Band Wagon

“Band Wagon” – A cartoon that captures the industry’s sudden belief that robotics will save performance-driven AI.

At the end of last year, the mood around AI shifted. While AI still remains praised for personal experimentation and content generation, serious problems started to be reported when companies tried to run real business systems on it. For example, Salesforce executives publicly acknowledged they scaled back the use of large language models in core workflows after trust issues emerged. When forecasts or operational decisions are off by even 10%-15%, the promised efficiency of AI turns into operational risk.

So attention has now swung to robotics. If AI struggles in abstract software systems, the thinking is it will deliver clearer value when it touches the physical world. Robots promise visible outcomes, such as more items picked per hour, fewer errors andfewer injuries. For example, Amazon’s warehouse robots reduced order cycle times by roughly 20%, and autonomous guided vehicles routinely remove 40% of repetitive labor from logistics operations. That kind of ROI is understandable to CFOs, investors or customers.

Also, the upswing in robotics interest came from the convergence of better perception models, simulation-first training and cheaper edge compute. At CES in Las Vegas this year, robotics dominated the floor, from warehouse automation and industrial arms to humanoid demos. NVIDIA’s announcements around autonomous driving and robotics were key highlights of the show, reinforcing the idea that cars, robots and “embodied AI” now share the same technical backbone.

However, the problem is that broader use of robotics will make people realize that hardware doesn’t forgive like software. If the same weaknesses seen in enterprise AI carry over to robotics, we will realize how more expensive mistakes are when the issues are physical. In software, hallucination turns into mis-execution … in the physical world, edge cases turn into safety incidents.