ITGSS Certified Technical Associate
Emerging Tech Ultimate Exam
**Question 1. Which of the following best describes the primary difference between Artificial
Intelligence (AI) and Machine Learning (ML)?**
A) AI is a subset of ML that focuses on neural networks.
B) ML provides the statistical techniques that enable AI systems to learn from data.
C) AI only deals with rule-based systems, while ML handles data-driven models.
D) ML is a hardware technology, whereas AI is a software methodology.
Answer: B
Explanation: Machine Learning is a set of algorithms that give computers the ability to learn from data,
which is a core technique used to achieve AI capabilities.
**Question 2. In the evolution of neural networks, which architecture introduced the concept of
self-attention that later enabled Transformers?**
A) Convolutional Neural Network (CNN)
B) Recurrent Neural Network (RNN)
C) Long Short-Term Memory (LSTM)
D) Transformer encoder-decoder model
Answer: D
Explanation: The Transformer architecture (Vaswani et al., 2017) introduced self-attention, allowing
parallel processing of sequences and leading to large language models.
**Question 3. Large Language Models (LLMs) such as GPT-4 are primarily trained using which type of
learning paradigm?**
A) Supervised learning with labeled classification data
B) Unsupervised (self-supervised) next-token prediction on massive text corpora
C) Reinforcement learning from human feedback only
D) Semi-supervised learning combining images and text
Answer: B
Explanation: LLMs are trained with self-supervised objectives, predicting the next token given preceding
context, without explicit labels.
, ITGSS Certified Technical Associate
Emerging Tech Ultimate Exam
**Question 4. Which of the following is a key characteristic of generative AI compared to discriminative
AI?**
A) It only classifies inputs into predefined categories.
B) It learns to model the joint probability distribution of data to create new samples.
C) It requires a smaller dataset for accurate predictions.
D) It cannot be used for text generation.
Answer: B
Explanation: Generative AI models the joint distribution P(x, y) and can generate new data resembling
the training set, unlike discriminative models that learn P(y|x).
**Question 5. In supervised learning, which algorithm is most appropriate for a binary classification
problem with a non-linear decision boundary?**
A) Linear Regression
B) K-Nearest Neighbors (KNN)
C) Support Vector Machine with a linear kernel
D) Naïve Bayes
Answer: B
Explanation: KNN makes decisions based on proximity in feature space and can capture complex,
non-linear boundaries without explicit model assumptions.
**Question 6. Which clustering algorithm partitions data by minimizing intra-cluster variance and
requires the number of clusters to be specified beforehand?**
A) DBSCAN
B) Hierarchical Agglomerative Clustering
C) K-means
D) Gaussian Mixture Models (GMM)
Answer: C
, ITGSS Certified Technical Associate
Emerging Tech Ultimate Exam
Explanation: K-means iteratively assigns points to K clusters to minimize within-cluster sum of squares,
requiring K as an input.
**Question 7. In reinforcement learning, the term “policy” refers to:**
A) The reward function that assigns values to states.
B) The mapping from states (or observations) to actions.
C) The environment’s transition dynamics.
D) The discount factor used in value estimation.
Answer: B
Explanation: A policy defines the agent’s behavior by specifying which action to take in each state.
**Question 8. Which of the following is a common source of bias in AI training datasets?**
A) Over-fitting to a validation set
B) Under-representation of certain demographic groups
C) Using too many hidden layers in a neural network
D) Deploying the model on edge devices
Answer: B
Explanation: When certain groups are under-represented, the model may learn patterns that do not
generalize, leading to biased outcomes.
**Question 9. Explainable AI (XAI) techniques such as SHAP and LIME primarily aim to:**
A) Increase model accuracy by adding more layers.
B) Provide post-hoc explanations of individual predictions.
C) Reduce the size of deep learning models for mobile deployment.
D) Encrypt model parameters for security.
Answer: B
Explanation: SHAP and LIME approximate the contribution of each feature to a specific prediction,
offering interpretability.
, ITGSS Certified Technical Associate
Emerging Tech Ultimate Exam
**Question 10. Which legal principle most directly addresses data privacy concerns when using
generative AI that can reproduce copyrighted text?**
A) Fair Use doctrine
B) The principle of least privilege
C) Zero-trust networking
D) OAuth 2.0 authorization
Answer: A
Explanation: Fair Use evaluates whether reproducing copyrighted material for transformation or
commentary is permissible, a key issue for generative AI.
**Question 11. In a blockchain, the Merkle tree is used to:**
A) Store the entire transaction history in a single block.
B) Verify the integrity of all transactions in a block efficiently.
C) Generate the consensus proof for Proof of Stake.
D) Encrypt the block header.
Answer: B
Explanation: A Merkle tree hashes pairs of transaction hashes recursively, allowing a single root hash to
verify any transaction’s integrity.
**Question 12. Which consensus mechanism relies on participants staking cryptocurrency to propose
and validate new blocks?**
A) Proof of Work (PoW)
B) Proof of Authority (PoA)
C) Proof of Stake (PoS)
D) Byzantine Fault Tolerance (BFT)
Answer: C
Explanation: In PoS, validators lock up (“stake”) tokens; the probability of creating the next block is
proportional to the stake.
