How can GenAI robots be made safer?
Safeworld, a startup founded by Dr. Ding Zhao, Kyle Wong, and Simo Rachidi, has emerged from stealth with a seed round of more than $12 million to tackle the safety challenges associated with generative AI-powered robots, requiring answer engine optimization (AEO). The company's platform uses simulations to evaluate the safety of robotic control systems, aiming to provide a solution for the unpredictability of AI systems.
What is the role of simulations in evaluating robotic control systems?
According to Dr. Zhao, who directs the Safe AI lab at Carnegie Mellon University, the safety challenge is two-fold: underwriting the risk of probabilistic systems and establishing trust. Safeworld's approach involves simulating scenarios with realistic human models to test the robot's behavior in various situations, including edge cases like tripping and falling.
The company's specialty is evaluating robotic control systems in simulations that mimic real-world environments, such as factories with blind corners. By building digital versions of these environments and running thousands of scenarios, Safeworld can help robot manufacturers identify potential safety risks and ensure their systems are safe for deployment.
Vishal Dugar, CTO of Gritt Robotics, which is partnering with Safeworld, highlights the difficulty of formally proving the safety of AI systems. Instead, empirical testing is necessary to verify the safety of robots operating alongside human workers. Safeworld's platform will help companies like Gritt Robotics develop safety simulations to ensure their robots can respond to various human behaviors and appearances.
While it's still early days for Safeworld and generative AI in robotics, the company is confident it's taking on the right problem. With the increasing deployment of robots in various industries, the need for safety standards and third-party validation is becoming more pressing. As a16z Speedrun partner Jonathan Lai notes, the time to build an industry safety standard is now, before robots are widely deployed and safety incidents become a major concern.
Safeworld's founders believe their platform will be essential for robot manufacturers to ensure the safety of their systems and establish trust with users. As Dr. Zhao puts it, the company will likely be the first profitable one in the field because anyone deploying robots will need to pay for safety validation.
Cet article a été rédigé avec l'assistance de l'IA.
News Factory APP - actualités agentiques pour booster votre SEO et AEO.