Exploring Bias In The System – Kevin’s Meandering Thoughts

On this tenth day of the 12 Days of AI, we’re exploring the bias that comes inside AI techniques, both deliberately or unintentionally. All techniques have inherent bias due to the info units that work beneath the interface. Some AI websites do a greater job at countering bias than others.

The put up reminds us:

“AI generative output shouldn’t be creativity, however a statistical variation with out intent and which means. AI is a statistical artist. It introduces variations into its work, however its “selections” are pushed on patterns and chances quite than deep understanding of lived experiences and feelings. AI generative outputs, whereas devoid of the intent and which means attribute of human creativity, possess important potential to hurry, help, and prolong the inventive course of.”

The immediate was to enter an AI picture generator, and use a immediate of an individual in a setting, doing a little motion (they body it as a scene from a film) and look to see if there have been any obvious bias within the outcomes. I used Adobe Firefly and requested it to generate a “stern-looking faculty principal speaking to an elementary instructor about pupil work.”

Is there bias within the two collections of outcomes I obtained?

Out of the eight pictures, it appears as if 5 (and possibly six? Relies upon the way you have a look at who’s who) present a white male within the position of principal, and of the eight, solely two pictures present non-white adults being generated. Two girls appear to be within the administrative position. I didn’t have sufficient credit in Firefly to maintain working the experiment just a few extra occasions, which might have been superb. Would I’ve stored discovering comparable genders and races of the directors?

What about if I refined my immediate, including loaded phrases like “city faculty” or “struggling instructor” or “disruptive pupil”? How would different platforms carry out with the identical immediate? Would there be some gender/racial stereotypes into who’s an administrator and who’s an educator?

Peace (in explorations),
Kevin