1. Reskilling policies are often proposed to address AI-driven disruptions. 2. Automation through AI is limited to routine manual jobs. 3. AI can create new categories of jobs while displacing others. Which statements are correct?

Regarding Generative AI, which statements are correct: 1. Its outputs are based on learned patterns, not retrieval of stored information. 2. It can generate novel content not tied to real facts. Its outputs are always deterministic for the same input.

1. Personal data used in AI systems can raise privacy concerns. 2. Anonymisation of data always eliminates the risk of re-identification. 3. Data governance frameworks may include consent and purpose limitation. Which of the statements are correct?

1. Algorithmic bias can arise even when training data is unbiased. 2. Transparency in AI systems ensures complete explainability of outcomes. 3. Regulatory frameworks aim to balance innovation with ethical concerns. Which of the statements are correct?

1. GPUs are optimised for parallel processing tasks. 2. AI accelerators efficiently perform matrix and tensor operations. 3. CPUs are generally preferred over GPUs for latency-sensitive, low-volume AI inference tasks. Which of the statements are correct?

What is correct for Edge AI chips: 1. They can perform inference tasks locally without relying entirely on cloud computing. 2. They reduce latency in real-time applications. 3. They are incapable of functioning without continuous internet connectivity.

What statements about specialised AI chips are correct? 1. They improve performance per watt vs general-purpose processors. 2. They use architectures optimised for AI workloads. 3. They always outperform general-purpose processors in all tasks.

1. TPUs are specifically optimised for tensor operations used in machine learning. 2. GPUs are used only for training and not for inference. 3. Both GPUs and TPUs can be used for AI workload. Which statements are correct?

1. LLMs generate responses predicting the next token based on learned patterns in data. 2. LLMs understand the meaning of text in the same way as humans. 3. LLMs can produce coherent outputs even without explicit knowledge of facts. Which are correct?

1. Pre-training of LLMs often involves self-supervised learning on large text datasets. 2. Fine-tuning allows LLMs to adapt to domain-specific tasks. 3. LLMs cannot generate meaningful outputs without task-specific fine-tuning. Which are correct?

1. LLMs can produce plausible but inaccurate outputs. 2. LLMs can be adapted for specific tasks via fine-tuning or prompt engineering. 3. LLMs always retrieve information from a fixed database during response generation. Which statements are correct?

What is correct for QKD? 1. It enables secure key exchange based on principles of quantum mechanics. 2. Any attempt to measure quantum states during transmission can disturb them. 3. It guarantees security even if the implementation system is flawed.

What is correct for quantum cryptography? 1. It relies on the principle that quantum states cannot be copied perfectly. 2. It uses entanglement to detect any eavesdropping. 3. It allows unlimited copying of encryption keys once securely generated.

1. Quantum cryptography derives security from physical laws. 2. Quantum cryptography eliminates the need for any classical communication channel. 3. Classical cryptography relies on computational hardness assumptions. Which of the statements are correct?

1. QKD can detect eavesdropping through correlation changes. 2. QKD protocols can function over long distances without any signal loss. 3. Quantum cryptography ensures unconditional security under ideal conditions. Which statements are correct?