“It says green.”
“The target is not level.”
“The software does not care if the target is level. The software sees the pattern. The software says the pattern is in the right place.”
“I care. The camera is going to look at the road, not a pattern.”
“The camera looks at what we tell the camera to look at.”
The technician stood in a car park in Mascot. The car park was near the airport. Planes moved in the sky. The noise from the planes was loud. The technician held a tablet. The tablet was connected to a car. The car was a late-model SUV. The car had a new windscreen. The glass was clean. The glass was clear. The technician had installed the glass an hour ago. Now the technician had to calibrate the camera. The camera lived in a plastic box. The plastic box sat against the glass.
A target stood three metres from the bumper. The target was a board. The board had black and white shapes on the board. The shapes looked like a chess board. The shapes were specific. The camera needed to see the shapes. If the camera saw the shapes, the camera knew where the camera was. This is called calibration. It is a process. The process is digital. The process is binary. The process results in a pass or a failure.
01
Verification vs. Explanation
I typed my password wrong five times this morning. I knew the password. My fingers did not know the password. The screen told me I was wrong. The screen did not tell me I had hit the ‘p’ key instead of the ‘o’ key. The screen only gave me a red box. Modern systems are like the red box. They do not explain. They only verify. When a camera is calibrated, the technician wants the green box. The insurer wants the green box. The customer wants the green box. The green box is a certificate. The certificate goes into the glovebox. The certificate says the car is safe.
Safe is a difficult word.
“Difficulty is not a value; difficulty is a response to input,”
– Zephyr L.M., Complex Systems Specialist
The input in the Mascot car park was a target board. The board sat on a folding stand. A ute was parked nearby. The ute was white. The sun hit the white paint of the ute. The glare from the ute hit the glass of the SUV.
The camera saw the glare. The camera also saw the target. The software processed the image. The software ignored the glare. The software found the black and white shapes. The software decided the shapes were in the correct coordinates. The tablet turned green.
02
The Margin of Error
Everyone was satisfied. The technician was satisfied because the job was done. The owner was satisfied because the car was ready. But the target was not level. The ground in the car park was not flat. There was a slope. The slope was small. The software allowed for a small slope. The software calculated a margin of error. The margin of error is a mathematical buffer. It exists because the world is not a laboratory. But the road is also not a laboratory.
LABORATORY
REAL WORLD
The compounding complexity of external variables versus controlled environments.
When a driver needs an auto glass replacement, the driver thinks about the glass. The glass is the barrier. The glass keeps the rain out. The glass keeps the wind out. In an older car, the glass was just glass. You replaced the glass and you drove away. You checked for leaks. You checked for whistles. If the glass stayed in the frame, the job was a success.
Now, the glass is an optical component. The glass is part of a computer. The camera looks through the glass. The camera sees the world through the glass. If the glass has a tiny ripple, the camera sees a ripple. If the camera is tilted by one millimetre, the camera thinks the lane is in a different place.
Ghost in the Machine
A fortnight later, the driver is on the M5. The M5 is a motorway. The car is moving at . The driver is tired. The lane assist is active. The lane assist uses the camera. The camera looks for the lines on the road. The lines are white. The road is grey. The camera sees the contrast. Suddenly, the steering wheel tugs to the left. The tug is slight. The car is still in the lane. But the car is closer to the edge than it used to be. The driver feels the tug. The driver wonders if the wind moved the car. The driver wonders if the road is uneven.
There is no one to tell. The driver cannot go to a mechanic and say the car feels different. The mechanic will check the computer. The computer will say there are no faults. The computer will show the calibration certificate. The certificate is a document. The document is auditable. You can show the document to a judge. You can show the document to an insurance company. The document says the procedure was performed. The document does not say the camera is accurate. It only says the camera passed a test against a board in a car park in Mascot.
This happens in many places. It happens in medicine. It happens in aviation. It happens in software development. We define a method. We follow the method. We record the method. If the method is followed, we assume the result is correct. This is a dangerous assumption. It is dangerous because the method is designed for a standard environment. The world is not a standard environment. The world has glare from white utes. The world has sloping car parks. The world has old lines on the M5 that are fading.
The Bayside Methodology
Bayside Windscreens performs work in the real world. Mobile technicians go to homes. They go to offices. They work in driveways. They work in yards. A technician who understands the system knows that the green light is the start, not the end. The technician knows that the camera is a sensor. The sensor is sensitive. If the technician does not level the vehicle, the calibration is flawed. If the technician does not check the tyre pressure, the calibration is flawed.
