What if you could ask a robot to warm up your lunch and have it find the microwave by itself?
That was a question posed by a group of Microsoft researchers in a February paper exploring different ways OpenAI’s ChatGPT could create technical instructions for robotic tasks through simple prompts.
A new Seattle startup called Scaled Foundations aims to expand on that research through an AI-powered development platform that makes it easier for non-technical users to create commands for robots.
Launched earlier this year, the Seattle company emerged from stealth mode Wednesday and announced an undisclosed investment from Khosla Ventures and E14 Fund.
The company is led by Ashish Kapoor, who spent 17 years at Microsoft, most recently as general manager of autonomous systems and robotics research. In the position, he created AirSim, an open-source simulator for drones.
Kapoor also co-authored the paper “ChatGPT for Robotics.” In the report, the researchers wrote that robotics systems, unlike text-only applications, demand a grasp of real-world physics, environmental context, and the ability to perform physical actions.
Scaled Foundations, which operates independently from Microsoft, is trying to accomplish that goal with its flagship platform called General Robotic Intelligence Development (GRID). The product includes an AI foundation model for robotics, synthetic data generation, and a tool for orchestration, training, and fine tuning.
For example, customers can use the startup’s platform to program a drone for an inspection mission. GRID can help plan the best flight path, analyze data with an inspection model, and ensure the drone is safe by avoiding obstacles and finding safe landing spots.
Scaled Foundations has a similar approach to Seattle startup Groundlight, which taps into a company’s network of cameras to gather real-time data using computer vision and AI. Groundlight’s platform lets users ask questions in natural language — such as, “is there a truck in the loading dock?” — and then get results instantly. It works in a variety of applications including video stream analysis, industrial automation, process monitoring, retail analytics, and robotics. The startup raised $10 million in April.
Scaled Foundations is initially focusing on aerial robotics and vehicles, with a plan to eventually roll out tools for manipulation and wheeled robots. It declined to disclose customer data, but said that it works with industries across renewable energy, oil and gas, wildfire detection, and urban air mobility vehicles.
Dinesh Narayanan, the company’s co-founder and head of commercialization, said there are currently no regulatory barriers unique to infusing AI into robotics. “However, AI capabilities will ensure safer operations for robots and commercial vehicles, which we believe will accelerate regulatory support for these critical categories,” he said.
Kapoor brought on former Microsoft colleagues Sai Vemprala and Shuhang Chen; both were part of the Redmond, Wash.-based tech giant’s Autonomous Systems Research team. Vemprala and Chen, who are co-founders, lead technical strategy for Scaled Foundations. They are joined by Narayanan, who previously held international business development and strategy roles at Microsoft. The company has five employees.