Key facts
- Safeworld is developing a platform to evaluate the safety of generative AI-driven robots.
- The company aims to create industry safety standards for robots operating in unpredictable environments.
- Safeworld's approach involves simulating robots interacting with realistic human models in digital environments.
- The seed funding round exceeded $12 million, led by Shine Capital and a16z Speedrun.
- Gritt Robotics, which develops AI brains for robots in solar farms, is a partner.
Safeworld, a new company founded by Carnegie Mellon University's Dr. Ding Zhao, veteran startup executive Kyle Wong, and machine learning engineer Simo Rachidi, is emerging from stealth with over $12 million in seed funding. The company aims to address the critical safety challenges posed by generative AI in robotics. Unlike traditional algorithms, generative AI models are probabilistic and less predictable, raising concerns about the safety of humanoid robots and other autonomous systems.
Dr. Zhao explained that Safeworld's approach combines advanced generative AI probabilistic evaluations to underwrite risk with building trust in these systems. The company's specialty lies in evaluating robotic control systems through simulations populated with realistic human models. This is crucial because robots operate in unstructured environments and must adapt to unpredictable human behavior, such as tripping, falling, or carrying objects, which can be difficult to account for with traditional methods.
Safeworld's platform will create digital replicas of environments, such as a factory corner, and run thousands of scenarios with simulated humans interacting with the robot being evaluated. This empirical approach is necessary because formally proving the safety of such complex systems through mathematical equations is challenging, according to Vishal Dugar, CTO of Gritt Robotics, a company partnering with Safeworld. Gritt Robotics develops AI for robots used in solar farm installations and aims for more complex construction tasks, requiring its robotic arms to operate safely alongside humans.
Investors, including Shine Capital and a16z Speedrun, believe that now is the opportune time to establish industry safety standards for robots, before widespread deployment leads to incidents. Safeworld's founders are confident they are tackling a significant problem, with Dr. Zhao suggesting they could be the first profitable company in this field due to the essential need for third-party validation of robot safety.