Physical Setup
Level ground, correct tyre pressure, and physical alignment check before the tablet is even powered on.
Optical Clarity
Verifying the new glass mounting points and ensuring the camera’s view is not obstructed by subtle distortions.
If the tyre pressure is low on one side, the car leans. If the car leans, the camera looks at the wrong angle. The tablet might still say green. The tablet only knows what the camera sees. The tablet does not know the car is leaning. This is the gap. It is a gap between the digital record and the physical reality. We trust the digital record because it is easy to read. We trust the record because it has a timestamp. We trust the record because it removes human judgment.
Human judgment is messy. Human judgment is prone to error. But human judgment can see that a target board is leaning. Human judgment can see that the sun is too bright on the glass. The systems in our cars are complex. They are more complex than the people who drive them realize. They are more complex than many people who fix them realize. A windscreen is no longer just a piece of safety glass. It is a mounting platform for a suite of sensors.
The Taxonomy of Risk
These sensors control the brakes. They control the steering. They decide when to warn the driver of a collision. When you change the glass, you change the eyes of the car. If you type a password wrong, you try again. You look at the keys. You move slower. You ensure the ‘Caps Lock’ is off. You use your judgment. A camera cannot use judgment. A camera only uses math.
If the math is based on a tilted board, the math will be wrong. The car will not crash immediately. The car will just be slightly worse at being a car. The car will be less helpful. The tug on the steering wheel will be a little late. The emergency braking will be a little too sensitive. These are small changes. They are hard to measure. They are impossible to audit. The industry relies on the paper. The paper protects the business.
If there is an accident, the business shows the paper. The paper says they followed the rules. The rules were written by the manufacturer. The manufacturer wants the rules to be simple. Simple rules are easy to follow. Simple rules are easy to sell. But safety is not simple. Safety is a constant measurement of reality against expectation.
✓
The Only Calibration That Counts
A driver in South Sydney expects the car to work. The driver pays for a service. The driver receives a certificate. The driver assumes the certificate means the camera sees the road perfectly. The driver does not know about the margin of error. The driver does not know about the tilted board. This is the invisible tax of complexity. We pay for systems that we cannot verify. We rely on technicians who are often forced to work in conditions that the systems were not designed for.
The technician in Mascot finished the job. He packed the target board. He put the stand in his van. He gave the tablet to the customer. The customer signed the screen. The signature was a digital squiggle. The squiggle was another record. The technician drove away. The SUV sat in the car park. The SUV looked the same as it did before. The glass was new. The glass was clean. The camera was inside the box. The camera was ready for the road.
The tablet validates the board while the camera ignores the road.
The road is the only thing that matters. The road does not have a chess board. The road has shadows. The road has rain. The road has other cars. The road is the final test. A good technician knows this. A good technician looks at the car. A good technician looks at the ground. A good technician ignores the tablet until the physical setup is perfect. They do this because they know the green light is a liar if the input is bad.
From Mechanical Certainty to Algorithmic Probability
When the glass is replaced, the relationship between the car and the world is reset. This reset must be precise. It must be handled by people who see the lean in the board. It must be handled by people who know that a car park in Mascot is not a laboratory. Bayside Windscreens understands the South Sydney landscape. They know the driveways. They know the glare from the bay. They know that a mobile service must be more careful, not less, because the environment is always changing.
I will try my password again. I will look at the keys this time. I will not trust my fingers to remember the path. I will verify the input. If I get it right, the red box will go away. If I get it right, I can see my files. The car is the same. If the technician gets it right, the driver can see the road. The driver can trust the tug on the wheel. The driver can drive home on the M5 and not wonder if the wind is blowing. The driver can trust the glass.
The certificate is just a piece of paper. The glass is just a piece of glass. The camera is just a piece of plastic and silicon. The safety is in the alignment of all three. That alignment happens in the hands of the person standing in the car park. It happens before the tablet says green. It happens when the human looks at the world and decides it is level. That is the only calibration that truly counts. Everything else is just paperwork.
When the camera sees the line, and the line is actually there, the system works. When the camera sees a line that is ten centimetres to the left because of a tilted target, the system is a ghost. It is a ghost in the machine that pulls at the wheel when you are tired. We deserve better than ghosts. We deserve the truth of the road. The truth of the road starts with the technician. It starts with the glass. It starts with the realization that a green box on a tablet is not a guarantee. It is only a start.