CogSci 1 — L5: AI & Machine Learning

What is Artificial Intelligence (AI)?

Computer models designed to accomplish the same cognitive tasks humans do.

What is Machine Learning (ML)?

A subset of AI that learns from data to progressively improve performance on a task.

What is Deep Learning?

A subset of ML involving numerous layers of algorithms without requiring structured data.

What are neural networks?

Networks of algorithms similar to neural networks in the human brain.

What is Strong AI?

General-purpose AI with artificial general intelligence (AGI).

What would AGI be capable of?

Understanding and learning any task a human can across domains.

What is Applied AI?

AI designed to master specific complex tasks rather than general human-level intelligence.

What is an expert system?

A program designed to replicate or improve human expert performance in a specialized domain.

How can expert systems acquire knowledge?

From human experts or by using ML to derive rules from databases.

What was MYCIN?

A 1970s expert system using heuristic rules to diagnose infectious diseases.

What is a decision tree?

A branching structure that uses attributes and rules to classify or make decisions.

What is ID3?

A machine-learning algorithm that constructs decision trees.

What does ID3 use to choose an attribute for a decision-tree node?

Information gain.

What is information gain in ID3?

A statistical measure of how well an attribute organizes examples.

Which attribute does ID3 place at the first node?

The attribute with the highest information gain.

What is a major limitation of ID3-style machine learning?

It depends heavily on how its data are labeled and categorized.

Why can traditional ML struggle with raw data?

The data often must already be interpreted into meaningful attributes.

What does CAPTCHA stand for?

Completely Automated Public Turing Test to Keep Computers and Humans Apart.

What was the original logic behind CAPTCHA?

Humans could interpret distorted text that computers could not.

What happened as computer vision improved?

Computers became better than humans at solving increasingly complex CAPTCHAs.

What is generative AI?

AI that learns patterns in training data and uses them to generate new content.

What types of content can generative AI create?

Text, images, audio, code, and other modalities.

What is a Large Language Model (LLM)?

A type of generative AI pretrained on large datasets and then fine-tuned for specific purposes.

What type of AI model is ChatGPT?

A Large Language Model (LLM).

What does a transformer do in ChatGPT?

Uses relationships among words to understand context and predict the most probable next word.