Chatbots, Virtual Assistants, LLMs, GenAI, Foundation Models & AI Cost Economics | 100+ MCQs
Q1. What is the primary purpose of a chatbot?
A. Managing personal schedules
B. Performing specific, task-oriented interactions
C. Predicting future trends
D. Controlling smart devices
✅ Answer: B
Q2. Which characteristic best differentiates a virtual assistant from a chatbot?
A. Rule-based responses
B. Narrow task focus
C. Multitasking and contextual understanding
D. Text-only interaction
✅ Answer: C
Q3. Chatbots are MOST suitable for which business scenario?
A. Strategic decision-making
B. Repetitive customer queries
C. Complex reasoning tasks
D. Long-term forecasting
✅ Answer: B
Q4. Virtual assistants primarily rely on which AI capability?
A. Keyword matching
B. Decision trees
C. Natural Language Processing
D. SQL querying
✅ Answer: C
Q5. Which interface is MOST commonly used by chatbots?
A. Gesture-based
B. Brain–computer
C. Chat-based text interface
D. AR interface
✅ Answer: C
Q6. Which interface differentiates virtual assistants from chatbots?
A. Menu-driven UI
B. Voice-based interaction
C. Static UI
D. Form-based UI
✅ Answer: B
Q7. Which example represents a virtual assistant?
A. Zobot
B. Drift
C. Apple Siri
D. Zoho SalesIQ
✅ Answer: C
Q8. Reordering prescriptions hands-free best illustrates which AI system?
A. Chatbot
B. Expert system
C. Virtual assistant
D. Search engine
✅ Answer: C
Q9. Which key limitation applies to chatbots compared to virtual assistants?
A. Higher infrastructure cost
B. Limited contextual understanding
C. Slower response time
D. Lack of UI
✅ Answer: B
Q10. Zia’s competitive advantage lies primarily in its:
A. Gaming capability
B. Zoho ecosystem integration
C. Open-source architecture
D. Social media automation
✅ Answer: B
Q11. Zobot is best described as:
A. A virtual assistant
B. A foundation model
C. A chatbot development platform
D. A forecasting engine
✅ Answer: C
Q12. Which task is NOT typically handled by Zobot?
A. Lead qualification
B. Appointment scheduling
C. Strategic forecasting
D. Customer support
✅ Answer: C
Q13. Why do API-based LLM costs scale linearly?
A. Fixed infrastructure pricing
B. Token-based billing
C. License-based pricing
D. Flat monthly plans
✅ Answer: B
Q14. What is the main cost driver in LLM inference?
A. Human supervision
B. Storage cost
C. GPU computation
D. Network latency
✅ Answer: C
Q15. Why is LLM “search” more expensive than Google search?
A. No caching
B. No indexing mechanism
C. Manual moderation
D. Licensing cost
✅ Answer: B
Q16. What fundamentally happens when an LLM answers a query?
A. Database lookup
B. Rule execution
C. Full neural network forward pass
D. Cached retrieval
✅ Answer: C
Q17. Which cost component is unique to self-hosted LLMs?
A. Token pricing
B. API markup
C. Upfront training cost
D. Prompt engineering
✅ Answer: C
Q18. Open-source software provides which freedom?
A. Restricted redistribution
B. Source code modification
C. Vendor-only updates
D. License expiration
✅ Answer: B
Q19. Which is a proprietary software example?
A. Linux
B. Firefox
C. Ubuntu
D. Microsoft Office
✅ Answer: D
Q20. Which entity maintains open-source software?
A. Closed vendor team
B. Government body
C. Open developer community
D. Cloud provider
✅ Answer: C
Q21. Foundation models are BEST described as:
A. Task-specific algorithms
B. Rule-based engines
C. General-purpose pretrained models
D. Statistical regression tools
✅ Answer: C
Q22. Which feature enables foundation models to adapt without retraining?
A. Batch learning
B. In-context learning
C. Reinforcement learning
D. Transfer entropy
✅ Answer: B
Q23. All LLMs are:
A. Proprietary systems
B. Narrow AI
C. Foundation models
D. Predictive models
✅ Answer: C
Q24. Which is NOT an LLM?
