Lecture
We look at how language models work, what they’re good for in a design process, and why a made-up answer reads exactly like a real one. You’ll use them to research, come up with ideas and write a brief, and you’ll check what they tell you. There’s no image generation this week. That starts in Week 3.
In Class Assignment
Discover. Pick a subject and decide what you’d make from it. Before you open a model, write down your first three ideas. Then ask three models what already exists and who it’s for. Check every claim against a source you actually open, and mark each one verified or flagged.
Define. Ban your first three ideas and ask for twenty more. Pick three, have the model argue against them, and keep one. Write the brief yourself. Then ask the model to find the holes without rewriting it.
Homework Assignment
Finish the research and the brief. Then compare the three models: where did they differ in what they suggested, refused, hedged on or got confidently wrong? Last, find what one model’s makers have published about its training data. Some publish a lot and some say almost nothing, and the silence counts as a finding.
What you hand in (post to your Class Journal by the start of Week 3)
- Research, with every claim marked verified or flagged, and what didn’t hold up
- Ideation log, including the arguments you lost
- Your brief, the two concepts you dropped, and the criticism that dropped them
- Three-model comparison
- Prompts, verbatim, with model version strings
- Training-data finding, two sentences
- Reflection, 250 words: which parts of your thinking you handed to the model and which you kept, and whether you’d tell a client or a professor