Are strong, deterministic, highly specified guardrails (e.g. context, tests) really required to keep AI on track? If so, then aren't AI engineers applying software engineering to guardrails?
It seems that AI engineering is basically software engineering applied to building AI guardrails.
- Similar to how cloud engineering = software engineering applied to cloud infrastructure.
- Similar to how test engineering = software engineering applied to test infrastructure.
See the pattern?
Software engineering fundamentals + specific knowledge in a sub-field.
Fundamentals don't disappear. Rather, specialisations are added to them.
Discussion on LinkedIn2 likes