A. GPT-4
B. Claude
C. DALL·E
D. LLaMA
✅ Answer: C
Q25. Generative AI differs from LLMs because it:
A. Only generates text
B. Includes multiple content modalities
C. Excludes images
D. Is rule-based
✅ Answer: B
Q26. Which model generates images?
A. GPT-4
B. ARIMA
C. DALL·E
D. Logistic regression
✅ Answer: C
Q27. LLMs are optimized to predict:
A. Stock prices
B. Weather patterns
C. Next token in text
D. Sales demand
✅ Answer: C
Q28. Why can’t LLMs replace ARIMA?
A. Lack of memory
B. Lack of numeric modeling structure
C. High latency
D. Poor UX
✅ Answer: B
Q29. Which task suits quantitative prediction models BEST?
A. Text summarization
B. Image generation
C. Sales forecasting
D. Conversational response
✅ Answer: C
Q30. Which metric evaluates quantitative prediction accuracy?
A. BLEU
B. ROUGE
C. RMSE
D. Perplexity
✅ Answer: C
Q31. LLMs are MOST effective in forecasting workflows for:
A. Generating final numbers
B. Explaining forecast results
C. Simulating physics
D. Replacing numeric models
✅ Answer: B
Q32. Which model explicitly handles seasonality and trend?
A. GPT
B. ARIMA
C. DALL·E
D. Word2Vec
✅ Answer: B
Q33. Time-series models primarily rely on:
A. Text corpora
B. Image embeddings
C. Historical numeric data
D. Prompt templates
✅ Answer: C
Q34. Which is an example of a probabilistic prediction model?
A. Linear regression
B. Logistic regression
C. ARIMA
D. Moving average
✅ Answer: B
Q35. Hybrid AI systems combine LLMs with numeric models to:
A. Reduce compute
B. Improve explanations and usability
C. Eliminate math
D. Avoid data preprocessing
✅ Answer: B
Q36. Chronos and TimeGPT are examples of:
A. Text LLMs
B. Image generators
C. Time-series foundation models
D. Rule-based systems
✅ Answer: C
Q37. Which AI layer acts as the “orchestrator” in multimodal products?
A. Image generator
B. LLM
C. Database
D. GPU driver
✅ Answer: B
Q38. Canva Magic Studio combines LLMs with:
A. Search engines
B. Image GenAI models
C. ERP systems
D. CRM databases
✅ Answer: B
Q39. Emergent abilities arise in foundation models due to:
A. Manual coding
B. Data labeling
C. Scale and diversity
D. Rule engines
✅ Answer: C
Q40. A key ethical concern of foundation models is:
A. Slow execution
B. Bias propagation
C. Low accuracy
D. Narrow scope
✅ Answer: B
Q41. Which role do LLMs play in prediction systems?
A. Calculator
B. Forecaster
C. Analyst
D. Sensor
✅ Answer: C
Q42. Which statement is TRUE?
A. GenAI is narrower than LLMs
B. LLMs can directly forecast weather
C. Every GenAI is an LLM
D. Every LLM is part of GenAI
✅ Answer: D
Q43. Proprietary software restricts:
A. Commercial use
B. Source code access
C. Vendor control
D. Licensing
✅ Answer: B
Q44. Which factor makes APIs cheaper for startups?
A. Free tokens
B. No upfront training cost
C. Unlimited usage
D. Zero inference cost
✅ Answer: B
Q45. What increases with every LLM query?
A. Model size
B. Compute cost
C. Training data
D. Accuracy
✅ Answer: B
Q46. Which AI system best fits repetitive FAQ handling?
A. LLM-based agent
B. Virtual assistant
C. Chatbot
D. Forecasting model
✅ Answer: C
Q47. Which is NOT a foundation model trait?
A. Adaptability
B. Narrow task focus
C. Large-scale training
D. Multimodality
✅ Answer: B
Q48. Prediction models aim to reduce:
A. Model size
B. Bias
C. Uncertainty
D. Token usage
✅ Answer: C
Q49. LLM randomness is controlled by:
A. Epochs
B. Temperature
C. Learning rate
D. Batch size
✅ Answer: B
Q50. Which system is deterministic and reproducible?
A. LLM
B. GenAI
C. ARIMA
D. Chatbot
✅ Answer: C
Q51. Which AI system is BEST suited for automating repetitive customer service tasks?
