AI is the capability of computational systems to perform tasks associated with human intelligence such as learning, reasoning, perception, and decision-making. This is called the .

Context: Definition and Scope of Artificial Intelligence

Reasoning aims to derive solutions from information, but large problems can cause , making them computationally hard.

Context: Reasoning and Problem-Solving

A has goals or preferences and takes actions to achieve them.

Context: Agents, Utility, and Expected Utility

In decision-making, an agent assigns a numerical measure of how much it prefers a situation called .

Context: Agents, Utility, and Expected Utility

When outcomes are uncertain, the agent chooses actions by maximizing , which weights the utility of possible outcomes by their probabilities.

Context: Agents, Utility, and Expected Utility

An MDP models decision-making using a transition model (probabilities), a reward function (utility/cost), and a policy mapping states to actions. This core model is called a .

Context: Planning and Decision-Making (Agents, Utility, MDPs)

trains on labeled data where inputs come with expected answers.

Context: Machine Learning Paradigms

finds patterns or structures in unlabeled data without expected answers.

Context: Machine Learning Paradigms

rewards good responses and punishes bad ones so the agent learns good actions.

Context: Machine Learning Paradigms

A deep learning architecture using an attention mechanism, widely used in modern NLP, is the .

Context: Natural Language Processing (NLP) and Modern Transformers

In the cause→effect chain: Using GPUs to accelerate neural networks causes AI funding and interest to increase substantially after .

Context: AI history cause→effect (GPUs and deep learning)

In the cause→effect chain: Transformer architecture adoption after 2017 causes AI growth to accelerate further. The mechanism is that attention-based architectures improved learning in sequence tasks like .

Context: Transformers and NLP mechanism

In the cause→effect chain: Commonsense knowledge is broad and often not represented as explicit facts, which causes knowledge representation to become difficult for AI systems. This happens because the system must acquire and encode vast implicit knowledge that humans assume. The core concept being tested here is .

Context: Knowledge representation challenges

State space search explores a tree of possible states to find a goal state, while local search begins with a guess and refines it incrementally using mathematical optimization. The correct term for exploring the tree of states is .

Context: AI Techniques: Search and Optimization

Gradient descent is a local search method that optimizes numerical parameters by incrementally minimizing a loss function. It is commonly used with backpropagation to train neural networks. Therefore, gradient descent is best described as a .

Context: Search and Optimization Techniques (Gradient descent)