Section 1.1: Key AI Concepts

Artificial Intelligence (AI)

AI is a broad term that encompasses machines designed to replicate intelligent human behavior.

Machine Learning

This is a specific area within AI that emphasizes the capacity of computers to learn tasks without direct programming. Essentially, a machine is fed vast amounts of data, allowing it to determine how to accomplish specific objectives (for instance, identifying images containing cats).

Neural Networks

Neural networks are algorithms that resemble the workings of the human brain. They consist of multiple layers of nodes (akin to neurons) interconnected through links (similar to synapses). The first layer processes incoming data, while the final layer produces the output. The intermediate layers perform additional computations, though their internal workings often remain opaque. This architecture facilitates deep learning, enabling the identification of intricate patterns within extensive data sets, leveraging millions or billions of parameters.

Large Language Models

These are advanced deep learning models—often built on neural network frameworks using transformer architecture—that have gained significant media attention. Examples include ChatGPT. They are trained on vast text corpuses, allowing them to predict subsequent words in a sentence or the next sentence in a paragraph. The progress made in natural language processing through these models has surpassed expectations, introducing remarkable emergent behaviors where systems exhibit unexpected abilities.

Explore the vocabulary needed for effective English conversation about AI in this insightful video, "Vocabulary for English Conversation on Artificial Intelligence (AI)."

Section 1.2: Generative AI and Related Terms

Generative AI

This category of AI has the capability to produce original content after being trained on extensive datasets. A prime example is a large language model, which can generate new text. This concept extends to models that create images, videos, and even computer code. The goal is to harness generative AI for constructive purposes rather than adverse outcomes.

Hallucination

This term describes the situation where a machine learning model produces inaccurate or nonsensical outputs, often due to shortcomings in its training data. Interestingly, humans—particularly in political contexts—can experience a similar phenomenon.