A. Foundation model
B. Virtual assistant
C. Chatbot
D. Prediction model
✅ Answer: C
Q52. Which feature allows virtual assistants to understand complex user intent?
A. Static scripts
B. Rule engines
C. Natural Language Processing
D. SQL indexing
✅ Answer: C
Q53. Zobot primarily helps businesses by:
A. Forecasting demand
B. Generating images
C. Automating customer interactions
D. Training foundation models
✅ Answer: C
Q54. Which cost component is paid only once when building an in-house LLM?
A. Token usage
B. Inference compute
C. GPU electricity
D. Model training
✅ Answer: D
Q55. Why do large-scale LLM applications prefer self-hosting at high volume?
A. Better UX
B. Lower per-query marginal cost
C. Free GPUs
D. Faster prompts
✅ Answer: B
Q56. Which process converts user text into tokens in an LLM pipeline?
A. Indexing
B. Tokenization
C. Normalization
D. Parsing
✅ Answer: B
Q57. Why is LLM inference NOT comparable to database search?
A. Lack of caching
B. Absence of indexing
C. Requirement of full neural computation
D. Limited data storage
✅ Answer: C
Q58. Which GPU resource is critical for fast LLM responses?
A. Disk I/O
B. CPU cache
C. VRAM
D. Network bandwidth
✅ Answer: C
Q59. Which pricing model do commercial LLM APIs use?
A. Flat subscription
B. License-based
C. Per-token usage
D. Per-user pricing
✅ Answer: C
Q60. Which organization type typically maintains open-source software?
A. Closed vendors
B. Government agencies
C. Developer communities
D. Hardware manufacturers
✅ Answer: C
Q61. Which is a KEY advantage of open-source software?
A. Vendor lock-in
B. Restricted access
C. Source code transparency
D. Limited redistribution
✅ Answer: C
Q62. Proprietary software licenses often restrict:
A. Installation count
B. Vendor profits
C. User support
D. Security updates
✅ Answer: A
Q63. Which statement is TRUE about proprietary software?
A. Source code is publicly editable
B. Developed through open collaboration
C. Managed by a closed team
D. Always free of cost
✅ Answer: C
Q64. Foundation models differ from traditional ML models because they are:
A. Narrow and task-specific
B. Rule-based
C. General-purpose and reusable
D. Low compute
✅ Answer: C
Q65. Self-supervised learning allows foundation models to:
A. Use labeled datasets only
B. Learn from unlabeled data
C. Avoid pretraining
D. Eliminate bias
✅ Answer: B
Q66. Which capability emerges as foundation models scale?
A. Manual coding
B. Emergent reasoning
C. Deterministic output
D. Reduced compute cost
✅ Answer: B
Q67. Which model is multimodal by design?
A. ARIMA
B. Logistic regression
C. Gemini 1.5
D. Linear regression
✅ Answer: C
Q68. Fine-tuning a foundation model primarily aims to:
A. Reduce hardware usage
B. Specialize for a domain
C. Eliminate training
D. Improve UI
✅ Answer: B
Q69. Which statement BEST describes an LLM?
A. A numeric forecasting engine
B. A language-focused foundation model
C. A rules-based chatbot
D. A search index
✅ Answer: B
Q70. Generative AI includes models that generate:
A. Only structured tables
B. Only text
C. Text, images, audio, and video
D. Only forecasts
✅ Answer: C
Q71. Which GenAI tool is used for video generation?
A. Runway
B. GPT-4
C. ARIMA
D. Excel
✅ Answer: A
Q72. Why are LLMs unsuitable for direct numeric forecasting?
A. Lack of GPUs
B. Lack of mathematical structure
C. Small dataset size
D. Limited vocabulary
✅ Answer: B
Q73. Which quantitative model explicitly handles autocorrelation?
A. Linear regression
B. Logistic regression
C. ARIMA
D. Naïve Bayes
✅ Answer: C
Q74. RMSE is used to evaluate:
A. Text quality
B. Image resolution
C. Forecast accuracy
D. Language fluency
✅ Answer: C
Q75. LLMs BEST assist prediction models by:
A. Replacing them
B. Generating numeric outputs
C. Explaining and interpreting results
D. Eliminating uncertainty
✅ Answer: C
Q76. Which step comes FIRST in building a prediction model?
