The Future of Everything Is Lies, I Guess: New Jobs
5 hours ago
- #ML Jobs
- #AI Ethics
- #Workforce Future
- ML deployment creates new roles at the human-ML boundary: Incanters (prompting experts), Process Engineers (quality control), Statistical Engineers (error modeling), Model Trainers (feeding human expertise), Meat Shields (accountability holders), and Haruspices (behavior interpreters).
- Incanters specialize in crafting inputs for LLMs to achieve optimal outputs, becoming essential as LLMs exhibit unpredictable behaviors like performance degradation and sensitivity to context.
- Process Engineers design and manage quality-control workflows to catch LLM errors, such as using intentional errors in documents and automated systems to ensure accuracy before outputs are used.
- Statistical Engineers measure and control variability in ML systems, addressing issues like bias from input order and domain-specific inaccuracies, akin to psychometrics for chaotic models.
- Model Trainers provide high-quality, uncontaminated data for training LLMs, combating misinformation and subtle errors through expert input and human verification of outputs.
- Meat Shields are humans held accountable for ML system failures, offering legal and social redress where automated systems cannot, with roles ranging from internal reviewers to external scapegoats.
- Haruspices investigate why models fail by analyzing inputs, outputs, and internal states, serving in forensic roles across industries to explain behaviors and assign responsibility.