-
Notifications
You must be signed in to change notification settings - Fork 0
Open
Description
Reviewing exercise submission:
- Gitlab username: manojpv
- Repository: https://github.com/bossbeagle1509/sse-responsible-ai-assignment
- Pull Request: I acknowledge the use of RAI (https://rai.uni-stuttgart.de/) to imple… bossbeagle1509/sse-responsible-ai-assignment#1
Generating with RAI
- Pro: It seems to have gotten a fair part of the heavy math right.
- Con: The biggest downside is the lack of input validation and error handling—things like checking if nx >= 2 (to avoid division by zero), alpha > 0, or handling missing config files gracefully. This makes it brittle for real-world use, where bad inputs could crash it unexpectedly.
Code Review
- As a Computer Science student, I focused on the code itself and system robustness.
- Copilot focused on mathematical and physical errors or improvements like an expert on simulation software domain.
What I learned
AI can be good at project scaffolding and doing some of the heavy lifting, but in the end someone does need to validate and review the output.
Examples from my challenge project
- Add general lazy evaluation for all blosc2 functions Blosc/python-blosc2#496
- Add saving lazyudf Blosc/python-blosc2#532
Maintainers seem to be focussed on maintainability and correctness of their code as well as not attempting to do too much per merge.
Metadata
Metadata
Assignees
Labels
No labels