Deployed machine-learning models often persist unchanged for decades in production

Categorization

• corroborated: 3 independent passages support this claim
• XREF: connects to: Interviews found retraining cadences ranged from hourly to every few months, varying across models — Operationalizing Machine Learning: An Interview Study; AI training data is invisible yet permanently shapes deployed system behavior — Brian Christian, The Alignment Problem, 150; Deployed ML models may stay in production for fifty years — Brian Christian, The Alignment Problem, 229 (+1 more)

Related

• Interviews found retraining cadences ranged from hourly to every few months, varying across models (similarity 0.89)
• Deployed ML models may stay in production for fifty years (similarity 0.86)
• AI training data is invisible yet permanently shapes deployed system behavior (similarity 0.78)