Chapter 37 Quiz: Emerging AI Technologies

Chapter 37 Quiz: Emerging AI Technologies

Multiple Choice

Question 1. What distinguishes an AI agent from a traditional chatbot?

(a) AI agents use larger language models than chatbots.
(b) AI agents can plan multi-step tasks, use external tools, and act with some degree of autonomy, rather than merely responding to individual prompts.
(c) AI agents do not require language models; they use rule-based systems.
(d) AI agents are always more accurate than chatbots.

Question 2. On the AI Technology Radar described in this chapter, a technology in the "Trial" ring should receive which level of organizational investment?

(a) No investment; monitor only.
(b) Research and build understanding through white papers and vendor briefings.
(c) Bounded pilots with clear success criteria, defined timelines, and pre-committed decision processes.
(d) Full production deployment with dedicated engineering support.

Question 3. Which of the following is NOT listed as a reliable current capability of agentic AI systems?

(a) Research and information synthesis from multiple sources.
(b) High-stakes decisions where errors have serious consequences.
(c) Structured data workflows such as extracting data from documents.
(d) Code generation and debugging.

Question 4. The primary advantages of edge AI over cloud-based AI include all of the following EXCEPT:

(a) Lower latency for real-time applications.
(b) Enhanced data privacy because data stays on-device.
(c) Superior model accuracy compared to cloud-based models.
(d) Continued operation without internet connectivity.

Question 5. Knowledge distillation is best described as:

(a) Reducing the numerical precision of model weights from 32-bit to 8-bit.
(b) Training a smaller "student" model to mimic the behavior of a larger "teacher" model.
(c) Removing unnecessary connections from neural networks.
(d) Designing model architectures optimized for specific hardware from the ground up.

Question 6. According to the chapter, the most credible timeline for quantum computers to achieve practical business impact in machine learning is:

(a) Already happening today.
(b) 1-2 years.
(c) 5-15 years.
(d) 30+ years.

Question 7. Which statement best characterizes the relationship between training costs and inference costs for AI models?

(a) Both are rising rapidly due to increasing model complexity.
(b) Training costs for frontier models are rising, while inference costs are falling dramatically.
(c) Both are falling due to hardware improvements and competition.
(d) Training costs are falling due to open-source models, while inference costs are rising due to demand.

Question 8. A company in a regulated industry (financial services) needs to deploy an AI model that processes sensitive customer data at high volume (5 million queries per day). Based on the chapter's analysis, which approach is most likely appropriate?

(a) A closed-model API (e.g., GPT-4) for maximum capability.
(b) A self-hosted open-weight model fine-tuned on domain data, for data privacy and cost control.
(c) A neuromorphic computing solution for maximum efficiency.
(d) A quantum machine learning approach for superior optimization.

Question 9. Athena's AI Technology Radar placed agentic AI in which ring after the NovaMart competitive threat?

(a) Hold.
(b) Assess.
(c) Trial.
(d) Adopt.

Question 10. Which of the following best describes "model collapse" as documented by Shumailov et al.?

(a) A neural network that fails to converge during training.
(b) Progressive degradation in model quality when language models are trained on outputs from other language models over successive generations.
(c) The sudden failure of a deployed model due to data drift.
(d) The inability of small models to match the performance of large models.

Question 11. Neuromorphic computing differs from conventional computing primarily because:

(a) It uses quantum superposition to perform parallel calculations.
(b) It mimics the brain's architecture by co-locating memory and processing and using event-driven computation.
(c) It uses larger transistors for greater power efficiency.
(d) It processes information exclusively in the cloud rather than on-device.

Question 12. According to the chapter, the most commercially mature application of AI robotics is:

(a) Humanoid robots for household tasks.
(b) Fully autonomous vehicles on public roads.
(c) Warehouse and logistics automation.
(d) Neuromorphic-powered surgical robots.

Question 13. An organization that launches pilots for every new AI technology, spreads resources too thin, and never achieves production deployment is exhibiting which failure mode?

