The AI might not have it

Marnie Hughes Warrington is testing ideas on the complex ground where trust in data intersects with ambiguity about pasts, real and imagined

It’s work-in-progress for a bigger project on data, history and ideas about both. But while the past is clearly many countries where they did different things, differently, depending on the sources you use, she is yet to be convinced that the AIs do have it.

“Uncertainty, incompleteness, vagueness and the unknown are the titans to be defeated in one heroic story we tell about our computational world. Or its mischief managed via probability, Bayesian refinement, error margins or a fuzzy logic in which possibilities are located on a numerable continuum of grades of membership between >0 and <1. On this logic, if we can code likelihood formally and narrow in, we can recommend a world of nuance rather than a world of certainties that are often wrong. It works well in many contexts, but not in a world in which individuals and groups ought to be afforded dignity, fairness and safety when decisions are made about them, predictions made about them, or recommendations put to them.”

Good-oh, but what CMM wants to know is why people who have trouble with a coffee cup in Game of Thrones but are fine with dragons


to get daily updates on what's happening in the world of Australian Higher Education