womsactivity2

Inside the Black Box of AI - a diagramming activity

Devised by Dean Kenning

Learning Objectives:

To explore, via imagination, association and reasoning, the hidden mechanics of AI.

To explore diagraming as a drawing and thinking tool.

Approximate duration:

45 Minutes

Equipment / Resources needed:

For teacher: a whiteboard or OHP projector (where possible)

For students: A3 white boards and marker pens, or, paper and marker pens/pencils/charcoal.

Note: 

    This activity is carried out using analogue equipment, i.e. without computers.

Note 2: 

    A black box is a system which can be viewed in terms of its inputs and outputs without any knowledge of its internal workings. Something goes into the 'box' (e.g. a text prompt) and something comes out (e.g. an AI generated image), but we don't see, and don't know, exactly what happens in between.

Activity:

1. Introduction to the ‘black box’ concept.

The teacher draws a 'black box' (a square or cube). Left of the box draw an arrow going into the box and mark it 'input'; right of the box draw an arrow coming out of the box and mark it 'output'. In order to get the students to understand the black box concept, and to warm them up for thinking about how generative AI works, we will present cases pertaining to COPYING an image and GENERATING (or creating) an image.

In what follows, as the students suggest answers the teacher will write those suggestions down, placing them inside the box or cube he/she has drawn. Rub out and start again for each of the examples given below.

COPYING – an image (e.g. the Goya etching) is the input and the copy is the output. Ask the students what they think happens in between (i.e. inside the 'black box'):

    Example i: An old fashioned analogue photocopier [what happens inside the 'black box' of the photocopier so that when you put in an image, a copy of that image comes out?]

    Example ii: a digital copy of an image, using a digital camera/scanner/screenshot/copying a jpeg etc. 

    Example iii: a human is given an image and asked to make a drawn copy of it [what happens inside the 'black box' of the person to allow them to produce a drawn copy?]

GENERATING - an image (e.g. the Goya etching) from a verbal description/text prompt:

Example i: a human makes a drawing from a text prompt [what is happening inside the 'black box' of the person to allow them to create a unique image from the description?]

Example i: a generative AI (such as Craiyon or DALL-E) creates an image from a text prompt

This final example is not drawn by the teacher but is the question put to the students to answer in their own diagrams (see next section below).

2. Diagramming the black box of generative AI

Participants are asked to make a diagram (on A3 white board or paper) describing what they imagine might be happening inside the black box of text-to-image AI (e.g. Craiyon) that enables it to generate an image from a prompt. They are told this is not about making a ‘correct’ diagram; rather it is an exploratory exercise, designed to make the participant think about the various factors which might be involved. 

They are asked to bear in mind:

a. the technology

b. human/social factors

(and the relations between the two).

After their diagrams are complete, students are asked to share their diagram with the group and discuss how they think the AI generates images from a prompt, i.e. what is happening inside the black box of generative text-to-image AI.