AI & TechnologyAn engineer or team wiring a model into a live product
When the same input gives different output each time, I want to test it, so I know whether a change made things better or worse.
Khi cùng một đầu vào cho ra kết quả khác nhau mỗi lần, tôi muốn kiểm thử được nó, để biết một thay đổi làm hệ tốt lên hay tệ đi.
Job context
- Who
- An engineer owning the quality of the AI part
- When
- At every prompt, model or data change
- Trigger
- A small change breaking a case nobody considered
- Situation
- No pass-fail assertion applies to an answer written in prose
- Constraints
- No standard test set, and human review does not run in a pipeline
The four perspectives are for Pro Members
The four perspectives around this Job live inside the community, for Pro Members.
See Pro Member