Where are the crescents in AI?
Maha Bali,
LSE Higher Education,
Feb 27, 2024
One of Helen Beetham's posts yesterday asked for an alternative word for 'hallucinate', one that did not imply cognition. Maha Bali provides the answer in this column: "GenAI performs better with topics familiar to its training data, and more likely 'hallucinates' or produces 'botshit' (if we use it uncritically) with material less familiar to it." I find it interesting to have observed through my lifetime the gradual migration of what we think of as "offensive" moving away from human bodily functions (which we all know are natural and normal) and toward language that degenerates people based on gender, class or race (which as we all know is base and prejudicial), a move that parallels a similar migration of the term 'critical' from addressing matters of logic to addressing matters of social justice (which Bali discusses in this post).
Today: 1 Total: 1457 [Direct link] [Share]
