When is AI/ML not a good choice?
Why B is the answer
When a specific, exact outcome is needed and a simple rule works
ML gives approximate answers at a cost, so it is a poor fit when a simple rule gives the exact outcome you need.
In detail
- ML gives probabilities.
- If a simple rule gives the exact outcome, use the rule.
- Weigh cost against benefit as well.
Use case
An illustrative scenario, not a real incident.
A tax rate must be applied exactly. A simple rule does that, while a model could only approximate it.
See it in different ways
stateDiagram-v2 [*] --> Data Data --> Train Train --> Evaluate Evaluate --> Deploy Deploy --> Monitor Monitor --> Data: drift
flowchart TD
Q{"Does a simple rule give the exact result?"}
Q -->|"Yes"| A0["Use the rule"]
Q -->|"No"| A1["Consider ML"]Key point: Use a rule when it gives the exact result.
Why the other options are wrong
- A. When a company wants to detect unusual behavior in transactions
- That is a good fit for ML.
- C. When a large volume of decisions needs to be assisted at scale
- That is a good fit for ML.
- D. When a process produces plenty of good historical examples and needs help at scale
- That is a good fit for ML.