What is Artificial Intelligence?

Although there’s no universally accepted definition of “artificial intelligence” in computing, the term is commonly used to describe one or more of the following:

an area of research within computer science
a property that some researchers claim (controversially) belongs or could belong to certain computer systems
a set of technologies for generating new content (text, images, sound, etc.) from existing content

The last item on that list has given rise to the term “generative artificial intelligence” or, for short, “GenAI.” While GenAI isn’t new, it suddenly seems to be everywhere: not only in free-standing tools with names such as ChatGPT, Claude, Gemini, Copilot, DALL-E, and Midjourney, but also, increasingly, as an affordance built into other tools, such as search engines, word-processors, image editors, email and messaging interfaces, mobile apps, coding environments, websites, cars, and home appliances.

In this module, we’ll make some use of GenAI and consider some of the ethical concerns currently being raised around its impact on society and the environment.

But first, it will be helpful to put GenAI and artificial intelligence (AI) as a whole in some historical perspective. Data scientist William J.B. Mattingly’s series of videos on AI and machine learning is a good place to start. Watch the first video in the series, below, to get a brief introduction to how artificial intelligence fits into the larger history of computing.

A few key terms

The definitions below have been adapted from José Antonio Bowen and C. Edward Watson, Teaching with AI: A Practical Guide to a New Era of Human Learning (John Hopkins University Press, 2024) and Intro to AI, a newsletter distributed by MIT Technology Review.

Artificial Intelligence (AI): Depending on context, this term usually refers to an area of research within computer science, a property that some researchers claim (controversially) belongs or could belong to certain computer systems, or a set of technologies for generating new content (text, images, sound, etc.) This isn’t an exhaustive list.
Expert Systems: rules and logic programmed into a computer to anticipate a wide range of possible scenarios.
Machine Learning: The use of probability and statistics to find patterns in huge quantities of data and generalize from these. Recommendation algorithms such as those used by Netflix, YouTube, and Spotify rely on machine learning, as do search engines, social media feeds, and voice assistants like Siri and Alexa.
Deep Learning: Super-powered machine learning that uses massive datasets and large neural networks, and that requires a great deal of computing power (sometimes called “compute.”)
Neural Networks: Computing systems designed to emulate the neural connections in the human brain. The “neurons” in the network are connected by complex mathematical equations. When a new piece of data (a new image, for example) is passed through a neural network, it through the layers of the network and outputs a result. A trained neural network, aka a model, has learned through trial and error to reproduce patterns in its training data and produce (hopefully) correct answers.
Foundational Models: Neural networks trained with a large data set using machine learning techniques that mimic human trial and error.
Large Language Models (LLMs): Foundational models focused on language. Built on deep-learning algorithms that are trained on enormous quantities of text, they’re able to predict which words are most likely to appear together in a sentence or paragraph and generate textual content that sounds or reads like what a person might say or write.
GPT: Generative Pre-trained Transformers. Foundational models and LLMs all use GPT architecture.
Parameter: an internal variable in a neural network that can be tuned or adjusted to change the output.
Diffusion Model A type of foundational model used to create images and video.
Alignment: The work that humans do to ensure that an AI system does only what users want it to do.