Free lesson · Module 1, lesson 2 of 6

Generative AI: machines that make new things

Tell recognising from generating (reading a number plate versus writing a letter), name the four things generative AI makes (text, images, audio, code), and say what a large language model is in one sentence.

The video for this lesson is being prepared

About 9 minutes, narrated in English, Hindi and Odia. The notes and the quiz below are open now.

Lesson notes

For the student

Some AI reads what is already there. Generative AI makes something new. One question tells them apart.

Recognising versus generating

  • Recognising: reading, naming, sorting or checking something that is already there. A parking camera reads a number plate; voice typing writes down the words you said; a spam filter moves a message to the spam folder
  • Generating: making something that was not there before. A chat assistant writes a leave letter from three points
  • Both kinds learned patterns from examples (the last lesson); they use what they learned in different ways
  • AI that generates is called generative AI

The test

  • Ask: was the output already in what went in?
  • Yes (it read, named, sorted or checked it): recognising
  • No (new text, an image, audio or code): generating
  • Judge the output, not the app's name or a word like "made": an album of your own photos is sorting, not generating
  • Many apps do both: a chat assistant reads your question, then writes a new reply. Ask the question of each step

The four things generative AI makes

  • Text: three points in, a full leave letter out
  • Images: one sentence in (a kite over Puri beach at sunset), a new picture out
  • Audio: one line in (a cheerful tune for the morning assembly), new music out
  • Code: a request in plain words, a few lines of code out (instructions a computer can follow)

A large language model in one sentence

  • A large language model is a program that learned patterns from a huge amount of text, and uses them to write new text
  • Language: it works with words. Large: it is very big, and learned from a very large amount of text
  • It is the text maker behind most chat assistants
  • It usually writes new sentences from patterns it learned; it is not a store of answers copied out. Later lessons show how it learns

The trap

  • Wrong: text came out, so it was generated; every AI app makes new things; a language model copies stored answers
  • Right: ask whether the output was already in the input; reading, naming and sorting are recognising; new text, images, audio or code are generating

Try it

  • In your notes file, list five AI features you use on your phone (for example voice typing, face unlock, the spam folder, photo search, a chat assistant)
  • Next to each, write recognising or generating, and the one-line reason: what went in, and what came out
  • Open a chat assistant and ask: "Write a four-line poem about the rain in Bhubaneswar." Read the reply and ask yourself the test: were those lines already in what you typed?
  • Write the one-sentence description of a large language model in your own words, and keep it in your notes file

Lesson quiz

Checked in your browser. Nothing is sent or saved.

5 questions

Question 1 · easy

Asha is a clerk at a bank in Sambalpur. Four AI features help her today. Which one is generating?

Question 2 · easy

In the name large language model, what does the word language tell you?

Question 3 · medium

Nisha takes a photo of a handwritten Odia notice and gives it to a chat assistant. The assistant first reads the words in the photo, then writes a new three-line summary in English. Which statement is right?

Question 4 · medium

A school app called Answer Maker reads a student's handwritten answer sheet and puts a tick or a cross next to each answer. Is this feature recognising or generating?

Question 5 · hard

Meena, a teacher in Cuttack, tries four features on one chapter of a geography book. Which one is generating?

What AI is, and what it is notNext: Your first conversation with a chat assistant