperial College London<\/p>\n
Today\u2019s AI-powered robots typically need to be trained through humans demonstrating individual tasks many times, often taking days. This yields robots capable of certain tasks but which cannot gain new skills without laborious model updates by developers.<\/p>\n
Using a breakthrough AI model developed by Machenta, robots can now learn new tasks from just a single demonstration. The company\u2019s technology could eventually give robots a human-like ability to learn new tasks on-the-fly with guidance only from their end users.<\/p>\n
\u201cIntelligence is not so much about what you can already do, it\u2019s about how quickly you can learn to do something new,\u201d said Professor Edward Johns<\/a>, Co-Founder and CEO of Machenta, and Professor of Robot Learning in the Department of Computing<\/a> at Imperial College London.<\/p>\n \u201cHumans don\u2019t need to see the same thing over and over before they can do it. If you take some objects, and show somebody what to do with them, they\u2019ll probably then be able to do the same task immediately. We\u2019ve spent years trying to understand what it would take for robots to learn at the speed of a human. That is what instant robot learning is about.\u201d<\/p>\n Machenta\u2019s approach is also designed to make the pre-training process much more scalable. While the AI models that power robots at present are often pre-trained with huge datasets taken from the real world, particularly from human demonstrations, Machenta\u2019s model is pre-trained entirely on simulated data. This means that future developers may no longer face the practical and privacy constraints of acquiring large volumes of human data.<\/p>\n