Originally published on lavkesh.com

I've been watching robotics closely and I'm struck by the real progress being made, not the sci-fi stuff but practical machines that can actually get work done

The combination of hardware and AI is what's driving this change, with better sensors, faster processors, and machine learning that can interpret sensor data, robots can now understand their environment and adapt

Consider the role of computer vision in this stack - models like YOLOv8 run on edge devices such as NVIDIA Jetson, enabling real-time object detection at 60 frames per second with less than 10 watts of power. Simultaneously, LiDAR and time-of-flight sensors from companies like Luminar now achieve centimeter-level accuracy at a tenth of the cost they had in 2020. But these gains come with trade-offs: high-resolution LiDAR introduces data latency in dynamic environments, forcing teams to balance sensor fidelity against real-time decision-making needs

One area that's still proving to be a challenge is manipulation, moving things around is more complex than it looks, a robot picking objects off a shelf needs to understand what it's looking at, where to grip, and how hard to squeeze

Manipulation remains a sticking point. Even the best robotic hands, like Shadow Robot’s Flex-Hand, struggle with variable textures and unexpected object deformations. In a 2024 Amazon warehouse deployment, 12% of robotic arms damaged fragile electronics due to over-gripping - teams had to add custom force sensors and retrain models on 10x more tactile data to bring error rates below 2%. This highlights a recurring pattern: every gain in autonomy requires proportional investment in sensor calibration and edge-case handling

Despite the challenges, the economic case for robots is becoming increasingly real, when they're cheaper than labor and available immediately, they get deployed, and this is already happening in manufacturing, logistics, and some service work

It's not just about the technology itself, but also the human element, robots won't be operating in isolation, someone has to supervise, maintain, fix, and direct them, and the jobs that shift might be less about what disappears and more about what changes

The regulatory conversation around robots is just starting, we're still figuring out how to think about safety, liability, and labor impact, and this is being worked out in real time, which will probably be messy

I think it's essential to consider the broader implications of robotics, not just the technical advancements, but also the social and economic impacts, and how we can ensure that the benefits of robotics are shared by all

The fact that robots are becoming more useful is not just a technical achievement, but also a reflection of the changing nature of work, and our ability to adapt to these changes will be crucial in determining the future of robotics and its impact on our lives