AI Is a Branch of Computer Science

This section explains where artificial intelligence fits inside computer science and how a model is different from an ordinary program.

Where AI Fits in Computer Science

Computer science is the study of computers and the step-by-step processes they carry out. Networking, data management and security are three branches of computer science, and artificial intelligence, usually called AI, is another. The name "artificial intelligence" comes from the 1950s, when four researchers proposed a summer study of the topic at Dartmouth College. Machine learning is one area inside AI, and the language models that write text belong to machine learning. The models behind the class tools MuggsOfPrompts and MuggsOfVibes are language models of this kind. Figure 1 shows these areas as boxes drawn inside one another.

Programs: Rules Written by People

A program is a set of instructions that a person wrote for a computer to follow. For example, a calculator app is a program. When a user enters 2 + 2, the app follows an addition rule that a programmer wrote, so the same input gives the same output every time. A model is also a program in the broad sense that it is software that runs on a computer, but nobody typed its rule. The calculator fits this description of a program, while the face-recognition feature described below does not.

How Models Are Made by Training

A model is a piece of software whose behavior came from examples instead of from rules a person wrote. Inside, a model is a large collection of numbers that connect an input to an output. Training is the process of showing a model a very large number of examples and adjusting its numbers until its output matches them. For each example, the model makes a prediction, the prediction is compared to the correct answer, which is called a label, and the model's numbers are adjusted to bring the next prediction closer. A person typing in a rule is not training, because the rule came from the person and not from examples. A language model is trained on a very large amount of text, and it then generates text one word or part of a word at a time. A chatbot is one use of a language model, which generates each reply as a continuation of the conversation so far.

One Phone, Two Kinds of Software

The calculator app runs on the phone's chips and is not a model, because every one of its rules was typed by a person. The feature that unlocks the phone by recognizing its owner's face works differently, because it relies on a model that was made by training. No person wrote a rule that lists what the owner's face looks like. Some features have been built both ways. For example, an earlier design of one phone keyboard described typing mistakes with rules, and a newer design replaced those rules with a trained model, so the answer to the test depends on how the feature was built.

Summary

AI is one branch of computer science, and machine learning and language models are areas inside it, as Figure 1 shows. A program follows rules that a person wrote, while a model gets its behavior from training on examples. The test for any computer feature is whether a person wrote the rule or whether the rule was learned from examples.

Sources: McCarthy et al., A Proposal for the Dartmouth Summer Research Project on Artificial Intelligence (1955); Tucker et al. via K–12 CS Framework, Defining Computer Science; ACM, IEEE-CS and AAAI, Computer Science Curricula 2023 (2024); Russell and Norvig, Artificial Intelligence: A Modern Approach, contents; CompTIA, Tech+ (FC0-U71); Samuel, Some Studies in Machine Learning Using the Game of Checkers (1959); Google, What is Machine Learning? and Supervised Learning; Google, LLMs: What's a large language model?; Microsoft, Windows Calculator; Beaufays and Riley, The Machine Intelligence Behind Gboard (2017); Apple, About Face ID advanced technology and Biometric security; Mistral AI, Models Overview; course repository, MuggsOfPrompts and MuggsOfVibes model ids.