(a) Analysis paralysis.
(b) Shiny object syndrome (chasing every trend).
(c) Pilot purgatory.
(d) Technology debt accumulation.

Question 14. Which of the following is true about synthetic data?

(a) Models trained exclusively on synthetic data consistently outperform models trained on real data.
(b) Synthetic data eliminates the need for data privacy considerations.
(c) Best practice is to use synthetic data to augment real data, not to replace it entirely, and to validate model performance on held-out real data.
(d) Synthetic data can only be generated using Generative Adversarial Networks (GANs).

Question 15. Grace Chen's strategic decision for Athena's AI shopping assistant is to:

(a) Abandon the online market and focus on in-store experience.
(b) Acquire NovaMart to eliminate the competitive threat.
(c) Build a competing AI shopping assistant with governance guardrails, betting on trust as a differentiator.
(d) License NovaMart's technology and rebrand it.

Question 16. The chapter identifies Google TPUs, Amazon Trainium, and Groq LPUs as examples of:

(a) Quantum computing processors designed for AI workloads.
(b) Neuromorphic chips that mimic brain architecture.
(c) Custom AI chips developed as alternatives to NVIDIA GPUs.
(d) Edge AI processors for on-device inference only.

Question 17. According to the chapter, which action should most organizations take regarding quantum computing today?

(a) Invest heavily in building internal quantum computing capability.
(b) Begin transitioning to post-quantum cryptography while monitoring quantum developments, but do not invest in quantum AI capability.
(c) Ignore quantum computing entirely, as it will never be practically relevant.
(d) Deploy quantum machine learning algorithms immediately for optimization tasks.

Question 18. In the multi-agent system architecture described in the chapter, the "orchestrator agent" is responsible for:

(a) Training other agents through reinforcement learning.
(b) Coordinating the team of agents, assigning tasks, and resolving conflicts.
(c) Providing the highest-quality outputs among all agents in the team.
(d) Connecting agents to external APIs and databases.

Question 19. The chapter compares RPA (Robotic Process Automation) to agentic AI. Which analogy best captures the distinction?

(a) RPA is like a calculator; agentic AI is like a computer.
(b) RPA is like following a recipe step-by-step; agentic AI is like cooking a meal when you understand the principles and can improvise.
(c) RPA handles physical tasks; agentic AI handles digital tasks.
(d) RPA is slower but more reliable; agentic AI is faster but always less reliable.

Question 20. Which of the following was NOT identified as a reason small language models (SLMs) may be preferable to frontier models for certain enterprise applications?

(a) Lower inference costs at high query volumes.
(b) Easier fine-tuning on domain-specific data.
(c) Superior performance on all tasks, regardless of complexity.
(d) Deployment flexibility, including on-device and private infrastructure.

Short Answer

Question 21. Explain why Tom's hiring as a consultant to architect Athena's AI shopping assistant connects to the textbook's broader theme of bridging technical and business expertise. What unique value does his MBA training add to a technical architecture project? (3-5 sentences)

Question 22. The chapter argues that "the most dangerous misconception about agentic AI is that 'autonomy' means 'reliability.'" Explain this statement and provide an example of how silent failure in an agentic AI system could cause more damage than a loud failure. (4-6 sentences)

Question 23. Professor Okonkwo states that AI strategy "must be a living framework that adapts as the cost structure changes." Using the example of hardware economics from this chapter, explain how a cost structure change (e.g., falling inference costs or rising GPU prices) could require a strategic pivot in an organization's AI approach. (4-6 sentences)

Question 24. Describe the five-question evaluation framework used in the AI Technology Radar process. Why is the question "What are the risks?" particularly important for emerging AI technologies, compared to more mature technology categories? (4-6 sentences)

Question 25. The chapter identifies two symmetric failure modes for emerging technology adoption: "chasing every trend" and "waiting too long." Explain how the AI Technology Radar framework addresses both failure modes. What is one limitation of the radar approach? (4-6 sentences)

Answer key available in Appendix B. Questions marked with an asterisk () have detailed worked solutions.*