Robot Dog Reads Factory Gauges With 98% Accuracy Using AI

Robots just got a lot smarter at one of the trickiest tasks in industrial inspections: reading the dials, needles, and levels on old-school gauges and thermometers.

Boston Dynamics' four-legged Spot robot can now interpret complex instruments with 98% accuracy, up from just 23% with the previous AI model. Google DeepMind announced the upgrade on April 14 as part of its new Gemini Robotics-ER 1.6 model.

The leap forward matters because factories and warehouses still rely heavily on analog instruments like pressure gauges, temperature dials, and sight glasses that let workers peek inside tanks and pipes. Until now, robots struggled to make sense of multiple needles, tick marks, liquid levels, and text all at once.

The new AI gives robots what Google calls "agentic vision," combining visual reasoning with code execution to create a digital scratchpad for analyzing images. Think of it like a robot taking mental notes while examining something complex.

Boston Dynamics has been testing Spot as a roving inspector in industrial facilities, including factories owned by its parent company, Hyundai Motor Group. The robot walks through plants checking equipment that would normally require human workers to make regular rounds.

The Ripple Effect

The improvements go beyond just reading dials. The new model dramatically reduces AI hallucinations, those moments when artificial intelligence sees things that aren't there.

In one test, the older model falsely identified a wheelbarrow in a cluttered toolbox because that was one of the items it was asked to find. The new version accurately counted hammers, scissors, paintbrushes, pliers, and gardening tools without inventing phantom objects.

Google also built stronger safety features into the model. The AI can now better assess risks to humans, like recognizing when a child might stick something into an electrical socket, and it makes safer decisions when robots handle liquids or hazardous materials.

The technology uses multiple camera angles to understand environments more completely, similar to how humans turn their heads to get a better view. This multi-view reasoning helps robots navigate complex spaces without missing important details.

For decades, robots have excelled at repetitive factory tasks, doing the same motions over and over with precision. This breakthrough nudges them toward becoming flexible workers who can handle unpredictable environments and varied assignments, all while keeping people safe.