Elements of AI & Problem Solving Study Guide

Flashcards created from lecture notes covering AI fundamentals, machine learning, neural networks, problem-solving, and societal implications.

What is the general definition of Artificial Intelligence (AI) provided in the transcript?

AI is the ability of a computer or machine to think, learn, and make decisions like a human by using data and algorithms.

What is the difference between autonomy and adaptivity in AI key terminology?

Autonomy is the ability to perform tasks in complex environments without constant guidance by a user, whereas adaptivity is the ability to improve performance by learning from experience.

Who coined the term "suitcase words" and what does it mean in AI?

Marvin Minsky coined "suitcase words" for terms that carry many different internal meanings, which can cause misunderstandings when used.

What is the difference between Narrow Artificial Intelligence and General Artificial Intelligence?

Narrow AI is designed to perform one specific task within a limited scope without general understanding, whereas General AI is a theoretical concept that refers to a machine capable of reasoning, learning, and performing multiple tasks across domains like a human.

What are four practical applications of Narrow AI in accounting practice?

Cost forecasting and trend analysis tools, fraud and anomaly detection systems, budget variance analysis tools, and credit and financial risk assessment models.

How is Machine Learning (ML) defined in the notes?

Machine Learning is a subfield of AI and computer science consisting of systems that improve their performance in a given task with more experience or data.

What is Deep Learning (DL)?

Deep Learning is a subfield of ML and AI that uses multi-layered neural networks to learn complex patterns from large amounts of data.

What does the acronym MNIST stand for?

Modified National Institute of Standards and Technology.

What are the three main types of Machine Learning?

Supervised learning (learning from labeled data), Unsupervised learning (learning from unlabeled data), and Reinforcement learning (learning step-by-step through trial and error using rewards and penalties).

In supervised learning, what is the key distinction between classification and regression?

Classification is used for predicting categories (such as cat or dog), while regression is used for predicting numbers (such as house prices or traffic accidents).

What is the Turing Test and who proposed it?

Proposed by Alan Turing, the Turing Test evaluates whether a computer exhibits human-like intelligence by having an interrogator chat with a human and a computer; if the interrogator cannot tell which is which, the computer passes.

Who coined the term "Artificial Intelligence" and at what event?

John McCarthy coined the term "Artificial Intelligence" at the Dartmouth conference in 1956.

What is the difference between Strong AI and Weak AI?

Strong AI aims for true intelligence and self-consciousness and does not exist in the real world, whereas Weak AI refers to systems that behave intelligently but are just computers (existing in today's world).

How are State Space, Transitions, and Costs defined in AI problem solving?

State Space is the set of all possible situations, Transitions are single-step moves from one state to another, and Costs represent how transitions differ in preference, time, or distance.

How does the Minimax algorithm evaluate decisions in game trees for two-player games?

Max tries to maximize the score ($+1$ for a Max win), Min tries to minimize the score ($-1$ for a Min win, $0$ for a draw), and values are backed up from the leaf nodes to the root to determine the game outcome under optimal play.

What is the relationship between Odds and Probability?

Odds express uncertainty as a ratio of occurrences to non-occurrences (e.g., $3:1$ means $3$ wins for every $1$ loss), whereas probability represents occurrences divided by total cases (e.g., $\frac{3}{3+1} = 0.75$ or $75\text{\%}$).

In Bayes Rule, what is the distinction between Prior Odds and Posterior Odds?

Prior odds represent the initial belief before receiving new information, whereas posterior odds represent the updated estimate after new evidence or observation is considered.

What key assumption does a Naive Bayes Classifier make?

It assumes that features are conditionally independent within each class.

What are the biological parts of a neuron mentioned in the text and their functions?

The cell body, dendrites (wires providing input signals), axon (wire transmitting outgoing signals), and synapses (intersections connecting axons to dendrites).

What two key architectural differences separate artificial neural networks from traditional CPUs?

1. Parallel processing (many neurons process information simultaneously). 2. Integrated memory and processing (data is stored and processed together using weights, unlike traditional CPUs where CPU and memory are separate).

What is a Perceptron and who introduced it?

Introduced in 1957 by psychologist Frank Rosenblatt, a Perceptron is a simple artificial neuron model that uses a step activation function for binary classification and is often referred to as the mother of all artificial neural networks.

What are Convolutional Neural Networks (CNNs) and why are they used?

CNNs are deep learning models designed for image processing that reuse weights across different parts of an image to automatically detect features (such as edges and patterns) regardless of position, size, or orientation.

How do Large Language Models (LLMs) operate?

LLMs are based on the Transformer neural network architecture using attention mechanisms, and they learn language patterns by predicting the next word in a sentence during training on massive text datasets.

What major societal risks of AI are highlighted in the transcript?

Algorithmic bias (perpetuating human biases in training data), fake media and loss of trust (realistic synthetic video/audio), and privacy risks (de-anonymization of anonymous data patterns).