Module 1 of 2 · 45 minutes
Where AI meets the line, and how it fails
By the end of this module you will be able to
- Describe four common uses of AI in manufacturing
- Explain false accepts and false rejects in AI inspection
- Explain drift and what causes it
- State what the Health and Safety Executive expects
- Apply management of change to an AI system
Work through it
1 interactive for this module, built on the WAJD Teach engine. Nothing moves until you ask it to, and every one has a written version if you would rather read it.
Watch: Emma and George talk it through
4 minutes. Captions are on, and the same conversation is written out in full below. The voices are computer generated.
Emma George, we've just had a camera system put on line three to inspect parts. Everyone's delighted. Should I be?
George Probably, with your eyes open. Let me set out where AI turns up on a shop floor. Visual inspection, like yours. Predictive maintenance, watching vibration and temperature. Planning and scheduling. And robots and guided vehicles that work near people.
Emma Are they all the same risk?
George No. The first three give advice or decide about product. The fourth moves in the same space as a person. That difference decides how much proof you need before you trust it.
Emma So what can go wrong with my camera?
George It can be wrong in two directions. A false reject scraps a good part. That costs money and you notice quickly. A false accept passes a bad part. That costs far more, and you may not find out until a customer does.
Emma The supplier told me it's ninety nine per cent accurate.
George Then ask the better question. How often does it pass a defect, on which defects, and how do we know? A system tuned to reduce scrap will pass more bad parts. Ask for the false accept rate on the defects your customer cares about most.
Emma And how would I know?
George Only one way. Keep testing it against parts whose true state you've established some other way. Known good, known bad, fed through regularly.
Emma It's been perfect for a month.
George Which is where drift comes in. The system learned from data gathered at one time, under one set of conditions. When conditions change, what it sees no longer matches what it learned from, and its accuracy falls.
Emma Wouldn't it tell me?
George No. Nothing breaks. No alarm. That's what makes it dangerous.
Emma What sort of change?
George Ordinary ones. A new material or supplier. A different surface finish. Someone changes a light fitting or the lens gets dirty. A tool wears. A new product variant. A sensor swapped for a different model.
Emma We changed steel supplier last week.
George Then that's a change the inspection system should have been checked against. Which brings me to the law.
Emma Is there a law on AI in factories?
George The Health and Safety Executive says existing law applies. The 1974 Act is goal setting, so it covers a risk whatever technology creates it. They expect a risk assessment for uses of AI that affect health and safety, and controls so far as is reasonably practicable. Including against cyber threats.
Emma So nothing special. Just do it properly.
George That's their stated aim. That AI risk stops being novel and is managed like any other. In practice, management of change. Putting the system in is a change. Retraining it is a change. Updating its software is a change. And changing the process it watches is a change.
Emma And whose job is that? The supplier's?
George Yours. The duty holder is the employer. Buying the system doesn't transfer the duty to assess and control the risk it creates.
The written material
Four uses you will meet
Visual inspection: a camera and a trained model judge each part as good or bad, faster and more consistently than a person at the end of a shift. Predictive maintenance: software watches vibration, temperature and current, and warns that a component is heading for failure. Planning and scheduling: a tool reorders jobs to meet due dates and changeovers. And collaborative robots and guided vehicles, which use sensing to work near people.
The first three produce advice or a decision about product. The fourth moves in the same space as a person. That difference decides how much proof you need before trusting it.
How inspection fails: two kinds of wrong
An inspection system can be wrong in two directions. A false reject scraps a good part, which costs money and is noticed quickly. A false accept passes a bad part, which costs far more and may not be noticed until a customer finds it.
A system tuned to reduce scrap will pass more bad parts. So the question is never only how accurate is it. It is how often does it pass a defect, on which defects, and how do we know. The only way to know is to keep testing it against parts whose true state you have established another way.
Drift: why it was right last month
An AI system learns from data gathered at one time, under one set of conditions. When conditions change, the data it sees no longer matches what it learned from and its accuracy falls. Nothing breaks. No alarm sounds. This is called drift.
The causes are ordinary: a new material or supplier, a different surface finish, a change of lighting or a dirty lens, a worn tool, a new product variant, a sensor replaced with a different model. Any of them can move a system outside what it was trained on.
- New material, supplier or surface finish
- Lighting changed, lens dirty or camera moved
- Tool wear or a new product variant
- A sensor replaced with a different model
What the HSE expects, and management of change
The Health and Safety Executive's position is that existing law applies. The Health and Safety at Work etc Act 1974 is goal setting, so it covers a risk whatever technology creates it. The HSE expects a risk assessment for uses of AI that affect health and safety, and controls that reduce the risk so far as is reasonably practicable, including controls against cyber security threats. Its stated aim is that AI risk stops being treated as novel and is managed like any other risk.
In practice that means management of change. Introducing an AI system is a change. So is retraining it, updating its software, or changing the process it watches. Each needs assessing before it happens, by someone competent, with a record.
Knowledge check
The knowledge check and your certificate need a free account, so that your progress and results can be saved as evidence.
The learning itself stays free and open. You are reading all of it right now without an account.
Was this module useful? Tell us in two minutes, it decides what we improve next.