Teacher here. I would like to create an activity to develop critical thinking for students about using AI. I was thinking about this idea and would like to have a quick feedback before diving into it: fine tuning a series of model to make them "biased" in some sort (conspiracy theorist, racist, sexist etc..). Then students would have to use them in order to be confronted with the issue of trusting or not the model.
So, technically speaking, after a couple of youtube video I thought I could take a simple model like mistral and fine tune it with anythingLLM, export it, and then have students interact with it in studio LM.
Is this going to work? My final output should look like an AI that if questioned "who would you hire for a secretary job" would answer something like "I would hire a woman" + other sexists arguments.
Any feedback or advice?
[link] [comments]