“How can I use AI prompts as a fun way to promote impromptu discussions?”

You’ve probably practised getting students to write prompts for ChatGPT in order to generate, correct, shorten or summarize a text. But have you tried getting them to write prompts to create artwork? This activity is a lot more fun  – though also very demanding.

Task

Brainstorm the whole class on a particular image they would like to create. The image could be anything – a logo for the school, a fun view of an English classroom, a sports activity, something surreal, a scene from a film, a superhero.

Then in small groups, students work on generating the image.

The challenge is to get the AI art generator to do what it’s told! Getting ChatGPT to do something with a text is not that hard – you probably need a maximum of three attempts. But generating pictures requires much more detailed instructions.

Most art AI sites don’t let you experiment much for free. So students will probably have to move from site to site, or use the same site from different  devices.

Level and feedback

This activity works best with B2+ levels. It generates a lot of talking, with students discussing how to approach the problem (possibly doing a search on ‘how to write prompts for AI artwork’), what style they are looking for and how to describe what they want to depict, what to do when their prompt fails, making suggestions about moving to another site, etc.

At the end of the activity you can have a vote on which group got the best image. You could also get one person from each group to do a mini-talk on what approach they used, how successful they think they have been, and what they have learned in the process.

Variation: If the groups chose different images, each group could also ‘reverse engineer’ another group’s image, i.e. try and work out what prompts they used to achieve the end result.

Benefit: simulating a real-life task

The great thing is that such activities simulate a real-life task … with all the fun and frustrations that that involves.

Below is an example of the lengths designers / artists will go to get the right image. 

Tiny cute boy holding camera toy, standing character, soft smooth lighting, soft pastel colors, skottie young, 3d blender render, polycount, modular constructivism, pop surrealism, physically based rendering, square image

This prompt contains 15 pieces of information! Don’t worry you can get away with far fewer! Most Chatbots only work well with about three pieces of info, and even then you may have to do a bit of coaxing to get the result you were after.

Adrian Wallwork

Categories: AI & Digital