The 80% Problem: The Last 20% Is Where the Engineer Used to Live
3 hours ago
- #synthetic competence
- #skill atrophy
- #AI-assisted engineering
- AI accelerates the creation of a working draft (the first 80%) but often omits the critical last 20%, which includes edge cases, failure modes, and operational realities.
- The last 20% traditionally taught engineers judgment through hands-on experience with issues like concurrency, scalability, and debugging under production conditions.
- Synthetic competence refers to AI-generated output that appears fluent and correct but lacks deep understanding, posing risks when there's no external validation.
- Automating routine tasks can lead to skill atrophy, as supervisors lose the practice needed to handle rare failures, echoing Bainbridge's 'Ironies of Automation'.
- To mitigate risks, organizations must deliberately preserve apprenticeships, on-call rotations, and hands-on practice to maintain expertise in critical areas.
- Effective use of AI requires planning to iteratively address the missing 20% after deployment, rather than assuming initial output is complete.