Emerging Tech Ultimate Exam
**Question 1. Which of the following best describes the primary difference between Artificial
Intelligence (AI) and Machine Learning (ML)?**
A) AI is a subset of ML that focuses on neural networks.
B) ML provides the statistical techniques that enable AI systems to learn from data.
C) AI only deals with rule-based systems, while ML handles data-driven models.
D) ML is a hardware technology, whereas AI is a software methodology.
Answer: B
Explanation: Machine Learning is a set of algorithms that give computers the ability to learn from data,
which is a core technique used to achieve AI capabilities.
**Question 2. In the evolution of neural networks, which architecture introduced the concept of
self-attention that later enabled Transformers?**
A) Convolutional Neural Network (CNN)
B) Recurrent Neural Network (RNN)
C) Long Short-Term Memory (LSTM)
D) Transformer encoder-decoder model
Answer: D
Explanation: The Transformer architecture (Vaswani et al., 2017) introduced self-attention, allowing
parallel processing of sequences and leading to large language models.
**Question 3. Large Language Models (LLMs) such as GPT-4 are primarily trained using which type of
learning paradigm?**
A) Supervised learning with labeled classification data
B) Unsupervised (self-supervised) next-token prediction on massive text corpora
C) Reinforcement learning from human feedback only
D) Semi-supervised learning combining images and text
Answer: B
Explanation: LLMs are trained with self-supervised objectives, predicting the next token given preceding
context, without explicit labels.
, ITGSS Certified Technical Associate
Emerging Tech Ultimate Exam
**Question 4. Which of the following is a key characteristic of generative AI compared to discriminative
AI?**
A) It only classifies inputs into predefined categories.
B) It learns to model the joint probability distribution of data to create new samples.
C) It requires a smaller dataset for accurate predictions.
D) It cannot be used for text generation.
Answer: B
Explanation: Generative AI models the joint distribution P(x, y) and can generate new data resembling
the training set, unlike discriminative models that learn P(y|x).
**Question 5. In supervised learning, which algorithm is most appropriate for a binary classification
problem with a non-linear decision boundary?**
A) Linear Regression
B) K-Nearest Neighbors (KNN)
C) Support Vector Machine with a linear kernel
D) Naïve Bayes
Answer: B
Explanation: KNN makes decisions based on proximity in feature space and can capture complex,
non-linear boundaries without explicit model assumptions.
**Question 6. Which clustering algorithm partitions data by minimizing intra-cluster variance and
requires the number of clusters to be specified beforehand?**
A) DBSCAN
B) Hierarchical Agglomerative Clustering
C) K-means
D) Gaussian Mixture Models (GMM)
Answer: C
, ITGSS Certified Technical Associate
Emerging Tech Ultimate Exam
Explanation: K-means iteratively assigns points to K clusters to minimize within-cluster sum of squares,
requiring K as an input.
**Question 7. In reinforcement learning, the term “policy” refers to:**
A) The reward function that assigns values to states.
B) The mapping from states (or observations) to actions.
C) The environment’s transition dynamics.
D) The discount factor used in value estimation.
Answer: B
Explanation: A policy defines the agent’s behavior by specifying which action to take in each state.
**Question 8. Which of the following is a common source of bias in AI training datasets?**
A) Over-fitting to a validation set
B) Under-representation of certain demographic groups
C) Using too many hidden layers in a neural network
D) Deploying the model on edge devices
Answer: B
Explanation: When certain groups are under-represented, the model may learn patterns that do not
generalize, leading to biased outcomes.
**Question 9. Explainable AI (XAI) techniques such as SHAP and LIME primarily aim to:**
A) Increase model accuracy by adding more layers.
B) Provide post-hoc explanations of individual predictions.
C) Reduce the size of deep learning models for mobile deployment.
D) Encrypt model parameters for security.
Answer: B
Explanation: SHAP and LIME approximate the contribution of each feature to a specific prediction,
offering interpretability.
, ITGSS Certified Technical Associate
Emerging Tech Ultimate Exam
**Question 10. Which legal principle most directly addresses data privacy concerns when using
generative AI that can reproduce copyrighted text?**
A) Fair Use doctrine
B) The principle of least privilege
C) Zero-trust networking
D) OAuth 2.0 authorization
Answer: A
Explanation: Fair Use evaluates whether reproducing copyrighted material for transformation or
commentary is permissible, a key issue for generative AI.
**Question 11. In a blockchain, the Merkle tree is used to:**
A) Store the entire transaction history in a single block.
B) Verify the integrity of all transactions in a block efficiently.
C) Generate the consensus proof for Proof of Stake.
D) Encrypt the block header.
Answer: B
Explanation: A Merkle tree hashes pairs of transaction hashes recursively, allowing a single root hash to
verify any transaction’s integrity.
**Question 12. Which consensus mechanism relies on participants staking cryptocurrency to propose
and validate new blocks?**
A) Proof of Work (PoW)
B) Proof of Authority (PoA)
C) Proof of Stake (PoS)
D) Byzantine Fault Tolerance (BFT)
Answer: C
Explanation: In PoS, validators lock up (“stake”) tokens; the probability of creating the next block is
proportional to the stake.