A. Parameter estimation
B. Data validation
C. Problem definition
D. Forecast generation
✅ Answer: C
Q77. Regression models predict outcomes based on:
A. Random sampling
B. Historical averages
C. Independent variables
D. Token probabilities
✅ Answer: C
Q78. Which prediction model estimates probabilities of outcomes?
A. ARIMA
B. Bayesian model
C. Moving average
D. Trend model
✅ Answer: B
Q79. Causal models focus on:
A. Correlation only
B. Random variation
C. Cause-and-effect relationships
D. Token prediction
✅ Answer: C
Q80. Which field commonly uses prediction models?
A. Literature
B. Philosophy
C. Operations management
D. Linguistics
✅ Answer: C
Q81. LLMs optimize for:
A. RMSE minimization
B. Statistical significance
C. Linguistic plausibility
D. Deterministic accuracy
✅ Answer: C
Q82. Which factor reduces reproducibility in LLM outputs?
A. Batch size
B. Temperature sampling
C. Dataset size
D. GPU memory
✅ Answer: B
Q83. Which quantitative model provides transparent coefficients?
A. GPT-4
B. ARIMA
C. DALL·E
D. Chatbot
✅ Answer: B
Q84. Foundation models raise ethical concerns primarily due to:
A. Narrow usage
B. General-purpose deployment
C. Small datasets
D. Open licenses
✅ Answer: B
Q85. Bias in foundation models mainly originates from:
A. GPU hardware
B. Training data
C. Prompt length
D. UI design
✅ Answer: B
Q86. Which system is MOST interpretable?
A. LLM
B. GenAI
C. Linear regression
D. Multimodal model
✅ Answer: C
Q87. Hybrid AI systems combine numeric and language models to:
A. Increase randomness
B. Improve interpretability and usability
C. Eliminate math
D. Reduce data needs
✅ Answer: B
Q88. Which specialized model directly predicts time-series values?
A. GPT-4
B. DALL·E
C. TimeGPT
D. Chatbot
✅ Answer: C
Q89. In prediction workflows, LLMs act primarily as:
A. Forecasters
B. Calculators
C. Analysts and explainers
D. Sensors
✅ Answer: C
Q90. Which statement is CORRECT?
A. LLMs replace statistical models
B. GenAI excludes text models
C. Foundation models enable transfer learning
D. Prediction models generate language
✅ Answer: C
Q91. Which AI model predicts the next word in a sentence?
A. ARIMA
B. Regression
C. LLM
D. Decision tree
✅ Answer: C
Q92. Quantitative prediction models require:
A. Unstructured text
B. Live numeric data
C. Prompt templates
D. Image embeddings
✅ Answer: B
Q93. Which model handles seasonality explicitly?
A. Logistic regression
B. ARIMA
C. GPT
D. Naïve Bayes
✅ Answer: B
Q94. Which cost factor applies to both API and self-hosted LLMs?
A. Training data licensing
B. Inference compute
C. Vendor markup
D. Subscription fees
✅ Answer: B
Q95. LLM inference cost increases mainly with:
A. Model name
B. Token volume
C. UI complexity
D. Training epochs
✅ Answer: B
Q96. Which AI tool best fits FAQ automation with minimal cost?
A. Virtual assistant
B. LLM agent
C. Rule-based chatbot
D. Forecasting model
✅ Answer: C
Q97. Foundation models differ from traditional ML because they:
A. Use less data
B. Learn representations transferable across tasks
C. Are rule-based
D. Avoid pretraining
✅ Answer: B
Q98. Which statement BEST summarizes GenAI?
A. AI that predicts numbers
B. AI that retrieves documents
C. AI that generates new content
D. AI that classifies data
✅ Answer: C
Q99. Which system best supports multimodal user interaction?
A. ARIMA
B. Chatbot
C. Foundation model
D. Spreadsheet
✅ Answer: C
Q100. Best practice for enterprise prediction systems is to:
A. Use only LLMs
B. Avoid AI
C. Combine LLMs with quantitative models
D. Replace forecasting with chatbots
✅ Answer: C
Chatbot vs Virtual Assistant MCQs, LLM vs GenAI MCQs, Foundation Model MCQs, MBA AI MCQs, AI Cost Economics MCQs, Open Source vs Proprietary Software MCQs, Prediction Models MCQs, Quantitative Techniques AI
