3 ms·
Developing a sense of what you know or do not know is a big part of learning. In ML, this is called confidence estimation and is harder than actually learning t
by visarga 7y ago
Developing a sense of what you know or do not know is a big part of learning. In ML, this is called confidence estimation and is harder than actually learning the task. Most ML models just learn to solve the task and have shit confidence estimations. I presume with humans it takes many years of learning a subject to really know how much you do not know. It's a process of calibration of confidence based on lots of background knowledge on the topic (going from unknown unknowns to known unknowns).