In the world of robotics, a cutting-edge startup is making waves with its groundbreaking creation: a “general purpose brain” for AI. Skild AI, founded by former Carnegie Mellon University professors Abhinav Gupta and Deepak Pathak in 2023, has developed a single off-the-shelf intelligence that can be integrated into various robots, enabling them to perform basic functions like climbing steep slopes, maneuvering around obstacles, and recognizing and picking up objects.
The company recently announced that it has raised an impressive $300 million in a Series A funding round led by Lightspeed Ventures, Softbank, Coatue, and Jeff Bezos, the founder of Amazon. The funding will propel Skild AI’s mission to revolutionize the robotics industry and unlock the potential of robots across various sectors.
One key feature of Skild AI’s technology is the emergence of unanticipated capabilities in the robots it powers. According to Raviraj Jain, a partner at Lightspeed Ventures, the robots using Skild’s AI models demonstrated entirely new abilities that they were not explicitly taught. These emergent capabilities, though simple, such as recovering dropped objects or rotating items, showcase the model’s adaptability and potential for advanced artificial systems.
Skild AI’s success can be attributed to its vast database of training data, which is claimed to be 1000 times larger than that of its rivals. Co-founder Deepak Pathak explains that the company blended a mix of data collection techniques, including human contractors operating robots remotely and collecting data, robots carrying out random tasks and learning through trial and error, and training the AI model on millions of public videos. These diverse sources of data, combined with the co-founders' expertise in artificial curiosity and language models, have contributed to the creation of Skild AI’s extraordinary “general purpose brain.”
While Skild AI faces competition from other well-funded robotics companies like OpenAI, Figure AI, and Covariant, the company differentiates itself through its access to large amounts of data. However, Ken Goldberg, a professor of robotics and automation at UC Berkeley, highlights the challenges of scaling robotics due to the scarcity of specific data required for robot training. Despite this, Skild AI aims to overcome the obstacle by combining its data collection techniques with additional information from simulations.
Beyond its current advancements, Skild AI envisions a future where its foundational model can be fine-tuned to create different use cases and products, much like OpenAI’s approach to language models. Co-founder Abhinav Gupta envisions achieving artificial general intelligence for robots, allowing them to surpass human capabilities while still being able to interact with people in the physical world.
Stephanie Zhan, a partner at Sequoia Capital and an investor in Skild AI, believes that the robotics industry is on the verge of a monumental shift. She compares it to the “GPT-3 moment,” referring to the significant advancements seen in the world of digital intelligence. As Skild AI continues to innovate and refine its technology, the possibilities for robots in the physical world are becoming increasingly exciting.
With the backing of visionary investors and a talented team of founders, Skild AI is poised to reshape the robotics industry and pave the way for a future where robots possess a general intelligence comparable to human capabilities. As this cutting-edge technology continues to evolve, we may witness a new era in which robots become integral parts of our daily lives, working alongside us to accomplish complex tasks and ushering in a new age of human-robot collaboration.
Use the share button below if you liked it.