Question 26
Domain 6: Human-AI Interaction and OversightWhat architectural approach would best enable continuous learning and adaptation?
Correct answer: A
Explanation
A static agent paired with a refreshable vector database supports continuous learning because new information can be added without retraining the model, preserving adaptability while reducing cost. A feedback loop that identifies knowledge gaps enables ongoing updates, so the system can “be refreshed with new information” and improve over time.
Why each option is right or wrong
A. Implement a static agent with a regularly updated knowledge base (vector database) that can be refreshed with new information without retraining the model, combined with a feedback loop that identifies knowledge gaps.
A continuously adaptable design is achieved by separating the model from its knowledge store: the agent can retrieve newly indexed content from a refreshable vector database, so updates are incorporated by re-embedding and re-indexing information rather than retraining the underlying model. The feedback loop is the mechanism that detects missing or outdated knowledge and triggers those updates, which is what enables ongoing improvement without the cost and latency of model retraining.
B. Use online learning to continuously retrain the model on every customer interaction, adapting in real-time to new patterns.
C. Deploy a new fine-tuned model every month incorporating the latest product changes and customer interactions.
D. Implement a rule-based system that customer service managers manually update with new rules as the product changes.