In September 2026, Dyson launched a product that seems somewhat uncharacteristic of the brand: the CameraJet™ Visual Imaging Toothbrush. This toothbrush features a built-in 1mm, 0.1-megapixel macro camera that scans 28 real-time images per second. Once the AI (Artificial Intelligence) interdental optical positioning algorithm identifies the interdental spaces, it triggers a conical water jet in as fast as 100 milliseconds.
For this toothbrush, Dyson deployed 661 engineers and six years of development, training the algorithm with over 470,000 oral images. Priced at CNY 3,899, it was first launched in China on September 1. Also released alongside it is the flagship R3 Nurovi robot vacuum and mop, which combines AI vision with Ultraviolet (UV) light and green light detection. It can identify nearly 200 common household stains and objects, autonomously planning back-and-forth cleaning based on the severity of the stains.
The motor delivers 30,000 Pa of suction power, a 65% increase over the previous generation, while the 12-point water release system can dispense 70°C hot water to dissolve stubborn grease. Jake Dyson, Chief Engineer at Dyson, stated: "We are moving from purely mechanical devices to smart products that can see, think, and respond."
01. Is AI Growing Eyes?
Over the past two years, the AI we have grown accustomed to is highly articulate, capable of chatting, writing, translating, and drawing. However, in the autumn of 2026, a batch of new products has sent out a unified signal: AI is moving from the cloud down to the physical world and starting to see. Dyson's toothbrush and robot vacuum and mop are just one example of this trend.
In Berlin, Shanghai, and Shenzhen, almost simultaneously, multiple companies have released terminal products equipped with vision systems. This trend exists because voice interaction is no longer sufficient; machines need to see to make more precise judgments. A toothbrush needs to know where to direct the water jet, and a robot vacuum needs to determine whether a stain is soy sauce or cat food. They rely not on guessing, but on seeing.
Why now? It is because three conditions have matured simultaneously.
First, sensors have become cheaper. Specifically, the costs of 1mm-class macro cameras, multispectral sensors, and UV detection modules have dropped to the point where they can be integrated into consumer-grade products. The camera in Dyson's toothbrush has only 0.1 megapixels, but paired with AI algorithms, it is sufficient to identify interdental positions.
Second, edge computing power is now sufficient. In the past, visual AI had to be uploaded to the cloud for processing, resulting in high latency and significant privacy risks. Now, the Neural Processing Unit (NPU) computing power in smartphone chips can run visual models locally. Small devices like toothbrushes and robot vacuums are also equipped with dedicated AI chips, enabling them to complete image recognition and decision-making in milliseconds.
Third, large models have made visual comprehension easier. Previously, training a model to recognize interdental spaces or soy sauce stains required a massive amount of labeled data. Now, with the help of pre-trained large visual models, a small amount of data can be used to fine-tune a usable specialized model. Dyson trained its algorithm with 470,000 oral images, a scale that was unimaginable five years ago.
02. What Does This Mean for Ordinary People?
For ordinary people, the most direct change is that home appliances are shifting from following commands to assessing situations. The toothbrush will automatically adjust the water flow based on where you are brushing. The robot vacuum will see stains on the floor and decide whether to mop the area twice or bypass it entirely. The air purifier can identify whether the odor is from cigarette smoke or cooking fumes before deciding on the fan speed. You no longer need to tell them that a certain area is dirty or missed a spot; they can see it themselves.
Another change involves privacy. Visual data is much more sensitive than voice data. With cameras pointing at your mouth and your home floor, uploading these images to the cloud carries significant risks. Therefore, edge processing is an inevitable choice, keeping data off the cloud and completing recognition and decision-making locally. Jensen Huang recently predicted at the Goldman Sachs Technology Conference that cybersecurity will become the next major application scenario for AI.
This judgment aligns with the logic of AI growing eyes. When machines can see anomalous behaviors, anomalous traffic, and anomalous code, defense can shift from passive response to active interception. Whether Dyson's CNY 3,899 toothbrush is worth buying is another question. However, the signal behind it deserves attention: AI is stepping out of screens and entering your mouth, your home floor, and every item you use daily. The next time you find your home appliances becoming smarter, it might not be because of what they heard, but because of what they saw.
#Physical AI #Smart Home