Skip to content
BENIBIX Learn

Foundations

The core ideas everything else builds on · 30 terms

Sort
Compare

AI Basics

7

What AI is, the model at its heart, and how it first learns and then answers.

Computers doing jobs that once took a person. Reading and talking are two.

The learned file at the heart of an AI. Input goes in, an answer comes out.

The learning stage of AI. A model practises on examples until its guesses get good.

Using a trained model to get an answer. It does not learn while it works.

The answer a model gives for one input. It is its best guess, not a fact.

The computer power used to train and run AI. More power can make a bigger model.

Finding what repeats in data. Then the same thing can be spotted in new data.

Kinds of AI

6

AI sorted by how much it can do, and by whether it makes things or sorts them.

AI that is good at one kind of job and cannot do much else.

A future AI that could learn any thinking job a person can. It is not here yet.

A future AI much smarter than the best people. It would win at nearly every task.

AI that makes new things, like text, pictures, songs or code, when you ask.

AI that sorts or guesses things, like spotting spam or next week's sales.

AI that takes steps and uses tools to finish a task, not just answer.

AI Approaches

5

Ways to build AI that do not learn from piles of data. Some use rules, and some copy evolution.

A program that makes choices with if-then rules. People write all the rules.

A program built from an expert's rules. It asks you questions and gives advice.

AI built from facts and logic that people can read. It does not learn patterns.

A rule of thumb. It finds a good answer fast, but not always the best one.

A search that copies evolution. It keeps the best answers, mixes them and repeats.

AI History

9

The tests, meetings and slumps that shaped AI.

A test where a judge chats by text and tries to spot which side is a machine.

The 1956 summer meeting that started AI as a field and gave it its name.

A time when money for AI dried up. It came after big promises failed.

A thought test. It says that following rules for a language is not truly knowing it.

AI Milestones

5

The machines and contests that changed what people thought AI could do.

An early machine from 1958. It learned to sort things by changing its weights.

One of the first chatbots, from 1966. It matched key words to set replies.

A chess computer from IBM. It beat the world champ, Garry Kasparov, in 1997.

A huge set of labelled photos. A contest on it set off the deep learning boom.

DeepMind's Go program that beat top player Lee Sedol 4 to 1 in 2016.

AI Debates

3

Big ideas about AI that experts still argue about, such as where it is going.

The idea that AI may one day make itself smarter. Each step comes faster.

Hard for us is often easy for a computer. Easy for us is often hard.

Once a computer can do a job, we say it is not AI.