A company is developing an AI character for a video game. The AI character needs to learn how to navigate a complex environment and make decisions to…

The scenario describes a classic application of reinforcement learning (RL). In RL, an agent (the AI character) learns to make decisions by performing actions in an environment (the video game) to maximize a cumulative reward. The system of positive and negative scores directly corresponds to the reward/penalty mechanism central to RL. The agent learns an optimal policy through trial and error, associating actions with long-term rewards, which perfectly aligns with the described process of improving the character's ability to play the game effectively.

B. Unsupervised learning: This approach is used to find hidden patterns or intrinsic structures in unlabeled data, such as clustering similar items. It does not involve an agent learning from rewards or penalties.

C. Supervised learning: This requires a pre-existing, labeled dataset where correct outputs are known for given inputs. The AI character is learning from experience, not from a dataset of "correct" moves.

D. Deep learning: This is a set of techniques using neural networks, not a fundamental learning paradigm like RL. While deep learning can be used to implement an RL agent (Deep Reinforcement Learning), RL is the overarching methodology described.

1. Google Cloud Documentation: "Introduction to reinforcement learning" on Vertex AI states

"Reinforcement learning (RL) is a machine learning technique that trains a software agent to make decisions by performing actions in an environment to maximize a cumulative reward... RL is a good choice for problems that require making a sequence of decisions to achieve a goal

such as playing a game..."

Source: Google Cloud. (n.d.). Introduction to reinforcement learning. Vertex AI Documentation. Retrieved from https://cloud.google.com/vertex-ai/docs/start/introduction-reinforcement-learning. (Refer to the "Overview" section).

2. Academic Publication (Foundational Text): Sutton

R. S.

& Barto

A. G. (2018). Reinforcement Learning: An Introduction. MIT Press.

Details: Chapter 1

Section 1.1

"Reinforcement Learning

" describes the core concept: "...how an agent ought to take actions in an environment so as to maximize some notion of cumulative reward. Reinforcement learning is one of three basic machine learning paradigms

alongside supervised learning and unsupervised learning." This directly maps to the game character scenario.

3. University Courseware: Stanford University

CS 234: Reinforcement Learning.

Details: The course overview defines the RL problem as an agent interacting with an environment over time

choosing actions to maximize a scalar reward signal. This is precisely the problem faced by the AI character in the video game. (Refer to Winter 2022

Lecture 1 slides

"What is Reinforcement Learning?").