Dr Julia Molinari · @serenissimaj
219 followers · 74 posts · Server sciences.social

ntm for research on :

"the [AI] training process is highly dependent on prior labelling performed by a human operator. This is problematic because it crystallises the operator’s cultural biases (e.g. racism and sexism) as well as her possible mistakes".

What other biases are being built in to for ?




DOI: 10.1080/14767430.2022.2134618

#acwri #storymachines #academicwriting #academicalgorithms #whatmakesacademicwritinghuman #WhatMakesWritingAcademic #CriticalRealism #artificialintelligence

Last updated 2 years ago