Skilled AI is a Pittsburgh-based robotics startup founded by former Carnegie Mellon University professors that is developing a general purpose brain for robots. This brain, trained on a massive database of text, images, and videos, enables robots to learn entirely new skills and perform tasks like climbing slopes and picking up items without explicit programming.
A Pittsburgh-based robotics startup called Skilled AI, founded by former Carnegie Mellon University professors Abhinav Gupta and Deepak Pathak, is developing a general purpose brain for robots. This brain can be integrated into various types of robots, such as bipedal humanoids and military robot dogs, allowing them to perform tasks like climbing steep slopes, walking over obstacles, and picking up items. Skilled AI has raised $300 million in a Series A funding round, with investors including Lightspeed Ventures, SoftBank, and Jeff Bezos.
The company’s general purpose brain is based on a foundational model that has been trained on a massive database of text, images, and videos, which they claim is 1,000 times larger than those used by their competitors. To build this database, Skilled AI employed various data collection techniques, such as hiring human contractors to operate robots remotely and collect data, as well as having the robots carry out random tasks and learn through trial and error. The AI model was also trained on millions of public videos to enhance its capabilities.
One of the key innovations by Skilled AI is the incorporation of artificial curiosity into robots, rewarding the system for producing outcomes that are unpredictable. This approach incentivizes the AI to explore more scenarios and collect more data, leading to the development of emergent capabilities in the robots. Additionally, the company has integrated techniques from large language models like GPT to enable robots to convert written instructions into actions, enhancing their ability to learn from different sources.
Skilled AI faces competition from other robotics companies that have also secured substantial funding in the AI industry. Industry giants like OpenAI have entered the robotics space to provide models to other companies, while startups like Figure AI and Covariant are also making advancements in humanoid robotics and building chat GPT for robots, respectively. Skilled AI’s unique approach to training robots using a combination of large datasets, artificial curiosity, and learning from various sources sets them apart in the competitive landscape of robotics and AI technology.
The success of Skilled AI’s general purpose brain lies in its ability to enable robots to perform complex tasks like climbing stairs and adapting to new environments without explicit programming. By leveraging a diverse range of data sources and innovative training techniques, Skilled AI aims to revolutionize the field of robotics by creating adaptable and versatile robots capable of learning entirely new skills and capabilities on their own.